ruby-fann 0.7.1
Sign up to get free protection for your applications and to get access to all the features.
- data/History.txt +4 -0
- data/License.txt +20 -0
- data/Manifest.txt +28 -0
- data/README.txt +1 -0
- data/Rakefile +4 -0
- data/config/hoe.rb +71 -0
- data/config/requirements.rb +17 -0
- data/ext/ruby_fann/extconf.rb +5 -0
- data/ext/ruby_fann/fann_augment.h +110 -0
- data/ext/ruby_fann/neural_network.c +1499 -0
- data/lib/ruby_fann.rb +5 -0
- data/lib/ruby_fann/neurotica.rb +102 -0
- data/lib/ruby_fann/version.rb +9 -0
- data/log/debug.log +0 -0
- data/script/destroy +14 -0
- data/script/generate +14 -0
- data/script/txt2html +74 -0
- data/setup.rb +1585 -0
- data/tasks/deployment.rake +34 -0
- data/tasks/environment.rake +7 -0
- data/tasks/website.rake +17 -0
- data/test/test_helper.rb +2 -0
- data/test/test_neurotica.rb +36 -0
- data/test/test_ruby_fann.rb +390 -0
- data/test/test_ruby_fann_functional.rb +88 -0
- data/website/index.html +90 -0
- data/website/index.txt +35 -0
- data/website/javascripts/rounded_corners_lite.inc.js +285 -0
- data/website/stylesheets/screen.css +138 -0
- data/website/template.rhtml +48 -0
- metadata +82 -0
data/History.txt
ADDED
data/License.txt
ADDED
@@ -0,0 +1,20 @@
|
|
1
|
+
Copyright (c) 2007 FIXME full name
|
2
|
+
|
3
|
+
Permission is hereby granted, free of charge, to any person obtaining
|
4
|
+
a copy of this software and associated documentation files (the
|
5
|
+
"Software"), to deal in the Software without restriction, including
|
6
|
+
without limitation the rights to use, copy, modify, merge, publish,
|
7
|
+
distribute, sublicense, and/or sell copies of the Software, and to
|
8
|
+
permit persons to whom the Software is furnished to do so, subject to
|
9
|
+
the following conditions:
|
10
|
+
|
11
|
+
The above copyright notice and this permission notice shall be
|
12
|
+
included in all copies or substantial portions of the Software.
|
13
|
+
|
14
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
15
|
+
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
16
|
+
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
17
|
+
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
|
18
|
+
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
|
19
|
+
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
|
20
|
+
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
data/Manifest.txt
ADDED
@@ -0,0 +1,28 @@
|
|
1
|
+
History.txt
|
2
|
+
License.txt
|
3
|
+
Manifest.txt
|
4
|
+
README.txt
|
5
|
+
Rakefile
|
6
|
+
config/hoe.rb
|
7
|
+
config/requirements.rb
|
8
|
+
ext/ruby_fann/extconf.rb
|
9
|
+
ext/ruby_fann/fann_augment.h
|
10
|
+
ext/ruby_fann/neural_network.c
|
11
|
+
lib/ruby_fann.rb
|
12
|
+
lib/ruby_fann/version.rb
|
13
|
+
lib/ruby_fann/neurotica.rb
|
14
|
+
log/debug.log
|
15
|
+
script/destroy
|
16
|
+
script/generate
|
17
|
+
script/txt2html
|
18
|
+
setup.rb
|
19
|
+
tasks/deployment.rake
|
20
|
+
tasks/environment.rake
|
21
|
+
tasks/website.rake
|
22
|
+
test/test_helper.rb
|
23
|
+
test/test_ruby_fann.rb
|
24
|
+
website/index.html
|
25
|
+
website/index.txt
|
26
|
+
website/javascripts/rounded_corners_lite.inc.js
|
27
|
+
website/stylesheets/screen.css
|
28
|
+
website/template.rhtml
|
data/README.txt
ADDED
@@ -0,0 +1 @@
|
|
1
|
+
README
|
data/Rakefile
ADDED
data/config/hoe.rb
ADDED
@@ -0,0 +1,71 @@
|
|
1
|
+
require 'ruby_fann/version'
|
2
|
+
|
3
|
+
AUTHOR = 'Steven Miers' # can also be an array of Authors
|
4
|
+
EMAIL = "steven@7bpeople.com"
|
5
|
+
DESCRIPTION = "Bindings to use FANN from within ruby/rails environment."
|
6
|
+
GEM_NAME = 'ruby-fann' # what ppl will type to install your gem
|
7
|
+
RUBYFORGE_PROJECT = 'ruby-fann' # The unix name for your project
|
8
|
+
HOMEPATH = "http://#{RUBYFORGE_PROJECT}.rubyforge.org"
|
9
|
+
DOWNLOAD_PATH = "http://rubyforge.org/projects/#{RUBYFORGE_PROJECT}"
|
10
|
+
|
11
|
+
@config_file = "~/.rubyforge/user-config.yml"
|
12
|
+
@config = nil
|
13
|
+
RUBYFORGE_USERNAME = "unknown"
|
14
|
+
def rubyforge_username
|
15
|
+
unless @config
|
16
|
+
begin
|
17
|
+
@config = YAML.load(File.read(File.expand_path(@config_file)))
|
18
|
+
rescue
|
19
|
+
puts <<-EOS
|
20
|
+
ERROR: No rubyforge config file found: #{@config_file}
|
21
|
+
Run 'rubyforge setup' to prepare your env for access to Rubyforge
|
22
|
+
- See http://newgem.rubyforge.org/rubyforge.html for more details
|
23
|
+
EOS
|
24
|
+
exit
|
25
|
+
end
|
26
|
+
end
|
27
|
+
RUBYFORGE_USERNAME.replace @config["username"]
|
28
|
+
end
|
29
|
+
|
30
|
+
|
31
|
+
REV = nil
|
32
|
+
# UNCOMMENT IF REQUIRED:
|
33
|
+
# REV = `svn info`.each {|line| if line =~ /^Revision:/ then k,v = line.split(': '); break v.chomp; else next; end} rescue nil
|
34
|
+
VERS = RubyFann::VERSION::STRING + (REV ? ".#{REV}" : "")
|
35
|
+
RDOC_OPTS = ['--quiet', '--title', 'ruby_fann documentation',
|
36
|
+
"--opname", "index.html",
|
37
|
+
"--line-numbers",
|
38
|
+
"--main", "README",
|
39
|
+
"--inline-source"]
|
40
|
+
|
41
|
+
class Hoe
|
42
|
+
def extra_deps
|
43
|
+
@extra_deps.reject! { |x| Array(x).first == 'hoe' }
|
44
|
+
@extra_deps
|
45
|
+
end
|
46
|
+
end
|
47
|
+
|
48
|
+
# Generate all the Rake tasks
|
49
|
+
# Run 'rake -T' to see list of generated tasks (from gem root directory)
|
50
|
+
hoe = Hoe.new(GEM_NAME, VERS) do |p|
|
51
|
+
p.author = AUTHOR
|
52
|
+
p.description = DESCRIPTION
|
53
|
+
p.email = EMAIL
|
54
|
+
p.summary = DESCRIPTION
|
55
|
+
p.url = HOMEPATH
|
56
|
+
p.rubyforge_name = RUBYFORGE_PROJECT if RUBYFORGE_PROJECT
|
57
|
+
p.test_globs = ["test/**/test_*.rb"]
|
58
|
+
p.clean_globs |= ['**/.*.sw?', '*.gem', '.config', '**/.DS_Store'] #An array of file patterns to delete on clean.
|
59
|
+
|
60
|
+
# == Optional
|
61
|
+
p.changes = p.paragraphs_of("History.txt", 0..1).join("\n\n")
|
62
|
+
#p.extra_deps = [] # An array of rubygem dependencies [name, version], e.g. [ ['active_support', '>= 1.3.1'] ]
|
63
|
+
|
64
|
+
#p.spec_extras = {} # A hash of extra values to set in the gemspec.
|
65
|
+
|
66
|
+
end
|
67
|
+
|
68
|
+
CHANGES = hoe.paragraphs_of('History.txt', 0..1).join("\\n\\n")
|
69
|
+
PATH = (RUBYFORGE_PROJECT == GEM_NAME) ? RUBYFORGE_PROJECT : "#{RUBYFORGE_PROJECT}/#{GEM_NAME}"
|
70
|
+
hoe.remote_rdoc_dir = File.join(PATH.gsub(/^#{RUBYFORGE_PROJECT}\/?/,''), 'rdoc')
|
71
|
+
hoe.rsync_args = '-av --delete --ignore-errors'
|
@@ -0,0 +1,17 @@
|
|
1
|
+
require 'fileutils'
|
2
|
+
include FileUtils
|
3
|
+
|
4
|
+
require 'rubygems'
|
5
|
+
%w[rake hoe newgem rubigen].each do |req_gem|
|
6
|
+
begin
|
7
|
+
require req_gem
|
8
|
+
rescue LoadError
|
9
|
+
puts "This Rakefile requires the '#{req_gem}' RubyGem."
|
10
|
+
puts "Installation: gem install #{req_gem} -y"
|
11
|
+
exit
|
12
|
+
end
|
13
|
+
end
|
14
|
+
|
15
|
+
$:.unshift(File.join(File.dirname(__FILE__), %w[.. lib]))
|
16
|
+
|
17
|
+
require 'ruby_fann'
|
@@ -0,0 +1,110 @@
|
|
1
|
+
|
2
|
+
FANN_EXTERNAL struct fann_train_data * FANN_API fann_create_train_from_rb_ary2(
|
3
|
+
unsigned int num_data,
|
4
|
+
unsigned int num_input,
|
5
|
+
unsigned int num_output)
|
6
|
+
{
|
7
|
+
return 0;
|
8
|
+
}
|
9
|
+
|
10
|
+
/*
|
11
|
+
* Copied from fann_create_train_from_callback/file & modified to ease
|
12
|
+
* allocating from ruby arrays:
|
13
|
+
*/
|
14
|
+
FANN_EXTERNAL struct fann_train_data * FANN_API fann_create_train_from_rb_ary(
|
15
|
+
VALUE inputs,
|
16
|
+
VALUE outputs
|
17
|
+
)
|
18
|
+
{
|
19
|
+
unsigned int i, j;
|
20
|
+
fann_type *data_input, *data_output;
|
21
|
+
struct fann_train_data *data = (struct fann_train_data *)malloc(sizeof(struct fann_train_data));
|
22
|
+
unsigned int num_input = RARRAY(RARRAY(inputs)->ptr[0])->len;
|
23
|
+
unsigned int num_output =RARRAY(RARRAY(outputs)->ptr[0])->len;
|
24
|
+
unsigned int num_data = RARRAY(inputs)->len;
|
25
|
+
|
26
|
+
if(data == NULL) {
|
27
|
+
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
|
28
|
+
return NULL;
|
29
|
+
}
|
30
|
+
|
31
|
+
fann_init_error_data((struct fann_error *) data);
|
32
|
+
|
33
|
+
data->num_data = num_data;
|
34
|
+
data->num_input = num_input;
|
35
|
+
data->num_output = num_output;
|
36
|
+
|
37
|
+
data->input = (fann_type **) calloc(num_data, sizeof(fann_type *));
|
38
|
+
if(data->input == NULL)
|
39
|
+
{
|
40
|
+
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
|
41
|
+
fann_destroy_train(data);
|
42
|
+
return NULL;
|
43
|
+
}
|
44
|
+
|
45
|
+
data->output = (fann_type **) calloc(num_data, sizeof(fann_type *));
|
46
|
+
if(data->output == NULL)
|
47
|
+
{
|
48
|
+
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
|
49
|
+
fann_destroy_train(data);
|
50
|
+
return NULL;
|
51
|
+
}
|
52
|
+
|
53
|
+
data_input = (fann_type *) calloc(num_input * num_data, sizeof(fann_type));
|
54
|
+
if(data_input == NULL)
|
55
|
+
{
|
56
|
+
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
|
57
|
+
fann_destroy_train(data);
|
58
|
+
return NULL;
|
59
|
+
}
|
60
|
+
|
61
|
+
data_output = (fann_type *) calloc(num_output * num_data, sizeof(fann_type));
|
62
|
+
if(data_output == NULL)
|
63
|
+
{
|
64
|
+
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
|
65
|
+
fann_destroy_train(data);
|
66
|
+
return NULL;
|
67
|
+
}
|
68
|
+
|
69
|
+
VALUE inputs_i, outputs_i;
|
70
|
+
for(i = 0; i != num_data; i++)
|
71
|
+
{
|
72
|
+
data->input[i] = data_input;
|
73
|
+
data_input += num_input;
|
74
|
+
|
75
|
+
inputs_i = RARRAY(inputs)->ptr[i];
|
76
|
+
outputs_i = RARRAY(outputs)->ptr[i];
|
77
|
+
|
78
|
+
if(RARRAY(inputs_i)->len != num_input)
|
79
|
+
{
|
80
|
+
rb_raise (
|
81
|
+
rb_eRuntimeError,
|
82
|
+
"Number of inputs at [%d] is inconsistent: (%d != %d)",
|
83
|
+
i, RARRAY(inputs_i)->len, num_input);
|
84
|
+
}
|
85
|
+
|
86
|
+
if(RARRAY(outputs_i)->len != num_output)
|
87
|
+
{
|
88
|
+
rb_raise (
|
89
|
+
rb_eRuntimeError,
|
90
|
+
"Number of outputs at [%d] is inconsistent: (%d != %d)",
|
91
|
+
i, RARRAY(outputs_i)->len, num_output);
|
92
|
+
}
|
93
|
+
|
94
|
+
|
95
|
+
for(j = 0; j != num_input; j++)
|
96
|
+
{
|
97
|
+
data->input[i][j]=NUM2DBL(RARRAY(inputs_i)->ptr[j]);
|
98
|
+
}
|
99
|
+
|
100
|
+
data->output[i] = data_output;
|
101
|
+
data_output += num_output;
|
102
|
+
|
103
|
+
for(j = 0; j != num_output; j++)
|
104
|
+
{
|
105
|
+
data->output[i][j]=NUM2DBL(RARRAY(outputs_i)->ptr[j]);
|
106
|
+
}
|
107
|
+
}
|
108
|
+
|
109
|
+
return data;
|
110
|
+
}
|
@@ -0,0 +1,1499 @@
|
|
1
|
+
#include "ruby.h"
|
2
|
+
// #include "fann.h"
|
3
|
+
#include "doublefann.h"
|
4
|
+
#include "fann_data.h"
|
5
|
+
#include "fann_augment.h"
|
6
|
+
|
7
|
+
static VALUE m_rb_fann_module;
|
8
|
+
static VALUE m_rb_fann_standard_class;
|
9
|
+
static VALUE m_rb_fann_shortcut_class;
|
10
|
+
static VALUE m_rb_fann_train_data_class;
|
11
|
+
|
12
|
+
#define RETURN_FANN_INT(fn) \
|
13
|
+
struct fann* f; \
|
14
|
+
Data_Get_Struct (self, struct fann, f); \
|
15
|
+
return INT2NUM(fn(f));
|
16
|
+
|
17
|
+
#define SET_FANN_INT(attr_name, fann_fn) \
|
18
|
+
Check_Type(attr_name, T_FIXNUM); \
|
19
|
+
struct fann* f; \
|
20
|
+
Data_Get_Struct(self, struct fann, f); \
|
21
|
+
fann_fn(f, NUM2INT(attr_name)); \
|
22
|
+
|
23
|
+
#define RETURN_FANN_UINT(fn) \
|
24
|
+
struct fann* f; \
|
25
|
+
Data_Get_Struct (self, struct fann, f); \
|
26
|
+
return UINT2NUM(fn(f));
|
27
|
+
|
28
|
+
#define SET_FANN_UINT(attr_name, fann_fn) \
|
29
|
+
Check_Type(attr_name, T_FIXNUM); \
|
30
|
+
struct fann* f; \
|
31
|
+
Data_Get_Struct(self, struct fann, f); \
|
32
|
+
fann_fn(f, NUM2UINT(attr_name)); \
|
33
|
+
|
34
|
+
// Converts float return values to a double with same precision, avoids floating point errors.
|
35
|
+
#define RETURN_FANN_FLT(fn) \
|
36
|
+
struct fann* f; \
|
37
|
+
Data_Get_Struct (self, struct fann, f); \
|
38
|
+
char buffy[20]; \
|
39
|
+
sprintf(buffy, "%0.6g", fn(f)); \
|
40
|
+
return rb_float_new(atof(buffy));
|
41
|
+
|
42
|
+
#define SET_FANN_FLT(attr_name, fann_fn) \
|
43
|
+
Check_Type(attr_name, T_FLOAT); \
|
44
|
+
struct fann* f; \
|
45
|
+
Data_Get_Struct(self, struct fann, f); \
|
46
|
+
fann_fn(f, NUM2DBL(attr_name)); \
|
47
|
+
|
48
|
+
#define RETURN_FANN_DBL(fn) \
|
49
|
+
struct fann* f; \
|
50
|
+
Data_Get_Struct (self, struct fann, f); \
|
51
|
+
return rb_float_new(fn(f));
|
52
|
+
|
53
|
+
#define SET_FANN_DBL SET_FANN_FLT
|
54
|
+
|
55
|
+
// Convert ruby symbol to corresponding FANN enum type for activation function:
|
56
|
+
enum fann_activationfunc_enum sym_to_activation_function(VALUE activation_func)
|
57
|
+
{
|
58
|
+
ID id=SYM2ID(activation_func);
|
59
|
+
enum fann_activationfunc_enum activation_function;
|
60
|
+
if(id==rb_intern("linear")) {
|
61
|
+
activation_function = FANN_LINEAR;
|
62
|
+
} else if(id==rb_intern("threshold")) {
|
63
|
+
activation_function = FANN_THRESHOLD;
|
64
|
+
} else if(id==rb_intern("threshold_symmetric")) {
|
65
|
+
activation_function = FANN_THRESHOLD_SYMMETRIC;
|
66
|
+
} else if(id==rb_intern("sigmoid")) {
|
67
|
+
activation_function = FANN_SIGMOID;
|
68
|
+
} else if(id==rb_intern("sigmoid_stepwise")) {
|
69
|
+
activation_function = FANN_SIGMOID_STEPWISE;
|
70
|
+
} else if(id==rb_intern("sigmoid_symmetric")) {
|
71
|
+
activation_function = FANN_SIGMOID_SYMMETRIC;
|
72
|
+
} else if(id==rb_intern("sigmoid_symmetric_stepwise")) {
|
73
|
+
activation_function = FANN_SIGMOID_SYMMETRIC_STEPWISE;
|
74
|
+
} else if(id==rb_intern("gaussian")) {
|
75
|
+
activation_function = FANN_GAUSSIAN;
|
76
|
+
} else if(id==rb_intern("gaussian_symmetric")) {
|
77
|
+
activation_function = FANN_GAUSSIAN_SYMMETRIC;
|
78
|
+
} else if(id==rb_intern("gaussian_stepwise")) {
|
79
|
+
activation_function = FANN_GAUSSIAN_STEPWISE;
|
80
|
+
} else if(id==rb_intern("elliot")) {
|
81
|
+
activation_function = FANN_ELLIOT;
|
82
|
+
} else if(id==rb_intern("elliot_symmetric")) {
|
83
|
+
activation_function = FANN_ELLIOT_SYMMETRIC;
|
84
|
+
} else if(id==rb_intern("linear_piece")) {
|
85
|
+
activation_function = FANN_LINEAR_PIECE;
|
86
|
+
} else if(id==rb_intern("linear_piece_symmetric")) {
|
87
|
+
activation_function = FANN_LINEAR_PIECE_SYMMETRIC;
|
88
|
+
} else if(id==rb_intern("sin_symmetric")) {
|
89
|
+
activation_function = FANN_SIN_SYMMETRIC;
|
90
|
+
} else if(id==rb_intern("cos_symmetric")) {
|
91
|
+
activation_function = FANN_COS_SYMMETRIC;
|
92
|
+
} else if(id==rb_intern("sin")) {
|
93
|
+
activation_function = FANN_SIN;
|
94
|
+
} else if(id==rb_intern("cos")) {
|
95
|
+
activation_function = FANN_COS;
|
96
|
+
} else {
|
97
|
+
rb_raise(rb_eRuntimeError, "Unrecognized activation function: [%s]", rb_id2name(SYM2ID(activation_func)));
|
98
|
+
}
|
99
|
+
return activation_function;
|
100
|
+
}
|
101
|
+
|
102
|
+
// Convert FANN enum type for activation function to corresponding ruby symbol:
|
103
|
+
VALUE activation_function_to_sym(enum fann_activationfunc_enum fn)
|
104
|
+
{
|
105
|
+
VALUE activation_function;
|
106
|
+
|
107
|
+
if(fn==FANN_LINEAR) {
|
108
|
+
activation_function = ID2SYM(rb_intern("linear"));
|
109
|
+
} else if(fn==FANN_THRESHOLD) {
|
110
|
+
activation_function = ID2SYM(rb_intern("threshold"));
|
111
|
+
} else if(fn==FANN_THRESHOLD_SYMMETRIC) {
|
112
|
+
activation_function = ID2SYM(rb_intern("threshold_symmetric"));
|
113
|
+
} else if(fn==FANN_SIGMOID) {
|
114
|
+
activation_function = ID2SYM(rb_intern("sigmoid"));
|
115
|
+
} else if(fn==FANN_SIGMOID_STEPWISE) {
|
116
|
+
activation_function = ID2SYM(rb_intern("sigmoid_stepwise"));
|
117
|
+
} else if(fn==FANN_SIGMOID_SYMMETRIC) {
|
118
|
+
activation_function = ID2SYM(rb_intern("sigmoid_symmetric"));
|
119
|
+
} else if(fn==FANN_SIGMOID_SYMMETRIC_STEPWISE) {
|
120
|
+
activation_function = ID2SYM(rb_intern("sigmoid_symmetric_stepwise"));
|
121
|
+
} else if(fn==FANN_GAUSSIAN) {
|
122
|
+
activation_function = ID2SYM(rb_intern("gaussian"));
|
123
|
+
} else if(fn==FANN_GAUSSIAN_SYMMETRIC) {
|
124
|
+
activation_function = ID2SYM(rb_intern("gaussian_symmetric"));
|
125
|
+
} else if(fn==FANN_GAUSSIAN_STEPWISE) {
|
126
|
+
activation_function = ID2SYM(rb_intern("gaussian_stepwise"));
|
127
|
+
} else if(fn==FANN_ELLIOT) {
|
128
|
+
activation_function = ID2SYM(rb_intern("elliot"));
|
129
|
+
} else if(fn==FANN_ELLIOT_SYMMETRIC) {
|
130
|
+
activation_function = ID2SYM(rb_intern("elliot_symmetric"));
|
131
|
+
} else if(fn==FANN_LINEAR_PIECE) {
|
132
|
+
activation_function = ID2SYM(rb_intern("linear_piece"));
|
133
|
+
} else if(fn==FANN_LINEAR_PIECE_SYMMETRIC) {
|
134
|
+
activation_function = ID2SYM(rb_intern("linear_piece_symmetric"));
|
135
|
+
} else if(fn==FANN_SIN_SYMMETRIC) {
|
136
|
+
activation_function = ID2SYM(rb_intern("sin_symmetric"));
|
137
|
+
} else if(fn==FANN_COS_SYMMETRIC) {
|
138
|
+
activation_function = ID2SYM(rb_intern("cos_symmetric"));
|
139
|
+
} else if(fn==FANN_SIN) {
|
140
|
+
activation_function = ID2SYM(rb_intern("sin"));
|
141
|
+
} else if(fn==FANN_COS) {
|
142
|
+
activation_function = ID2SYM(rb_intern("cos"));
|
143
|
+
} else {
|
144
|
+
rb_raise(rb_eRuntimeError, "Unrecognized activation function: [%d]", fn);
|
145
|
+
}
|
146
|
+
return activation_function;
|
147
|
+
}
|
148
|
+
|
149
|
+
|
150
|
+
// Unused for now:
|
151
|
+
static void fann_mark (struct fann* ann){}
|
152
|
+
|
153
|
+
// #define DEBUG 1
|
154
|
+
|
155
|
+
// Free memory associated with FANN:
|
156
|
+
static void fann_free (struct fann* ann)
|
157
|
+
{
|
158
|
+
fann_destroy(ann);
|
159
|
+
printf("Destroyed FANN network [%d].\n", ann);
|
160
|
+
}
|
161
|
+
|
162
|
+
// Free memory associated with FANN Training data:
|
163
|
+
static void fann_training_data_free (struct fann_train_data* train_data)
|
164
|
+
{
|
165
|
+
fann_destroy_train(train_data);
|
166
|
+
printf("Destroyed Training data [%d].\n", train_data);
|
167
|
+
}
|
168
|
+
|
169
|
+
// Create wrapper, but don't allocate anything...do that in
|
170
|
+
// initialize, so we can construct with args:
|
171
|
+
static VALUE fann_allocate (VALUE klass)
|
172
|
+
{
|
173
|
+
return Data_Wrap_Struct (klass, fann_mark, fann_free, 0);
|
174
|
+
}
|
175
|
+
|
176
|
+
// Create wrapper, but don't allocate annything...do that in
|
177
|
+
// initialize, so we can construct with args:
|
178
|
+
static VALUE fann_training_data_allocate (VALUE klass)
|
179
|
+
{
|
180
|
+
return Data_Wrap_Struct (klass, fann_mark, fann_training_data_free, 0);
|
181
|
+
}
|
182
|
+
|
183
|
+
|
184
|
+
/** Initialization routine for both standard, shortcut & filename forms of FANN:
|
185
|
+
Standard Initialization:
|
186
|
+
RubyFann::Standard.new(:num_inputs=>1, :hidden_neurons=>[3, 4, 3, 4], :num_outputs=>1)
|
187
|
+
|
188
|
+
Shortcut Initialization (e.g., for use in cascade training):
|
189
|
+
RubyFann::Shortcut.new(:num_inputs=>5, :num_outputs=>1)
|
190
|
+
|
191
|
+
File Initialization:
|
192
|
+
RubyFann::Standard.new(:filename=>'xor_float.net') */
|
193
|
+
static VALUE fann_initialize(VALUE self, VALUE hash)
|
194
|
+
{
|
195
|
+
// Get args:
|
196
|
+
VALUE filename = rb_hash_aref(hash, ID2SYM(rb_intern("filename")));
|
197
|
+
VALUE num_inputs = rb_hash_aref(hash, ID2SYM(rb_intern("num_inputs")));
|
198
|
+
VALUE num_outputs = rb_hash_aref(hash, ID2SYM(rb_intern("num_outputs")));
|
199
|
+
VALUE hidden_neurons = rb_hash_aref(hash, ID2SYM(rb_intern("hidden_neurons")));
|
200
|
+
|
201
|
+
struct fann* ann;
|
202
|
+
if (TYPE(filename)==T_STRING)
|
203
|
+
{
|
204
|
+
// Initialize with file:
|
205
|
+
// train_data = fann_read_train_from_file(StringValuePtr(filename));
|
206
|
+
// DATA_PTR(self) = train_data;
|
207
|
+
ann = fann_create_from_file(StringValuePtr(filename));
|
208
|
+
printf("Created RubyFann::Standard [%d] from file [%s].\n", ann, StringValuePtr(filename));
|
209
|
+
}
|
210
|
+
else if(rb_obj_is_kind_of(self, m_rb_fann_shortcut_class))
|
211
|
+
{
|
212
|
+
// Initialize as shortcut, suitable for cascade training:
|
213
|
+
//ann = fann_create_shortcut_array(num_layers, layers);
|
214
|
+
Check_Type(num_inputs, T_FIXNUM);
|
215
|
+
Check_Type(num_outputs, T_FIXNUM);
|
216
|
+
|
217
|
+
ann = fann_create_shortcut(2, NUM2INT(num_inputs), NUM2INT(num_outputs));
|
218
|
+
printf("Created RubyFann::Shortcut [%d].\n", ann);
|
219
|
+
}
|
220
|
+
else
|
221
|
+
{
|
222
|
+
// Initialize as standard:
|
223
|
+
Check_Type(num_inputs, T_FIXNUM);
|
224
|
+
Check_Type(hidden_neurons, T_ARRAY);
|
225
|
+
Check_Type(num_outputs, T_FIXNUM);
|
226
|
+
|
227
|
+
// Initialize layers:
|
228
|
+
unsigned int num_layers=RARRAY(hidden_neurons)->len + 2; // NUM2INT(num_inputs) + NUM2INT(num_outputs) + RARRAY(hidden_neurons)->len;
|
229
|
+
unsigned int layers[num_layers];
|
230
|
+
|
231
|
+
// Input:
|
232
|
+
layers[0]=NUM2INT(num_inputs);
|
233
|
+
// Output:
|
234
|
+
layers[num_layers-1]=NUM2INT(num_outputs);
|
235
|
+
// Hidden:
|
236
|
+
int i;
|
237
|
+
for (i=1; i<=num_layers-2; i++) {
|
238
|
+
layers[i]=NUM2UINT(RARRAY(hidden_neurons)->ptr[i-1]);
|
239
|
+
printf("Setting layer [%d] to [%d]\n", i, layers[i]);
|
240
|
+
}
|
241
|
+
|
242
|
+
ann = fann_create_standard_array(num_layers, layers);
|
243
|
+
printf("Created RubyFann::Standard [%d].\n", ann);
|
244
|
+
}
|
245
|
+
|
246
|
+
DATA_PTR(self) = ann;
|
247
|
+
return (VALUE)ann;
|
248
|
+
|
249
|
+
DATA_PTR(self) = ann;
|
250
|
+
return (VALUE)ann;
|
251
|
+
}
|
252
|
+
|
253
|
+
/**
|
254
|
+
Initialize in one of the following forms:
|
255
|
+
|
256
|
+
# This is a flat file with training data as described in FANN docs.
|
257
|
+
RubyFann::TrainData.new(:filename => 'path/to/training_file.train')
|
258
|
+
OR
|
259
|
+
# Train with inputs (array of arrays) & desired_outputs (array of arrays)
|
260
|
+
# inputs & desired outputs should be of same length
|
261
|
+
# All sub-arrays on inputs should be of same length
|
262
|
+
# All sub-arrays on desired_outputs should be of same length
|
263
|
+
# Sub-arrays on inputs & desired_outputs can be different sizes from one another
|
264
|
+
RubyFann::TrainData.new(:inputs=>[[0.2, 0.3, 0.4], [0.8, 0.9, 0.7]], :desired_outputs=>[[3.14], [6.33]])
|
265
|
+
*/
|
266
|
+
static VALUE fann_train_data_initialize(VALUE self, VALUE hash)
|
267
|
+
{
|
268
|
+
struct fann_train_data* train_data;
|
269
|
+
Check_Type(hash, T_HASH);
|
270
|
+
|
271
|
+
VALUE filename = rb_hash_aref(hash, ID2SYM(rb_intern("filename")));
|
272
|
+
VALUE inputs = rb_hash_aref(hash, ID2SYM(rb_intern("inputs")));
|
273
|
+
VALUE desired_outputs = rb_hash_aref(hash, ID2SYM(rb_intern("desired_outputs")));
|
274
|
+
|
275
|
+
if (TYPE(filename)==T_STRING)
|
276
|
+
{
|
277
|
+
train_data = fann_read_train_from_file(StringValuePtr(filename));
|
278
|
+
DATA_PTR(self) = train_data;
|
279
|
+
}
|
280
|
+
else if (TYPE(inputs)==T_ARRAY)
|
281
|
+
{
|
282
|
+
if (TYPE(desired_outputs)!=T_ARRAY)
|
283
|
+
{
|
284
|
+
rb_raise (rb_eRuntimeError, "[desired_outputs] must be present when [inputs] used.");
|
285
|
+
}
|
286
|
+
|
287
|
+
// The data is here, start constructing:
|
288
|
+
if(RARRAY(inputs)->len != RARRAY(desired_outputs)->len)
|
289
|
+
{
|
290
|
+
rb_raise (
|
291
|
+
rb_eRuntimeError,
|
292
|
+
"Number of inputs must match number of outputs: (%d != %d)",
|
293
|
+
RARRAY(inputs)->len,
|
294
|
+
RARRAY(desired_outputs)->len);
|
295
|
+
}
|
296
|
+
|
297
|
+
train_data = fann_create_train_from_rb_ary(inputs, desired_outputs);
|
298
|
+
DATA_PTR(self) = train_data;
|
299
|
+
}
|
300
|
+
else
|
301
|
+
{
|
302
|
+
rb_raise (rb_eRuntimeError, "Must construct with a filename(string) or inputs/desired_outputs(arrays). All args passed via hash with symbols as keys.");
|
303
|
+
}
|
304
|
+
|
305
|
+
return (VALUE)train_data;
|
306
|
+
}
|
307
|
+
|
308
|
+
|
309
|
+
/** Save to given filename */
|
310
|
+
static VALUE training_save(VALUE self, VALUE filename)
|
311
|
+
{
|
312
|
+
Check_Type(filename, T_STRING);
|
313
|
+
struct fann_train_data* t;
|
314
|
+
Data_Get_Struct (self, struct fann_train_data, t);
|
315
|
+
fann_save_train(t, StringValuePtr(filename));
|
316
|
+
}
|
317
|
+
|
318
|
+
/** Shuffles training data, randomizing the order.
|
319
|
+
This is recommended for incremental training, while it will have no influence during batch training. */
|
320
|
+
static VALUE shuffle(VALUE self)
|
321
|
+
{
|
322
|
+
struct fann_train_data* t;
|
323
|
+
Data_Get_Struct (self, struct fann_train_data, t);
|
324
|
+
fann_shuffle_train_data(t);
|
325
|
+
}
|
326
|
+
|
327
|
+
/** Length of training data */
|
328
|
+
static VALUE length_train_data(VALUE self)
|
329
|
+
{
|
330
|
+
struct fann_train_data* t;
|
331
|
+
Data_Get_Struct (self, struct fann_train_data, t);
|
332
|
+
return(UINT2NUM(fann_length_train_data(t)));
|
333
|
+
}
|
334
|
+
|
335
|
+
/** Set the activation function for neuron number *neuron* in layer number *layer*,
|
336
|
+
counting the input layer as layer 0. activation_func must be one of the following symbols:
|
337
|
+
:linear, :threshold, :threshold_symmetric, :sigmoid, :sigmoid_stepwise, :sigmoid_symmetric,
|
338
|
+
:sigmoid_symmetric_stepwise, :gaussian, :gaussian_symmetric, :gaussian_stepwise, :elliot,
|
339
|
+
:elliot_symmetric, :linear_piece, :linear_piece_symmetric, :sin_symmetric, :cos_symmetric,
|
340
|
+
:sin, :cos
|
341
|
+
*/
|
342
|
+
static VALUE set_activation_function(VALUE self, VALUE activation_func, VALUE layer, VALUE neuron)
|
343
|
+
{
|
344
|
+
Check_Type(activation_func, T_SYMBOL);
|
345
|
+
Check_Type(layer, T_FIXNUM);
|
346
|
+
Check_Type(neuron, T_FIXNUM);
|
347
|
+
|
348
|
+
struct fann* f;
|
349
|
+
Data_Get_Struct(self, struct fann, f);
|
350
|
+
fann_set_activation_function(f, sym_to_activation_function(activation_func), NUM2INT(layer), NUM2INT(neuron));
|
351
|
+
}
|
352
|
+
|
353
|
+
/** Set the activation function for all of the hidden layers. activation_func must be one of the following symbols:
|
354
|
+
:linear, :threshold, :threshold_symmetric, :sigmoid, :sigmoid_stepwise, :sigmoid_symmetric,
|
355
|
+
:sigmoid_symmetric_stepwise, :gaussian, :gaussian_symmetric, :gaussian_stepwise, :elliot,
|
356
|
+
:elliot_symmetric, :linear_piece, :linear_piece_symmetric, :sin_symmetric, :cos_symmetric,
|
357
|
+
:sin, :cos*/
|
358
|
+
static VALUE set_activation_function_hidden(VALUE self, VALUE activation_func)
|
359
|
+
{
|
360
|
+
Check_Type(activation_func, T_SYMBOL);
|
361
|
+
struct fann* f;
|
362
|
+
Data_Get_Struct(self, struct fann, f);
|
363
|
+
fann_set_activation_function_hidden(f, sym_to_activation_function(activation_func));
|
364
|
+
}
|
365
|
+
|
366
|
+
/** Set the activation function for all the neurons in the layer number *layer*,
|
367
|
+
counting the input layer as layer 0. activation_func must be one of the following symbols:
|
368
|
+
:linear, :threshold, :threshold_symmetric, :sigmoid, :sigmoid_stepwise, :sigmoid_symmetric,
|
369
|
+
:sigmoid_symmetric_stepwise, :gaussian, :gaussian_symmetric, :gaussian_stepwise, :elliot,
|
370
|
+
:elliot_symmetric, :linear_piece, :linear_piece_symmetric, :sin_symmetric, :cos_symmetric,
|
371
|
+
:sin, :cos */
|
372
|
+
static VALUE set_activation_function_layer(VALUE self, VALUE activation_func, VALUE layer)
|
373
|
+
{
|
374
|
+
Check_Type(activation_func, T_SYMBOL);
|
375
|
+
Check_Type(layer, T_FIXNUM);
|
376
|
+
struct fann* f;
|
377
|
+
Data_Get_Struct(self, struct fann, f);
|
378
|
+
fann_set_activation_function_layer(f, sym_to_activation_function(activation_func), NUM2INT(layer));
|
379
|
+
}
|
380
|
+
|
381
|
+
/** Set the activation function for the output layer. activation_func must be one of the following symbols:
|
382
|
+
:linear, :threshold, :threshold_symmetric, :sigmoid, :sigmoid_stepwise, :sigmoid_symmetric,
|
383
|
+
:sigmoid_symmetric_stepwise, :gaussian, :gaussian_symmetric, :gaussian_stepwise, :elliot,
|
384
|
+
:elliot_symmetric, :linear_piece, :linear_piece_symmetric, :sin_symmetric, :cos_symmetric,
|
385
|
+
:sin, :cos */
|
386
|
+
static VALUE set_activation_function_output(VALUE self, VALUE activation_func)
|
387
|
+
{
|
388
|
+
Check_Type(activation_func, T_SYMBOL);
|
389
|
+
struct fann* f;
|
390
|
+
Data_Get_Struct(self, struct fann, f);
|
391
|
+
fann_set_activation_function_output(f, sym_to_activation_function(activation_func));
|
392
|
+
}
|
393
|
+
|
394
|
+
/** Get the activation steepness for neuron number neuron in layer number layer, counting the input layer as layer 0. */
|
395
|
+
static VALUE get_activation_steepness(self, layer, neuron)
|
396
|
+
VALUE self; VALUE layer; VALUE neuron;
|
397
|
+
{
|
398
|
+
Check_Type(layer, T_FIXNUM);
|
399
|
+
Check_Type(neuron, T_FIXNUM);
|
400
|
+
struct fann* f;
|
401
|
+
Data_Get_Struct(self, struct fann, f);
|
402
|
+
fann_type val = fann_get_activation_steepness(f, NUM2INT(layer), NUM2INT(neuron));
|
403
|
+
return rb_float_new(val);
|
404
|
+
}
|
405
|
+
|
406
|
+
/** Set the activation steepness for neuron number {neuron} in layer number (({layer})),
|
407
|
+
counting the input layer as layer 0. */
|
408
|
+
static VALUE set_activation_steepness(VALUE self, VALUE steepness, VALUE layer, VALUE neuron)
|
409
|
+
{
|
410
|
+
Check_Type(steepness, T_FLOAT);
|
411
|
+
Check_Type(layer, T_FIXNUM);
|
412
|
+
Check_Type(neuron, T_FIXNUM);
|
413
|
+
|
414
|
+
struct fann* f;
|
415
|
+
Data_Get_Struct(self, struct fann, f);
|
416
|
+
fann_set_activation_steepness(f, NUM2DBL(steepness), NUM2INT(layer), NUM2INT(neuron));
|
417
|
+
}
|
418
|
+
|
419
|
+
/** Set the activation steepness in all of the hidden layers. */
|
420
|
+
static VALUE set_activation_steepness_hidden(VALUE self, VALUE steepness)
|
421
|
+
{
|
422
|
+
SET_FANN_FLT(steepness, fann_set_activation_steepness_hidden);
|
423
|
+
}
|
424
|
+
|
425
|
+
/** Set the activation steepness all of the neurons in layer number *layer*,
|
426
|
+
counting the input layer as layer 0. */
|
427
|
+
static VALUE set_activation_steepness_layer(VALUE self, VALUE steepness, VALUE layer)
|
428
|
+
{
|
429
|
+
Check_Type(steepness, T_FLOAT);
|
430
|
+
Check_Type(layer, T_FIXNUM);
|
431
|
+
|
432
|
+
struct fann* f;
|
433
|
+
Data_Get_Struct(self, struct fann, f);
|
434
|
+
fann_set_activation_steepness_layer(f, NUM2DBL(steepness), NUM2INT(layer));
|
435
|
+
}
|
436
|
+
|
437
|
+
/** Set the activation steepness in the output layer. */
|
438
|
+
static VALUE set_activation_steepness_output(VALUE self, VALUE steepness)
|
439
|
+
{
|
440
|
+
SET_FANN_FLT(steepness, fann_set_activation_steepness_output);
|
441
|
+
}
|
442
|
+
|
443
|
+
/** Returns the bit fail limit used during training. */
|
444
|
+
static VALUE get_bit_fail_limit(VALUE self)
|
445
|
+
{
|
446
|
+
RETURN_FANN_DBL(fann_get_bit_fail_limit);
|
447
|
+
}
|
448
|
+
|
449
|
+
/** Sets the bit fail limit used during training. */
|
450
|
+
static VALUE set_bit_fail_limit(VALUE self, VALUE bit_fail_limit)
|
451
|
+
{
|
452
|
+
SET_FANN_FLT(bit_fail_limit, fann_set_bit_fail_limit);
|
453
|
+
}
|
454
|
+
|
455
|
+
/** The decay is a small negative valued number which is the factor that the weights
|
456
|
+
should become smaller in each iteration during quickprop training. This is used
|
457
|
+
to make sure that the weights do not become too high during training. */
|
458
|
+
static VALUE get_quickprop_decay(VALUE self)
|
459
|
+
{
|
460
|
+
RETURN_FANN_FLT(fann_get_quickprop_decay);
|
461
|
+
}
|
462
|
+
|
463
|
+
/** Sets the quickprop decay factor */
|
464
|
+
static VALUE set_quickprop_decay(VALUE self, VALUE quickprop_decay)
|
465
|
+
{
|
466
|
+
SET_FANN_FLT(quickprop_decay, fann_set_quickprop_decay);
|
467
|
+
}
|
468
|
+
|
469
|
+
/** The mu factor is used to increase and decrease the step-size during quickprop training.
|
470
|
+
The mu factor should always be above 1, since it would otherwise decrease the step-size
|
471
|
+
when it was suppose to increase it. */
|
472
|
+
static VALUE get_quickprop_mu(VALUE self)
|
473
|
+
{
|
474
|
+
RETURN_FANN_FLT(fann_get_quickprop_mu);
|
475
|
+
}
|
476
|
+
|
477
|
+
/** Sets the quickprop mu factor. */
|
478
|
+
static VALUE set_quickprop_mu(VALUE self, VALUE quickprop_mu)
|
479
|
+
{
|
480
|
+
SET_FANN_FLT(quickprop_mu, fann_set_quickprop_mu);
|
481
|
+
}
|
482
|
+
|
483
|
+
/** The increase factor is a value larger than 1, which is used to
|
484
|
+
increase the step-size during RPROP training. */
|
485
|
+
static VALUE get_rprop_increase_factor(VALUE self)
|
486
|
+
{
|
487
|
+
RETURN_FANN_FLT(fann_get_rprop_increase_factor);
|
488
|
+
}
|
489
|
+
|
490
|
+
/** The increase factor used during RPROP training. */
|
491
|
+
static VALUE set_rprop_increase_factor(VALUE self, VALUE rprop_increase_factor)
|
492
|
+
{
|
493
|
+
SET_FANN_FLT(rprop_increase_factor, fann_set_rprop_increase_factor);
|
494
|
+
}
|
495
|
+
|
496
|
+
/** The decrease factor is a value smaller than 1, which is used to decrease the step-size during RPROP training. */
|
497
|
+
static VALUE get_rprop_decrease_factor(VALUE self)
|
498
|
+
{
|
499
|
+
RETURN_FANN_FLT(fann_get_rprop_decrease_factor);
|
500
|
+
}
|
501
|
+
|
502
|
+
/** The decrease factor is a value smaller than 1, which is used to decrease the step-size during RPROP training. */
|
503
|
+
static VALUE set_rprop_decrease_factor(VALUE self, VALUE rprop_decrease_factor)
|
504
|
+
{
|
505
|
+
SET_FANN_FLT(rprop_decrease_factor, fann_set_rprop_decrease_factor);
|
506
|
+
}
|
507
|
+
|
508
|
+
/** The minimum step-size is a small positive number determining how small the minimum step-size may be. */
|
509
|
+
static VALUE get_rprop_delta_min(VALUE self)
|
510
|
+
{
|
511
|
+
RETURN_FANN_FLT(fann_get_rprop_delta_min);
|
512
|
+
}
|
513
|
+
|
514
|
+
/** The minimum step-size is a small positive number determining how small the minimum step-size may be. */
|
515
|
+
static VALUE set_rprop_delta_min(VALUE self, VALUE rprop_delta_min)
|
516
|
+
{
|
517
|
+
SET_FANN_FLT(rprop_delta_min, fann_set_rprop_delta_min);
|
518
|
+
}
|
519
|
+
|
520
|
+
/** The maximum step-size is a positive number determining how large the maximum step-size may be. */
|
521
|
+
static VALUE get_rprop_delta_max(VALUE self)
|
522
|
+
{
|
523
|
+
RETURN_FANN_FLT(fann_get_rprop_delta_max);
|
524
|
+
}
|
525
|
+
|
526
|
+
/** The maximum step-size is a positive number determining how large the maximum step-size may be. */
|
527
|
+
static VALUE set_rprop_delta_max(VALUE self, VALUE rprop_delta_max)
|
528
|
+
{
|
529
|
+
SET_FANN_FLT(rprop_delta_max, fann_set_rprop_delta_max);
|
530
|
+
}
|
531
|
+
|
532
|
+
/** The initial step-size is a positive number determining the initial step size. */
|
533
|
+
static VALUE get_rprop_delta_zero(VALUE self)
|
534
|
+
{
|
535
|
+
RETURN_FANN_FLT(fann_get_rprop_delta_zero);
|
536
|
+
}
|
537
|
+
|
538
|
+
/** The initial step-size is a positive number determining the initial step size. */
|
539
|
+
static VALUE set_rprop_delta_zero(VALUE self, VALUE rprop_delta_zero)
|
540
|
+
{
|
541
|
+
SET_FANN_FLT(rprop_delta_zero, fann_set_rprop_delta_zero);
|
542
|
+
}
|
543
|
+
|
544
|
+
/** Return array of bias(es) */
|
545
|
+
static VALUE get_bias_array(VALUE self)
|
546
|
+
{
|
547
|
+
struct fann* f;
|
548
|
+
unsigned int num_layers;
|
549
|
+
Data_Get_Struct (self, struct fann, f);
|
550
|
+
num_layers = fann_get_num_layers(f);
|
551
|
+
unsigned int layers[num_layers];
|
552
|
+
fann_get_bias_array(f, layers);
|
553
|
+
|
554
|
+
// Create ruby array & set outputs:
|
555
|
+
VALUE arr;
|
556
|
+
arr = rb_ary_new();
|
557
|
+
int i;
|
558
|
+
for (i=0; i<num_layers; i++)
|
559
|
+
{
|
560
|
+
rb_ary_push(arr, INT2NUM(layers[i]));
|
561
|
+
}
|
562
|
+
|
563
|
+
return arr;
|
564
|
+
}
|
565
|
+
|
566
|
+
/** The number of fail bits; means the number of output neurons which differ more
|
567
|
+
than the bit fail limit (see <fann_get_bit_fail_limit>, <fann_set_bit_fail_limit>).
|
568
|
+
The bits are counted in all of the training data, so this number can be higher than
|
569
|
+
the number of training data. */
|
570
|
+
static VALUE get_bit_fail(VALUE self)
|
571
|
+
{
|
572
|
+
RETURN_FANN_INT(fann_get_bit_fail);
|
573
|
+
}
|
574
|
+
|
575
|
+
/** Get the connection rate used when the network was created. */
|
576
|
+
static VALUE get_connection_rate(VALUE self)
|
577
|
+
{
|
578
|
+
RETURN_FANN_INT(fann_get_connection_rate);
|
579
|
+
}
|
580
|
+
|
581
|
+
/** Return array<hash> where each array element is a hash
|
582
|
+
representing a neuron. It contains the following keys:
|
583
|
+
:activation_function, symbol -- the activation function
|
584
|
+
:activation_steepness=float -- The steepness of the activation function
|
585
|
+
:sum=float -- The sum of the inputs multiplied with the weights
|
586
|
+
:value=float -- The value of the activation fuction applied to the sum
|
587
|
+
:connections=array<int> -- indices of connected neurons(inputs)
|
588
|
+
|
589
|
+
This could be done more elegantly (e.g., defining more ruby ext classes).
|
590
|
+
This method does not directly correlate to anything in FANN, and accesses
|
591
|
+
structs that are not guaranteed to not change.
|
592
|
+
*/
|
593
|
+
static VALUE get_neurons(VALUE self, VALUE layer)
|
594
|
+
{
|
595
|
+
struct fann_layer *layer_it;
|
596
|
+
struct fann_neuron *neuron_it;
|
597
|
+
|
598
|
+
struct fann* f;
|
599
|
+
unsigned int i;
|
600
|
+
Data_Get_Struct (self, struct fann, f);
|
601
|
+
|
602
|
+
VALUE neuron_array = rb_ary_new();
|
603
|
+
VALUE activation_function_sym = ID2SYM(rb_intern("activation_function"));
|
604
|
+
VALUE activation_steepness_sym = ID2SYM(rb_intern("activation_steepness"));
|
605
|
+
VALUE layer_sym = ID2SYM(rb_intern("layer"));
|
606
|
+
VALUE sum_sym = ID2SYM(rb_intern("sum"));
|
607
|
+
VALUE value_sym = ID2SYM(rb_intern("value"));
|
608
|
+
VALUE connections_sym = ID2SYM(rb_intern("connections"));
|
609
|
+
unsigned int layer_num = 0;
|
610
|
+
for(layer_it = f->first_layer; layer_it != f->last_layer; layer_it++)
|
611
|
+
{
|
612
|
+
for(neuron_it = layer_it->first_neuron; neuron_it != layer_it->last_neuron; neuron_it++)
|
613
|
+
{
|
614
|
+
// Create array of connection indicies:
|
615
|
+
VALUE connection_array = rb_ary_new();
|
616
|
+
for (i = neuron_it->first_con; i < neuron_it->last_con; i++) {
|
617
|
+
rb_ary_push(connection_array, INT2NUM(f->connections[i] - f->first_layer->first_neuron));
|
618
|
+
}
|
619
|
+
|
620
|
+
VALUE neuron = rb_hash_new();
|
621
|
+
|
622
|
+
// Set attributes on hash & push on array:
|
623
|
+
rb_hash_aset(neuron, activation_function_sym, activation_function_to_sym(neuron_it->activation_function));
|
624
|
+
rb_hash_aset(neuron, activation_steepness_sym, rb_float_new(neuron_it->activation_steepness));
|
625
|
+
rb_hash_aset(neuron, layer_sym, INT2NUM(layer_num));
|
626
|
+
rb_hash_aset(neuron, sum_sym, rb_float_new(neuron_it->sum));
|
627
|
+
rb_hash_aset(neuron, value_sym, rb_float_new(neuron_it->value));
|
628
|
+
rb_hash_aset(neuron, connections_sym, connection_array);
|
629
|
+
|
630
|
+
rb_ary_push(neuron_array, neuron);
|
631
|
+
}
|
632
|
+
++layer_num;
|
633
|
+
}
|
634
|
+
|
635
|
+
return neuron_array;
|
636
|
+
}
|
637
|
+
|
638
|
+
/** Get list of layers in array format where each element contains number of neurons in that layer */
|
639
|
+
static VALUE get_layer_array(VALUE self)
|
640
|
+
{
|
641
|
+
struct fann* f;
|
642
|
+
unsigned int num_layers;
|
643
|
+
Data_Get_Struct (self, struct fann, f);
|
644
|
+
num_layers = fann_get_num_layers(f);
|
645
|
+
unsigned int layers[num_layers];
|
646
|
+
fann_get_layer_array(f, layers);
|
647
|
+
|
648
|
+
// Create ruby array & set outputs:
|
649
|
+
VALUE arr;
|
650
|
+
arr = rb_ary_new();
|
651
|
+
int i;
|
652
|
+
for (i=0; i<num_layers; i++)
|
653
|
+
{
|
654
|
+
rb_ary_push(arr, INT2NUM(layers[i]));
|
655
|
+
}
|
656
|
+
|
657
|
+
return arr;
|
658
|
+
}
|
659
|
+
|
660
|
+
/** Reads the mean square error from the network. */
|
661
|
+
static VALUE get_MSE(VALUE self)
|
662
|
+
{
|
663
|
+
RETURN_FANN_DBL(fann_get_MSE);
|
664
|
+
}
|
665
|
+
|
666
|
+
/** Resets the mean square error from the network.
|
667
|
+
This function also resets the number of bits that fail. */
|
668
|
+
static VALUE reset_MSE(VALUE self)
|
669
|
+
{
|
670
|
+
struct fann* f;
|
671
|
+
Data_Get_Struct (self, struct fann, f);
|
672
|
+
fann_reset_MSE(f);
|
673
|
+
}
|
674
|
+
|
675
|
+
/** Get the type of network. Returns as ruby symbol (one of :fann_nettype_shortcut, :fann_nettype_layer) */
|
676
|
+
static VALUE get_network_type(VALUE self)
|
677
|
+
{
|
678
|
+
struct fann* f;
|
679
|
+
enum fann_nettype_enum net_type;
|
680
|
+
VALUE ret_val;
|
681
|
+
Data_Get_Struct (self, struct fann, f);
|
682
|
+
|
683
|
+
net_type = fann_get_network_type(f);
|
684
|
+
|
685
|
+
if(net_type==FANN_NETTYPE_LAYER)
|
686
|
+
{
|
687
|
+
ret_val = ID2SYM(rb_intern("fann_nettype_layer")); // (rb_str_new2("FANN_NETTYPE_LAYER"));
|
688
|
+
}
|
689
|
+
else if(net_type==FANN_NETTYPE_SHORTCUT)
|
690
|
+
{
|
691
|
+
ret_val = ID2SYM(rb_intern("fann_nettype_shortcut")); // (rb_str_new2("FANN_NETTYPE_SHORTCUT"));
|
692
|
+
}
|
693
|
+
return ret_val;
|
694
|
+
}
|
695
|
+
|
696
|
+
/** Get the number of input neurons. */
|
697
|
+
static VALUE get_num_input(VALUE self)
|
698
|
+
{
|
699
|
+
RETURN_FANN_INT(fann_get_num_input);
|
700
|
+
}
|
701
|
+
|
702
|
+
/** Get the number of layers in the network. */
|
703
|
+
static VALUE get_num_layers(VALUE self)
|
704
|
+
{
|
705
|
+
RETURN_FANN_INT(fann_get_num_layers);
|
706
|
+
}
|
707
|
+
|
708
|
+
/** Get the number of output neurons. */
|
709
|
+
static VALUE get_num_output(VALUE self)
|
710
|
+
{
|
711
|
+
RETURN_FANN_INT(fann_get_num_output);
|
712
|
+
}
|
713
|
+
|
714
|
+
/** Get the total number of connections in the entire network. */
|
715
|
+
static VALUE get_total_connections(VALUE self)
|
716
|
+
{
|
717
|
+
RETURN_FANN_INT(fann_get_total_connections);
|
718
|
+
}
|
719
|
+
|
720
|
+
/** Get the total number of neurons in the entire network. */
|
721
|
+
static VALUE get_total_neurons(VALUE self)
|
722
|
+
{
|
723
|
+
RETURN_FANN_INT(fann_get_total_neurons);
|
724
|
+
}
|
725
|
+
|
726
|
+
/** Sets the error function used during training. One of the following symbols:
|
727
|
+
:linear, :tanh */
|
728
|
+
static VALUE set_train_error_function(VALUE self, VALUE train_error_function)
|
729
|
+
{
|
730
|
+
Check_Type(train_error_function, T_SYMBOL);
|
731
|
+
|
732
|
+
ID id=SYM2ID(train_error_function);
|
733
|
+
enum fann_errorfunc_enum fann_train_error_function;
|
734
|
+
|
735
|
+
if(id==rb_intern("linear")) {
|
736
|
+
fann_train_error_function = FANN_ERRORFUNC_LINEAR;
|
737
|
+
} else if(id==rb_intern("tanh")) {
|
738
|
+
fann_train_error_function = FANN_ERRORFUNC_TANH;
|
739
|
+
} else {
|
740
|
+
rb_raise(rb_eRuntimeError, "Unrecognized train error function: [%s]", rb_id2name(SYM2ID(train_error_function)));
|
741
|
+
}
|
742
|
+
|
743
|
+
struct fann* f;
|
744
|
+
Data_Get_Struct (self, struct fann, f);
|
745
|
+
fann_set_train_error_function(f, fann_train_error_function);
|
746
|
+
}
|
747
|
+
|
748
|
+
/** Returns the error function used during training. One of the following symbols:
|
749
|
+
:linear, :tanh */
|
750
|
+
static VALUE get_train_error_function(VALUE self)
|
751
|
+
{
|
752
|
+
struct fann* f;
|
753
|
+
enum fann_errorfunc_enum train_error;
|
754
|
+
VALUE ret_val;
|
755
|
+
Data_Get_Struct (self, struct fann, f);
|
756
|
+
|
757
|
+
train_error = fann_get_train_error_function(f);
|
758
|
+
|
759
|
+
if(train_error==FANN_ERRORFUNC_LINEAR)
|
760
|
+
{
|
761
|
+
ret_val = ID2SYM(rb_intern("linear"));
|
762
|
+
}
|
763
|
+
else if(train_error==FANN_ERRORFUNC_TANH)
|
764
|
+
{
|
765
|
+
ret_val = ID2SYM(rb_intern("tanh"));
|
766
|
+
}
|
767
|
+
return ret_val;
|
768
|
+
}
|
769
|
+
|
770
|
+
/** Set the training algorithm. One of the following symbols:
|
771
|
+
:incremental, :batch, :rprop, :quickprop */
|
772
|
+
static VALUE set_training_algorithm(VALUE self, VALUE train_error_function)
|
773
|
+
{
|
774
|
+
Check_Type(train_error_function, T_SYMBOL);
|
775
|
+
|
776
|
+
ID id=SYM2ID(train_error_function);
|
777
|
+
enum fann_train_enum fann_train_algorithm;
|
778
|
+
|
779
|
+
if(id==rb_intern("incremental")) {
|
780
|
+
fann_train_algorithm = FANN_TRAIN_INCREMENTAL;
|
781
|
+
} else if(id==rb_intern("batch")) {
|
782
|
+
fann_train_algorithm = FANN_TRAIN_BATCH;
|
783
|
+
} else if(id==rb_intern("rprop")) {
|
784
|
+
fann_train_algorithm = FANN_TRAIN_RPROP;
|
785
|
+
} else if(id==rb_intern("quickprop")) {
|
786
|
+
fann_train_algorithm = FANN_TRAIN_QUICKPROP;
|
787
|
+
} else {
|
788
|
+
rb_raise(rb_eRuntimeError, "Unrecognized training algorithm function: [%s]", rb_id2name(SYM2ID(train_error_function)));
|
789
|
+
}
|
790
|
+
|
791
|
+
struct fann* f;
|
792
|
+
Data_Get_Struct (self, struct fann, f);
|
793
|
+
fann_set_training_algorithm(f, fann_train_algorithm);
|
794
|
+
}
|
795
|
+
|
796
|
+
/** Returns the training algorithm. One of the following symbols:
|
797
|
+
:incremental, :batch, :rprop, :quickprop */
|
798
|
+
static VALUE get_training_algorithm(VALUE self)
|
799
|
+
{
|
800
|
+
struct fann* f;
|
801
|
+
enum fann_train_enum fann_train_algorithm;
|
802
|
+
VALUE ret_val;
|
803
|
+
Data_Get_Struct (self, struct fann, f);
|
804
|
+
|
805
|
+
fann_train_algorithm = fann_get_training_algorithm(f);
|
806
|
+
|
807
|
+
if(fann_train_algorithm==FANN_TRAIN_INCREMENTAL) {
|
808
|
+
ret_val = ID2SYM(rb_intern("incremental"));
|
809
|
+
} else if(fann_train_algorithm==FANN_TRAIN_BATCH) {
|
810
|
+
ret_val = ID2SYM(rb_intern("batch"));
|
811
|
+
} else if(fann_train_algorithm==FANN_TRAIN_RPROP) {
|
812
|
+
ret_val = ID2SYM(rb_intern("rprop"));
|
813
|
+
} else if(fann_train_algorithm==FANN_TRAIN_QUICKPROP) {
|
814
|
+
ret_val = ID2SYM(rb_intern("quickprop"));
|
815
|
+
}
|
816
|
+
return ret_val;
|
817
|
+
}
|
818
|
+
|
819
|
+
/** Set the training stop function. One of the following symbols:
|
820
|
+
:mse, :bit */
|
821
|
+
static VALUE set_train_stop_function(VALUE self, VALUE train_stop_function)
|
822
|
+
{
|
823
|
+
Check_Type(train_stop_function, T_SYMBOL);
|
824
|
+
ID id=SYM2ID(train_stop_function);
|
825
|
+
enum fann_stopfunc_enum fann_train_stop_function;
|
826
|
+
|
827
|
+
if(id==rb_intern("mse")) {
|
828
|
+
fann_train_stop_function = FANN_STOPFUNC_MSE;
|
829
|
+
} else if(id==rb_intern("bit")) {
|
830
|
+
fann_train_stop_function = FANN_STOPFUNC_BIT;
|
831
|
+
} else {
|
832
|
+
rb_raise(rb_eRuntimeError, "Unrecognized stop function: [%s]", rb_id2name(SYM2ID(train_stop_function)));
|
833
|
+
}
|
834
|
+
|
835
|
+
struct fann* f;
|
836
|
+
Data_Get_Struct (self, struct fann, f);
|
837
|
+
fann_set_train_stop_function(f, fann_train_stop_function);
|
838
|
+
}
|
839
|
+
|
840
|
+
/** Returns the training stop function. One of the following symbols:
|
841
|
+
:mse, :bit */
|
842
|
+
static VALUE get_train_stop_function(VALUE self)
|
843
|
+
{
|
844
|
+
struct fann* f;
|
845
|
+
enum fann_stopfunc_enum train_stop;
|
846
|
+
VALUE ret_val;
|
847
|
+
Data_Get_Struct (self, struct fann, f);
|
848
|
+
|
849
|
+
train_stop = fann_get_train_stop_function(f);
|
850
|
+
|
851
|
+
if(train_stop==FANN_STOPFUNC_MSE)
|
852
|
+
{
|
853
|
+
ret_val = ID2SYM(rb_intern("mse")); // (rb_str_new2("FANN_NETTYPE_LAYER"));
|
854
|
+
}
|
855
|
+
else if(train_stop==FANN_STOPFUNC_BIT)
|
856
|
+
{
|
857
|
+
ret_val = ID2SYM(rb_intern("bit")); // (rb_str_new2("FANN_NETTYPE_SHORTCUT"));
|
858
|
+
}
|
859
|
+
return ret_val;
|
860
|
+
}
|
861
|
+
|
862
|
+
|
863
|
+
/** Will print the connections of the ann in a compact matrix,
|
864
|
+
for easy viewing of the internals of the ann. */
|
865
|
+
static VALUE print_connections(VALUE self)
|
866
|
+
{
|
867
|
+
struct fann* f;
|
868
|
+
Data_Get_Struct (self, struct fann, f);
|
869
|
+
fann_print_connections(f);
|
870
|
+
}
|
871
|
+
|
872
|
+
/** Print current NN parameters to stdout */
|
873
|
+
static VALUE print_parameters(VALUE self)
|
874
|
+
{
|
875
|
+
struct fann* f;
|
876
|
+
Data_Get_Struct (self, struct fann, f);
|
877
|
+
fann_print_parameters(f);
|
878
|
+
return Qnil;
|
879
|
+
}
|
880
|
+
|
881
|
+
/** Give each connection a random weight between *min_weight* and *max_weight* */
|
882
|
+
static VALUE randomize_weights(VALUE self, VALUE min_weight, VALUE max_weight)
|
883
|
+
{
|
884
|
+
Check_Type(min_weight, T_FLOAT);
|
885
|
+
Check_Type(max_weight, T_FLOAT);
|
886
|
+
struct fann* f;
|
887
|
+
Data_Get_Struct (self, struct fann, f);
|
888
|
+
fann_randomize_weights(f, NUM2DBL(min_weight), NUM2DBL(max_weight));
|
889
|
+
}
|
890
|
+
|
891
|
+
/** Run neural net with current parameters */
|
892
|
+
static VALUE run (VALUE self, VALUE inputs)
|
893
|
+
{
|
894
|
+
Check_Type(inputs, T_ARRAY);
|
895
|
+
|
896
|
+
struct fann* f;
|
897
|
+
int i;
|
898
|
+
fann_type* outputs;
|
899
|
+
|
900
|
+
// Convert inputs to type needed for NN:
|
901
|
+
unsigned int len = RARRAY(inputs)->len;
|
902
|
+
fann_type fann_inputs[len];
|
903
|
+
for (i=0; i<len; i++)
|
904
|
+
{
|
905
|
+
fann_inputs[i] = NUM2DBL(RARRAY(inputs)->ptr[i]);
|
906
|
+
}
|
907
|
+
|
908
|
+
|
909
|
+
// Obtain NN & run method:
|
910
|
+
Data_Get_Struct (self, struct fann, f);
|
911
|
+
outputs = fann_run(f, fann_inputs);
|
912
|
+
|
913
|
+
// Create ruby array & set outputs:
|
914
|
+
VALUE arr;
|
915
|
+
arr = rb_ary_new();
|
916
|
+
unsigned int output_len=fann_get_num_output(f);
|
917
|
+
for (i=0; i<output_len; i++)
|
918
|
+
{
|
919
|
+
rb_ary_push(arr, rb_float_new(outputs[i]));
|
920
|
+
}
|
921
|
+
|
922
|
+
return arr;
|
923
|
+
}
|
924
|
+
|
925
|
+
/** Initialize the weights using Widrow + Nguyen's algorithm. */
|
926
|
+
static VALUE init_weights(VALUE self, VALUE train_data)
|
927
|
+
{
|
928
|
+
|
929
|
+
Check_Type(train_data, T_DATA);
|
930
|
+
|
931
|
+
struct fann* f;
|
932
|
+
struct fann_train_data* t;
|
933
|
+
Data_Get_Struct (self, struct fann, f);
|
934
|
+
Data_Get_Struct (train_data, struct fann_train_data, t);
|
935
|
+
|
936
|
+
fann_init_weights(f, t);
|
937
|
+
}
|
938
|
+
|
939
|
+
|
940
|
+
|
941
|
+
/** Train with training data created with RubyFann::TrainData.new
|
942
|
+
max_epochs - The maximum number of epochs the training should continue
|
943
|
+
epochs_between_reports - The number of epochs between printing a status report to stdout.
|
944
|
+
desired_error - The desired <get_MSE> or <get_bit_fail>, depending on which stop function
|
945
|
+
is chosen by <set_train_stop_function>. */
|
946
|
+
static VALUE train_on_data(VALUE self, VALUE train_data, VALUE max_epochs, VALUE epochs_between_reports, VALUE desired_error)
|
947
|
+
{
|
948
|
+
Check_Type(train_data, T_DATA);
|
949
|
+
Check_Type(max_epochs, T_FIXNUM);
|
950
|
+
Check_Type(epochs_between_reports, T_FIXNUM);
|
951
|
+
Check_Type(desired_error, T_FLOAT);
|
952
|
+
|
953
|
+
struct fann* f;
|
954
|
+
struct fann_train_data* t;
|
955
|
+
Data_Get_Struct (self, struct fann, f);
|
956
|
+
Data_Get_Struct (train_data, struct fann_train_data, t);
|
957
|
+
|
958
|
+
unsigned int fann_max_epochs = NUM2INT(max_epochs);
|
959
|
+
unsigned int fann_epochs_between_reports = NUM2INT(epochs_between_reports);
|
960
|
+
float fann_desired_error = NUM2DBL(desired_error);
|
961
|
+
fann_train_on_data(f, t, fann_max_epochs, fann_epochs_between_reports, fann_desired_error);
|
962
|
+
}
|
963
|
+
|
964
|
+
/** Train one epoch with a set of training data, created with RubyFann::TrainData.new */
|
965
|
+
static VALUE train_epoch(VALUE self, VALUE train_data)
|
966
|
+
{
|
967
|
+
Check_Type(train_data, T_DATA);
|
968
|
+
struct fann* f;
|
969
|
+
struct fann_train_data* t;
|
970
|
+
Data_Get_Struct (self, struct fann, f);
|
971
|
+
Data_Get_Struct (train_data, struct fann_train_data, t);
|
972
|
+
return rb_float_new(fann_train_epoch(f, t));
|
973
|
+
}
|
974
|
+
|
975
|
+
/** Test a set of training data and calculates the MSE for the training data. */
|
976
|
+
static VALUE test_data(VALUE self, VALUE train_data)
|
977
|
+
{
|
978
|
+
Check_Type(train_data, T_DATA);
|
979
|
+
struct fann* f;
|
980
|
+
struct fann_train_data* t;
|
981
|
+
Data_Get_Struct (self, struct fann, f);
|
982
|
+
Data_Get_Struct (train_data, struct fann_train_data, t);
|
983
|
+
return rb_float_new(fann_test_data(f, t));
|
984
|
+
}
|
985
|
+
|
986
|
+
// Returns the position of the decimal point in the ann.
|
987
|
+
// Only available in fixed-point mode, which we don't need:
|
988
|
+
// static VALUE get_decimal_point(VALUE self)
|
989
|
+
// {
|
990
|
+
// struct fann* f;
|
991
|
+
// Data_Get_Struct (self, struct fann, f);
|
992
|
+
// return INT2NUM(fann_get_decimal_point(f));
|
993
|
+
// }
|
994
|
+
|
995
|
+
// returns the multiplier that fix point data is multiplied with.
|
996
|
+
|
997
|
+
// Only available in fixed-point mode, which we don't need:
|
998
|
+
// static VALUE get_multiplier(VALUE self)
|
999
|
+
// {
|
1000
|
+
// struct fann* f;
|
1001
|
+
// Data_Get_Struct (self, struct fann, f);
|
1002
|
+
// return INT2NUM(fann_get_multiplier(f));
|
1003
|
+
// }
|
1004
|
+
|
1005
|
+
// Train with inputs (array of arrays) & desired_outputs (array of arrays)
|
1006
|
+
// inputs & desired outputs should be of same length
|
1007
|
+
// All sub-arrays on inputs should be of same length
|
1008
|
+
// All sub-arrays on desired outputs should be of same length
|
1009
|
+
// Sub-arrays on inputs & desired_outputs can be different sizes from one another
|
1010
|
+
static VALUE cascadetrain_on_data(VALUE self, VALUE train_data, VALUE max_neurons, VALUE neurons_between_reports, VALUE desired_error)
|
1011
|
+
{
|
1012
|
+
Check_Type(train_data, T_DATA);
|
1013
|
+
Check_Type(max_neurons, T_FIXNUM);
|
1014
|
+
Check_Type(neurons_between_reports, T_FIXNUM);
|
1015
|
+
Check_Type(desired_error, T_FLOAT);
|
1016
|
+
|
1017
|
+
struct fann* f;
|
1018
|
+
struct fann_train_data* t;
|
1019
|
+
Data_Get_Struct (self, struct fann, f);
|
1020
|
+
Data_Get_Struct (train_data, struct fann_train_data, t);
|
1021
|
+
|
1022
|
+
unsigned int fann_max_neurons = NUM2INT(max_neurons);
|
1023
|
+
unsigned int fann_neurons_between_reports = NUM2INT(neurons_between_reports);
|
1024
|
+
float fann_desired_error = NUM2DBL(desired_error);
|
1025
|
+
|
1026
|
+
fann_cascadetrain_on_data(f, t, fann_max_neurons, fann_neurons_between_reports, fann_desired_error);
|
1027
|
+
|
1028
|
+
}
|
1029
|
+
|
1030
|
+
/** The cascade output change fraction is a number between 0 and 1 */
|
1031
|
+
static VALUE get_cascade_output_change_fraction(VALUE self)
|
1032
|
+
{
|
1033
|
+
RETURN_FANN_FLT(fann_get_cascade_output_change_fraction);
|
1034
|
+
}
|
1035
|
+
|
1036
|
+
/** The cascade output change fraction is a number between 0 and 1 */
|
1037
|
+
static VALUE set_cascade_output_change_fraction(VALUE self, VALUE cascade_output_change_fraction)
|
1038
|
+
{
|
1039
|
+
SET_FANN_FLT(cascade_output_change_fraction, fann_set_cascade_output_change_fraction);
|
1040
|
+
}
|
1041
|
+
|
1042
|
+
/** The number of cascade output stagnation epochs determines the number of epochs training is allowed to
|
1043
|
+
continue without changing the MSE by a fraction of <get_cascade_output_change_fraction>. */
|
1044
|
+
static VALUE get_cascade_output_stagnation_epochs(VALUE self)
|
1045
|
+
{
|
1046
|
+
RETURN_FANN_INT(fann_get_cascade_output_stagnation_epochs);
|
1047
|
+
}
|
1048
|
+
|
1049
|
+
/** The number of cascade output stagnation epochs determines the number of epochs training is allowed to
|
1050
|
+
continue without changing the MSE by a fraction of <get_cascade_output_change_fraction>. */
|
1051
|
+
static VALUE set_cascade_output_stagnation_epochs(VALUE self, VALUE cascade_output_stagnation_epochs)
|
1052
|
+
{
|
1053
|
+
SET_FANN_INT(cascade_output_stagnation_epochs, fann_set_cascade_output_stagnation_epochs);
|
1054
|
+
}
|
1055
|
+
|
1056
|
+
/** The cascade candidate change fraction is a number between 0 and 1 */
|
1057
|
+
static VALUE get_cascade_candidate_change_fraction(VALUE self)
|
1058
|
+
{
|
1059
|
+
RETURN_FANN_FLT(fann_get_cascade_candidate_change_fraction);
|
1060
|
+
}
|
1061
|
+
|
1062
|
+
/** The cascade candidate change fraction is a number between 0 and 1 */
|
1063
|
+
static VALUE set_cascade_candidate_change_fraction(VALUE self, VALUE cascade_candidate_change_fraction)
|
1064
|
+
{
|
1065
|
+
SET_FANN_FLT(cascade_candidate_change_fraction, fann_set_cascade_candidate_change_fraction);
|
1066
|
+
}
|
1067
|
+
|
1068
|
+
/** The number of cascade candidate stagnation epochs determines the number of epochs training is allowed to
|
1069
|
+
continue without changing the MSE by a fraction of <get_cascade_candidate_change_fraction>. */
|
1070
|
+
static VALUE get_cascade_candidate_stagnation_epochs(VALUE self)
|
1071
|
+
{
|
1072
|
+
RETURN_FANN_UINT(fann_get_cascade_candidate_stagnation_epochs);
|
1073
|
+
}
|
1074
|
+
|
1075
|
+
/** The number of cascade candidate stagnation epochs determines the number of epochs training is allowed to
|
1076
|
+
continue without changing the MSE by a fraction of <get_cascade_candidate_change_fraction>. */
|
1077
|
+
static VALUE set_cascade_candidate_stagnation_epochs(VALUE self, VALUE cascade_candidate_stagnation_epochs)
|
1078
|
+
{
|
1079
|
+
SET_FANN_UINT(cascade_candidate_stagnation_epochs, fann_set_cascade_candidate_stagnation_epochs);
|
1080
|
+
}
|
1081
|
+
|
1082
|
+
/** The weight multiplier is a parameter which is used to multiply the weights from the candidate neuron
|
1083
|
+
before adding the neuron to the neural network. This parameter is usually between 0 and 1, and is used
|
1084
|
+
to make the training a bit less aggressive. */
|
1085
|
+
static VALUE get_cascade_weight_multiplier(VALUE self)
|
1086
|
+
{
|
1087
|
+
RETURN_FANN_DBL(fann_get_cascade_weight_multiplier);
|
1088
|
+
}
|
1089
|
+
|
1090
|
+
/** The weight multiplier is a parameter which is used to multiply the weights from the candidate neuron
|
1091
|
+
before adding the neuron to the neural network. This parameter is usually between 0 and 1, and is used
|
1092
|
+
to make the training a bit less aggressive. */
|
1093
|
+
static VALUE set_cascade_weight_multiplier(VALUE self, VALUE cascade_weight_multiplier)
|
1094
|
+
{
|
1095
|
+
SET_FANN_DBL(cascade_weight_multiplier, fann_set_cascade_weight_multiplier);
|
1096
|
+
}
|
1097
|
+
|
1098
|
+
/** The candidate limit is a limit for how much the candidate neuron may be trained.
|
1099
|
+
The limit is a limit on the proportion between the MSE and candidate score. */
|
1100
|
+
static VALUE get_cascade_candidate_limit(VALUE self)
|
1101
|
+
{
|
1102
|
+
RETURN_FANN_DBL(fann_get_cascade_candidate_limit);
|
1103
|
+
}
|
1104
|
+
|
1105
|
+
/** The candidate limit is a limit for how much the candidate neuron may be trained.
|
1106
|
+
The limit is a limit on the proportion between the MSE and candidate score. */
|
1107
|
+
static VALUE set_cascade_candidate_limit(VALUE self, VALUE cascade_candidate_limit)
|
1108
|
+
{
|
1109
|
+
SET_FANN_DBL(cascade_candidate_limit, fann_set_cascade_candidate_limit);
|
1110
|
+
}
|
1111
|
+
|
1112
|
+
/** The maximum out epochs determines the maximum number of epochs the output connections
|
1113
|
+
may be trained after adding a new candidate neuron. */
|
1114
|
+
static VALUE get_cascade_max_out_epochs(VALUE self)
|
1115
|
+
{
|
1116
|
+
RETURN_FANN_UINT(fann_get_cascade_max_out_epochs);
|
1117
|
+
}
|
1118
|
+
|
1119
|
+
/** The maximum out epochs determines the maximum number of epochs the output connections
|
1120
|
+
may be trained after adding a new candidate neuron. */
|
1121
|
+
static VALUE set_cascade_max_out_epochs(VALUE self, VALUE cascade_max_out_epochs)
|
1122
|
+
{
|
1123
|
+
SET_FANN_UINT(cascade_max_out_epochs, fann_set_cascade_max_out_epochs);
|
1124
|
+
}
|
1125
|
+
|
1126
|
+
/** The maximum candidate epochs determines the maximum number of epochs the input
|
1127
|
+
connections to the candidates may be trained before adding a new candidate neuron. */
|
1128
|
+
static VALUE get_cascade_max_cand_epochs(VALUE self)
|
1129
|
+
{
|
1130
|
+
RETURN_FANN_UINT(fann_get_cascade_max_cand_epochs);
|
1131
|
+
}
|
1132
|
+
|
1133
|
+
/** The maximum candidate epochs determines the maximum number of epochs the input
|
1134
|
+
connections to the candidates may be trained before adding a new candidate neuron. */
|
1135
|
+
static VALUE set_cascade_max_cand_epochs(VALUE self, VALUE cascade_max_cand_epochs)
|
1136
|
+
{
|
1137
|
+
SET_FANN_UINT(cascade_max_cand_epochs, fann_set_cascade_max_cand_epochs);
|
1138
|
+
}
|
1139
|
+
|
1140
|
+
/** The number of candidates used during training (calculated by multiplying <get_cascade_activation_functions_count>,
|
1141
|
+
<get_cascade_activation_steepnesses_count> and <get_cascade_num_candidate_groups>). */
|
1142
|
+
static VALUE get_cascade_num_candidates(VALUE self)
|
1143
|
+
{
|
1144
|
+
RETURN_FANN_UINT(fann_get_cascade_num_candidates);
|
1145
|
+
}
|
1146
|
+
|
1147
|
+
/** The number of activation functions in the <get_cascade_activation_functions> array */
|
1148
|
+
static VALUE get_cascade_activation_functions_count(VALUE self)
|
1149
|
+
{
|
1150
|
+
RETURN_FANN_UINT(fann_get_cascade_activation_functions_count);
|
1151
|
+
}
|
1152
|
+
|
1153
|
+
/** The learning rate is used to determine how aggressive training should be for some of the
|
1154
|
+
training algorithms (:incremental, :batch, :quickprop).
|
1155
|
+
Do however note that it is not used in :rprop.
|
1156
|
+
The default learning rate is 0.7. */
|
1157
|
+
static VALUE get_learning_rate(VALUE self)
|
1158
|
+
{
|
1159
|
+
RETURN_FANN_FLT(fann_get_learning_rate);
|
1160
|
+
}
|
1161
|
+
|
1162
|
+
/** The learning rate is used to determine how aggressive training should be for some of the
|
1163
|
+
training algorithms (:incremental, :batch, :quickprop).
|
1164
|
+
Do however note that it is not used in :rprop.
|
1165
|
+
The default learning rate is 0.7. */
|
1166
|
+
static VALUE set_learning_rate(VALUE self, VALUE learning_rate)
|
1167
|
+
{
|
1168
|
+
SET_FANN_FLT(learning_rate, fann_set_learning_rate);
|
1169
|
+
}
|
1170
|
+
|
1171
|
+
/** Get the learning momentum. */
|
1172
|
+
static VALUE get_learning_momentum(VALUE self)
|
1173
|
+
{
|
1174
|
+
RETURN_FANN_FLT(fann_get_learning_momentum);
|
1175
|
+
}
|
1176
|
+
|
1177
|
+
/** Set the learning momentum. */
|
1178
|
+
static VALUE set_learning_momentum(VALUE self, VALUE learning_momentum)
|
1179
|
+
{
|
1180
|
+
SET_FANN_FLT(learning_momentum, fann_set_learning_momentum);
|
1181
|
+
}
|
1182
|
+
|
1183
|
+
/** The cascade activation functions is an array of the different activation functions used by
|
1184
|
+
the candidates. The default is [:sigmoid, :sigmoid_symmetric, :gaussian, :gaussian_symmetric, :elliot, :elliot_symmetric] */
|
1185
|
+
static VALUE set_cascade_activation_functions(VALUE self, VALUE cascade_activation_functions)
|
1186
|
+
{
|
1187
|
+
Check_Type(cascade_activation_functions, T_ARRAY);
|
1188
|
+
struct fann* f;
|
1189
|
+
Data_Get_Struct (self, struct fann, f);
|
1190
|
+
|
1191
|
+
unsigned int cnt = RARRAY(cascade_activation_functions)->len;
|
1192
|
+
enum fann_activationfunc_enum fann_activation_functions[cnt];
|
1193
|
+
int i;
|
1194
|
+
for (i=0; i<cnt; i++)
|
1195
|
+
{
|
1196
|
+
fann_activation_functions[i] = sym_to_activation_function(RARRAY(cascade_activation_functions)->ptr[i]);
|
1197
|
+
}
|
1198
|
+
|
1199
|
+
fann_set_cascade_activation_functions(f, fann_activation_functions, cnt);
|
1200
|
+
}
|
1201
|
+
|
1202
|
+
/** The cascade activation functions is an array of the different activation functions used by
|
1203
|
+
the candidates. The default is [:sigmoid, :sigmoid_symmetric, :gaussian, :gaussian_symmetric, :elliot, :elliot_symmetric] */
|
1204
|
+
static VALUE get_cascade_activation_functions(VALUE self)
|
1205
|
+
{
|
1206
|
+
struct fann* f;
|
1207
|
+
Data_Get_Struct (self, struct fann, f);
|
1208
|
+
unsigned int cnt = fann_get_cascade_activation_functions_count(f);
|
1209
|
+
enum fann_activationfunc_enum* fann_functions = fann_get_cascade_activation_functions(f);
|
1210
|
+
|
1211
|
+
// Create ruby array & set outputs:
|
1212
|
+
VALUE arr;
|
1213
|
+
arr = rb_ary_new();
|
1214
|
+
int i;
|
1215
|
+
for (i=0; i<cnt; i++)
|
1216
|
+
{
|
1217
|
+
rb_ary_push(arr, activation_function_to_sym(fann_functions[i]));
|
1218
|
+
}
|
1219
|
+
|
1220
|
+
return arr;
|
1221
|
+
}
|
1222
|
+
|
1223
|
+
/** The number of activation steepnesses in the <get_cascade_activation_functions> array. */
|
1224
|
+
static VALUE get_cascade_activation_steepnesses_count(VALUE self)
|
1225
|
+
{
|
1226
|
+
RETURN_FANN_UINT(fann_get_cascade_activation_steepnesses_count);
|
1227
|
+
}
|
1228
|
+
|
1229
|
+
/** The number of candidate groups is the number of groups of identical candidates which will be used
|
1230
|
+
during training. */
|
1231
|
+
static VALUE get_cascade_num_candidate_groups(VALUE self)
|
1232
|
+
{
|
1233
|
+
RETURN_FANN_UINT(fann_get_cascade_num_candidate_groups);
|
1234
|
+
}
|
1235
|
+
|
1236
|
+
/** The number of candidate groups is the number of groups of identical candidates which will be used
|
1237
|
+
during training. */
|
1238
|
+
static VALUE set_cascade_num_candidate_groups(VALUE self, VALUE cascade_num_candidate_groups)
|
1239
|
+
{
|
1240
|
+
SET_FANN_UINT(cascade_num_candidate_groups, fann_set_cascade_num_candidate_groups);
|
1241
|
+
}
|
1242
|
+
|
1243
|
+
/** The cascade activation steepnesses array is an array of the different activation functions used by
|
1244
|
+
the candidates. */
|
1245
|
+
static VALUE set_cascade_activation_steepnesses(VALUE self, VALUE cascade_activation_steepnesses)
|
1246
|
+
{
|
1247
|
+
Check_Type(cascade_activation_steepnesses, T_ARRAY);
|
1248
|
+
struct fann* f;
|
1249
|
+
Data_Get_Struct (self, struct fann, f);
|
1250
|
+
|
1251
|
+
unsigned int cnt = RARRAY(cascade_activation_steepnesses)->len;
|
1252
|
+
fann_type fann_activation_steepnesses[cnt];
|
1253
|
+
int i;
|
1254
|
+
for (i=0; i<cnt; i++)
|
1255
|
+
{
|
1256
|
+
fann_activation_steepnesses[i] = NUM2DBL(RARRAY(cascade_activation_steepnesses)->ptr[i]);
|
1257
|
+
}
|
1258
|
+
|
1259
|
+
fann_set_cascade_activation_steepnesses(f, fann_activation_steepnesses, cnt);
|
1260
|
+
}
|
1261
|
+
|
1262
|
+
/** The cascade activation steepnesses array is an array of the different activation functions used by
|
1263
|
+
the candidates. */
|
1264
|
+
static VALUE get_cascade_activation_steepnesses(VALUE self)
|
1265
|
+
{
|
1266
|
+
struct fann* f;
|
1267
|
+
Data_Get_Struct (self, struct fann, f);
|
1268
|
+
fann_type* fann_steepnesses = fann_get_cascade_activation_steepnesses(f);
|
1269
|
+
unsigned int cnt = fann_get_cascade_activation_steepnesses_count(f);
|
1270
|
+
|
1271
|
+
// Create ruby array & set outputs:
|
1272
|
+
VALUE arr;
|
1273
|
+
arr = rb_ary_new();
|
1274
|
+
int i;
|
1275
|
+
for (i=0; i<cnt; i++)
|
1276
|
+
{
|
1277
|
+
rb_ary_push(arr, rb_float_new(fann_steepnesses[i]));
|
1278
|
+
}
|
1279
|
+
|
1280
|
+
return arr;
|
1281
|
+
}
|
1282
|
+
|
1283
|
+
/** Save the entire network to configuration file with given name */
|
1284
|
+
static VALUE nn_save(VALUE self, VALUE filename)
|
1285
|
+
{
|
1286
|
+
struct fann* f;
|
1287
|
+
Data_Get_Struct (self, struct fann, f);
|
1288
|
+
int status = fann_save(f, StringValuePtr(filename));
|
1289
|
+
return INT2NUM(status);
|
1290
|
+
}
|
1291
|
+
|
1292
|
+
/** Initializes classes under RubyFann module/namespace. */
|
1293
|
+
void Init_neural_network ()
|
1294
|
+
{
|
1295
|
+
// RubyFann module/namespace:
|
1296
|
+
m_rb_fann_module = rb_define_module ("RubyFann");
|
1297
|
+
|
1298
|
+
// Standard NN class:
|
1299
|
+
m_rb_fann_standard_class = rb_define_class_under (m_rb_fann_module, "Standard", rb_cObject);
|
1300
|
+
rb_define_alloc_func (m_rb_fann_standard_class, fann_allocate);
|
1301
|
+
rb_define_method(m_rb_fann_standard_class, "initialize", fann_initialize, 1);
|
1302
|
+
rb_define_method(m_rb_fann_standard_class, "init_weights", init_weights, 1);
|
1303
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_function", set_activation_function, 3);
|
1304
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_function_hidden", set_activation_function_hidden, 1);
|
1305
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_function_layer", set_activation_function_layer, 1);
|
1306
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_function_output", set_activation_function_output, 1);
|
1307
|
+
rb_define_method(m_rb_fann_standard_class, "get_activation_steepness", get_activation_steepness, 2);
|
1308
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_steepness", set_activation_steepness, 3);
|
1309
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_steepness_hidden", set_activation_steepness_hidden, 1);
|
1310
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_steepness_layer", set_activation_steepness_layer, 2);
|
1311
|
+
rb_define_method(m_rb_fann_standard_class, "set_activation_steepness_output", set_activation_steepness_output, 1);
|
1312
|
+
rb_define_method(m_rb_fann_standard_class, "get_train_error_function", get_train_error_function, 0);
|
1313
|
+
rb_define_method(m_rb_fann_standard_class, "set_train_error_function", set_train_error_function, 1);
|
1314
|
+
rb_define_method(m_rb_fann_standard_class, "get_train_stop_function", get_train_stop_function, 0);
|
1315
|
+
rb_define_method(m_rb_fann_standard_class, "set_train_stop_function", set_train_stop_function, 1);
|
1316
|
+
rb_define_method(m_rb_fann_standard_class, "get_bit_fail_limit", get_bit_fail_limit, 0);
|
1317
|
+
rb_define_method(m_rb_fann_standard_class, "set_bit_fail_limit", set_bit_fail_limit, 1);
|
1318
|
+
rb_define_method(m_rb_fann_standard_class, "get_quickprop_decay", get_quickprop_decay, 0);
|
1319
|
+
rb_define_method(m_rb_fann_standard_class, "set_quickprop_decay", set_quickprop_decay, 1);
|
1320
|
+
rb_define_method(m_rb_fann_standard_class, "get_quickprop_mu", get_quickprop_mu, 0);
|
1321
|
+
rb_define_method(m_rb_fann_standard_class, "set_quickprop_mu", set_quickprop_mu, 1);
|
1322
|
+
rb_define_method(m_rb_fann_standard_class, "get_rprop_increase_factor", get_rprop_increase_factor, 0);
|
1323
|
+
rb_define_method(m_rb_fann_standard_class, "set_rprop_increase_factor", set_rprop_increase_factor, 1);
|
1324
|
+
rb_define_method(m_rb_fann_standard_class, "get_rprop_decrease_factor", get_rprop_decrease_factor, 0);
|
1325
|
+
rb_define_method(m_rb_fann_standard_class, "set_rprop_decrease_factor", set_rprop_decrease_factor, 1);
|
1326
|
+
rb_define_method(m_rb_fann_standard_class, "get_rprop_delta_max", get_rprop_delta_max, 0);
|
1327
|
+
rb_define_method(m_rb_fann_standard_class, "set_rprop_delta_max", set_rprop_delta_max, 1);
|
1328
|
+
rb_define_method(m_rb_fann_standard_class, "get_rprop_delta_min", get_rprop_delta_min, 0);
|
1329
|
+
rb_define_method(m_rb_fann_standard_class, "set_rprop_delta_min", set_rprop_delta_min, 1);
|
1330
|
+
rb_define_method(m_rb_fann_standard_class, "get_rprop_delta_zero", get_rprop_delta_zero, 0);
|
1331
|
+
rb_define_method(m_rb_fann_standard_class, "set_rprop_delta_zero", set_rprop_delta_zero, 1);
|
1332
|
+
rb_define_method(m_rb_fann_standard_class, "get_bias_array", get_bias_array, 0);
|
1333
|
+
rb_define_method(m_rb_fann_standard_class, "get_connection_rate", get_connection_rate, 0);
|
1334
|
+
rb_define_method(m_rb_fann_standard_class, "get_layer_array", get_layer_array, 0);
|
1335
|
+
rb_define_method(m_rb_fann_standard_class, "get_network_type", get_network_type, 0);
|
1336
|
+
rb_define_method(m_rb_fann_standard_class, "get_neurons", get_neurons, 0);
|
1337
|
+
rb_define_method(m_rb_fann_standard_class, "get_num_input", get_num_input, 0);
|
1338
|
+
rb_define_method(m_rb_fann_standard_class, "get_num_layers", get_num_layers, 0);
|
1339
|
+
rb_define_method(m_rb_fann_standard_class, "get_num_output", get_num_output, 0);
|
1340
|
+
rb_define_method(m_rb_fann_standard_class, "get_total_connections", get_total_connections, 0);
|
1341
|
+
rb_define_method(m_rb_fann_standard_class, "get_total_neurons", get_total_neurons, 0);
|
1342
|
+
rb_define_method(m_rb_fann_standard_class, "get_train_error_function", get_train_error_function, 0);
|
1343
|
+
rb_define_method(m_rb_fann_standard_class, "set_train_error_function", set_train_error_function, 1);
|
1344
|
+
rb_define_method(m_rb_fann_standard_class, "print_connections", print_connections, 0);
|
1345
|
+
rb_define_method(m_rb_fann_standard_class, "print_parameters", print_parameters, 0);
|
1346
|
+
rb_define_method(m_rb_fann_standard_class, "randomize_weights", randomize_weights, 2);
|
1347
|
+
rb_define_method(m_rb_fann_standard_class, "run", run, 1);
|
1348
|
+
rb_define_method(m_rb_fann_standard_class, "train_on_data", train_on_data, 4);
|
1349
|
+
rb_define_method(m_rb_fann_standard_class, "train_epoch", train_epoch, 1);
|
1350
|
+
rb_define_method(m_rb_fann_standard_class, "test_data", test_data, 1);
|
1351
|
+
rb_define_method(m_rb_fann_standard_class, "get_MSE", get_MSE, 0);
|
1352
|
+
rb_define_method(m_rb_fann_standard_class, "get_bit_fail", get_bit_fail, 0);
|
1353
|
+
rb_define_method(m_rb_fann_standard_class, "reset_MSE", reset_MSE, 0);
|
1354
|
+
rb_define_method(m_rb_fann_standard_class, "get_learning_rate", get_learning_rate, 0);
|
1355
|
+
rb_define_method(m_rb_fann_standard_class, "set_learning_rate", set_learning_rate, 1);
|
1356
|
+
rb_define_method(m_rb_fann_standard_class, "get_learning_momentum", get_learning_momentum, 0);
|
1357
|
+
rb_define_method(m_rb_fann_standard_class, "set_learning_momentum", set_learning_momentum, 1);
|
1358
|
+
rb_define_method(m_rb_fann_standard_class, "get_training_algorithm", get_training_algorithm, 0);
|
1359
|
+
rb_define_method(m_rb_fann_standard_class, "set_training_algorithm", set_training_algorithm, 1);
|
1360
|
+
|
1361
|
+
|
1362
|
+
// Cascade functions:
|
1363
|
+
rb_define_method(m_rb_fann_standard_class, "cascadetrain_on_data", cascadetrain_on_data, 4);
|
1364
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_output_change_fraction", get_cascade_output_change_fraction, 0);
|
1365
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_output_change_fraction", set_cascade_output_change_fraction, 1);
|
1366
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_output_stagnation_epochs", get_cascade_output_stagnation_epochs, 0);
|
1367
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_output_stagnation_epochs", set_cascade_output_stagnation_epochs, 1);
|
1368
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_candidate_change_fraction", get_cascade_candidate_change_fraction, 0);
|
1369
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_candidate_change_fraction", set_cascade_candidate_change_fraction, 1);
|
1370
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_candidate_stagnation_epochs", get_cascade_candidate_stagnation_epochs, 0);
|
1371
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_candidate_stagnation_epochs", set_cascade_candidate_stagnation_epochs, 1);
|
1372
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_weight_multiplier", get_cascade_weight_multiplier, 0);
|
1373
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_weight_multiplier", set_cascade_weight_multiplier, 1);
|
1374
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_candidate_limit", get_cascade_candidate_limit, 0);
|
1375
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_candidate_limit", set_cascade_candidate_limit, 1);
|
1376
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_max_out_epochs", get_cascade_max_out_epochs, 0);
|
1377
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_max_out_epochs", set_cascade_max_out_epochs, 1);
|
1378
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_max_cand_epochs", get_cascade_max_cand_epochs, 0);
|
1379
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_max_cand_epochs", set_cascade_max_cand_epochs, 1);
|
1380
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_num_candidates", get_cascade_num_candidates, 0);
|
1381
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_activation_functions_count", get_cascade_activation_functions_count, 0);
|
1382
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_activation_functions", get_cascade_activation_functions, 0);
|
1383
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_activation_functions", set_cascade_activation_functions, 1);
|
1384
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_activation_steepnesses_count", get_cascade_activation_steepnesses_count, 0);
|
1385
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_activation_steepnesses", get_cascade_activation_steepnesses, 0);
|
1386
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_activation_steepnesses", set_cascade_activation_steepnesses, 1);
|
1387
|
+
rb_define_method(m_rb_fann_standard_class, "get_cascade_num_candidate_groups", get_cascade_num_candidate_groups, 0);
|
1388
|
+
rb_define_method(m_rb_fann_standard_class, "set_cascade_num_candidate_groups", set_cascade_num_candidate_groups, 1);
|
1389
|
+
rb_define_method(m_rb_fann_standard_class, "save", nn_save, 1);
|
1390
|
+
|
1391
|
+
|
1392
|
+
// Uncomment for fixed-point mode (also recompile fann). Probably not going to be needed:
|
1393
|
+
//rb_define_method(clazz, "get_decimal_point", get_decimal_point, 0);
|
1394
|
+
//rb_define_method(clazz, "get_multiplier", get_multiplier, 0);
|
1395
|
+
|
1396
|
+
// Shortcut NN class (duplicated from above so that rdoc generation tools can find the methods:):
|
1397
|
+
m_rb_fann_shortcut_class = rb_define_class_under (m_rb_fann_module, "Shortcut", rb_cObject);
|
1398
|
+
rb_define_alloc_func (m_rb_fann_shortcut_class, fann_allocate);
|
1399
|
+
rb_define_method(m_rb_fann_shortcut_class, "initialize", fann_initialize, 1);
|
1400
|
+
rb_define_method(m_rb_fann_shortcut_class, "init_weights", init_weights, 1);
|
1401
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_function", set_activation_function, 3);
|
1402
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_function_hidden", set_activation_function_hidden, 1);
|
1403
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_function_layer", set_activation_function_layer, 1);
|
1404
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_function_output", set_activation_function_output, 1);
|
1405
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_activation_steepness", get_activation_steepness, 2);
|
1406
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_steepness", set_activation_steepness, 3);
|
1407
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_steepness_hidden", set_activation_steepness_hidden, 1);
|
1408
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_steepness_layer", set_activation_steepness_layer, 2);
|
1409
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_activation_steepness_output", set_activation_steepness_output, 1);
|
1410
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_train_error_function", get_train_error_function, 0);
|
1411
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_train_error_function", set_train_error_function, 1);
|
1412
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_train_stop_function", get_train_stop_function, 0);
|
1413
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_train_stop_function", set_train_stop_function, 1);
|
1414
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_bit_fail_limit", get_bit_fail_limit, 0);
|
1415
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_bit_fail_limit", set_bit_fail_limit, 1);
|
1416
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_quickprop_decay", get_quickprop_decay, 0);
|
1417
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_quickprop_decay", set_quickprop_decay, 1);
|
1418
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_quickprop_mu", get_quickprop_mu, 0);
|
1419
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_quickprop_mu", set_quickprop_mu, 1);
|
1420
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_rprop_increase_factor", get_rprop_increase_factor, 0);
|
1421
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_rprop_increase_factor", set_rprop_increase_factor, 1);
|
1422
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_rprop_decrease_factor", get_rprop_decrease_factor, 0);
|
1423
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_rprop_decrease_factor", set_rprop_decrease_factor, 1);
|
1424
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_rprop_delta_max", get_rprop_delta_max, 0);
|
1425
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_rprop_delta_max", set_rprop_delta_max, 1);
|
1426
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_rprop_delta_min", get_rprop_delta_min, 0);
|
1427
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_rprop_delta_min", set_rprop_delta_min, 1);
|
1428
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_rprop_delta_zero", get_rprop_delta_zero, 0);
|
1429
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_rprop_delta_zero", set_rprop_delta_zero, 1);
|
1430
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_bias_array", get_bias_array, 0);
|
1431
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_connection_rate", get_connection_rate, 0);
|
1432
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_layer_array", get_layer_array, 0);
|
1433
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_network_type", get_network_type, 0);
|
1434
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_neurons", get_neurons, 0);
|
1435
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_num_input", get_num_input, 0);
|
1436
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_num_layers", get_num_layers, 0);
|
1437
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_num_output", get_num_output, 0);
|
1438
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_total_connections", get_total_connections, 0);
|
1439
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_total_neurons", get_total_neurons, 0);
|
1440
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_train_error_function", get_train_error_function, 0);
|
1441
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_train_error_function", set_train_error_function, 1);
|
1442
|
+
rb_define_method(m_rb_fann_shortcut_class, "print_connections", print_connections, 0);
|
1443
|
+
rb_define_method(m_rb_fann_shortcut_class, "print_parameters", print_parameters, 0);
|
1444
|
+
rb_define_method(m_rb_fann_shortcut_class, "randomize_weights", randomize_weights, 2);
|
1445
|
+
rb_define_method(m_rb_fann_shortcut_class, "run", run, 1);
|
1446
|
+
rb_define_method(m_rb_fann_shortcut_class, "train_on_data", train_on_data, 4);
|
1447
|
+
rb_define_method(m_rb_fann_shortcut_class, "train_epoch", train_epoch, 1);
|
1448
|
+
rb_define_method(m_rb_fann_shortcut_class, "test_data", test_data, 1);
|
1449
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_MSE", get_MSE, 0);
|
1450
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_bit_fail", get_bit_fail, 0);
|
1451
|
+
rb_define_method(m_rb_fann_shortcut_class, "reset_MSE", reset_MSE, 0);
|
1452
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_learning_rate", get_learning_rate, 0);
|
1453
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_learning_rate", set_learning_rate, 1);
|
1454
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_learning_momentum", get_learning_momentum, 0);
|
1455
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_learning_momentum", set_learning_momentum, 1);
|
1456
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_training_algorithm", get_training_algorithm, 0);
|
1457
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_training_algorithm", set_training_algorithm, 1);
|
1458
|
+
|
1459
|
+
// Cascade functions:
|
1460
|
+
rb_define_method(m_rb_fann_shortcut_class, "cascadetrain_on_data", cascadetrain_on_data, 4);
|
1461
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_output_change_fraction", get_cascade_output_change_fraction, 0);
|
1462
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_output_change_fraction", set_cascade_output_change_fraction, 1);
|
1463
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_output_stagnation_epochs", get_cascade_output_stagnation_epochs, 0);
|
1464
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_output_stagnation_epochs", set_cascade_output_stagnation_epochs, 1);
|
1465
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_candidate_change_fraction", get_cascade_candidate_change_fraction, 0);
|
1466
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_candidate_change_fraction", set_cascade_candidate_change_fraction, 1);
|
1467
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_candidate_stagnation_epochs", get_cascade_candidate_stagnation_epochs, 0);
|
1468
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_candidate_stagnation_epochs", set_cascade_candidate_stagnation_epochs, 1);
|
1469
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_weight_multiplier", get_cascade_weight_multiplier, 0);
|
1470
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_weight_multiplier", set_cascade_weight_multiplier, 1);
|
1471
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_candidate_limit", get_cascade_candidate_limit, 0);
|
1472
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_candidate_limit", set_cascade_candidate_limit, 1);
|
1473
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_max_out_epochs", get_cascade_max_out_epochs, 0);
|
1474
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_max_out_epochs", set_cascade_max_out_epochs, 1);
|
1475
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_max_cand_epochs", get_cascade_max_cand_epochs, 0);
|
1476
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_max_cand_epochs", set_cascade_max_cand_epochs, 1);
|
1477
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_num_candidates", get_cascade_num_candidates, 0);
|
1478
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_activation_functions_count", get_cascade_activation_functions_count, 0);
|
1479
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_activation_functions", get_cascade_activation_functions, 0);
|
1480
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_activation_functions", set_cascade_activation_functions, 1);
|
1481
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_activation_steepnesses_count", get_cascade_activation_steepnesses_count, 0);
|
1482
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_activation_steepnesses", get_cascade_activation_steepnesses, 0);
|
1483
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_activation_steepnesses", set_cascade_activation_steepnesses, 1);
|
1484
|
+
rb_define_method(m_rb_fann_shortcut_class, "get_cascade_num_candidate_groups", get_cascade_num_candidate_groups, 0);
|
1485
|
+
rb_define_method(m_rb_fann_shortcut_class, "set_cascade_num_candidate_groups", set_cascade_num_candidate_groups, 1);
|
1486
|
+
rb_define_method(m_rb_fann_shortcut_class, "save", nn_save, 1);
|
1487
|
+
|
1488
|
+
|
1489
|
+
// TrainData NN class:
|
1490
|
+
m_rb_fann_train_data_class = rb_define_class_under (m_rb_fann_module, "TrainData", rb_cObject);
|
1491
|
+
rb_define_alloc_func (m_rb_fann_train_data_class, fann_training_data_allocate);
|
1492
|
+
rb_define_method(m_rb_fann_train_data_class, "initialize", fann_train_data_initialize, 1);
|
1493
|
+
rb_define_method(m_rb_fann_train_data_class, "length", length_train_data, 0);
|
1494
|
+
rb_define_method(m_rb_fann_train_data_class, "shuffle", shuffle, 0);
|
1495
|
+
rb_define_method(m_rb_fann_train_data_class, "save", training_save, 1);
|
1496
|
+
|
1497
|
+
printf("Initialized Ruby Bindings for FANN.\n");
|
1498
|
+
}
|
1499
|
+
|