ml 0.3.0 → 0.4.0
Sign up to get free protection for your applications and to get access to all the features.
- data/README.md +3 -0
- data/VERSION +1 -1
- data/lib/data/generator.rb +7 -3
- data/lib/method/adaptive_perceptron.rb +3 -3
- data/lib/method/cyclic_descent.rb +67 -0
- data/lib/method/decision_stump.rb +31 -9
- data/lib/method/linear_regression.rb +34 -0
- data/lib/method/logistic_regression.rb +50 -0
- data/lib/method/perceptron.rb +6 -42
- data/lib/method/pocket.rb +2 -2
- data/lib/method/toolbox.rb +73 -0
- data/lib/ml.rb +7 -0
- data/lib/util/util.rb +20 -0
- data/ml.gemspec +7 -2
- data/spec/learner_spec.rb +19 -98
- metadata +22 -17
data/README.md
CHANGED
@@ -8,6 +8,9 @@ Machine learning library for ruby
|
|
8
8
|
* Adaptive Perceptron (Adaline) Learning Algorithm
|
9
9
|
* Pocket Learning Algorithm
|
10
10
|
* Decision Stump Learning Algorithm
|
11
|
+
* Linear Regression Algorithm
|
12
|
+
* Logistic Regression Algorithm
|
13
|
+
* Cyclic Coordinate Descent
|
11
14
|
|
12
15
|
# Tools
|
13
16
|
|
data/VERSION
CHANGED
@@ -1 +1 @@
|
|
1
|
-
0.
|
1
|
+
0.4.0
|
data/lib/data/generator.rb
CHANGED
@@ -9,8 +9,8 @@ module ML
|
|
9
9
|
# @param [Integer] dim dimension
|
10
10
|
# @param [Numeric] scale the magnitude of the vector
|
11
11
|
# @param [Numeric] noise the percentage of noise
|
12
|
-
# @param [Symbol] model the noise model,
|
13
|
-
# all the element in a probability, while
|
12
|
+
# @param [Symbol] model the noise model, :random for flipping
|
13
|
+
# all the element in a probability, while :flip only flips a
|
14
14
|
# portion of elements randomly
|
15
15
|
def initialize dim, scale = 1, noise = 0, model = :random
|
16
16
|
@dim = dim
|
@@ -85,10 +85,14 @@ module ML
|
|
85
85
|
# @param [Integer] x_range x range
|
86
86
|
# @param [Integer] y_range y range
|
87
87
|
# @param [Numeric] noise the percentage of noise
|
88
|
-
|
88
|
+
# @param [Symbol] model the noise model, :random for flipping
|
89
|
+
# all the element in a probability, while :flip only flips a
|
90
|
+
# portion of elements randomly
|
91
|
+
def initialize x_range = 100, y_range = 100, noise = 0, model = :random
|
89
92
|
@x_range = x_range
|
90
93
|
@y_range = y_range
|
91
94
|
@noise = noise
|
95
|
+
@model = model
|
92
96
|
end
|
93
97
|
|
94
98
|
# Generate two groups of points on 2d plain
|
@@ -8,18 +8,18 @@ module ML
|
|
8
8
|
#
|
9
9
|
# @param [Integer] dim the number of dimension
|
10
10
|
# @param [Float] the eta parameter
|
11
|
-
def initialize dim, eta
|
11
|
+
def initialize dim, eta = 0.1
|
12
12
|
super(dim)
|
13
13
|
@eta = eta
|
14
14
|
end
|
15
15
|
|
16
16
|
protected
|
17
17
|
def wrongly_classify x, y
|
18
|
-
|
18
|
+
classify_inner(x) * y <= 1
|
19
19
|
end
|
20
20
|
|
21
21
|
def update_vector x, y
|
22
|
-
|
22
|
+
self.current_vector += @eta * (y - classify_inner(x))* x
|
23
23
|
end
|
24
24
|
end
|
25
25
|
end
|
@@ -0,0 +1,67 @@
|
|
1
|
+
require 'matrix'
|
2
|
+
|
3
|
+
module ML
|
4
|
+
module Learner
|
5
|
+
# Implementation of cyclic coordinate descent learner
|
6
|
+
class CyclicDescentLearner
|
7
|
+
include Toolbox
|
8
|
+
include LinearToolbox
|
9
|
+
|
10
|
+
# Initialize a learner
|
11
|
+
#
|
12
|
+
# @param [Integer] dim dimension
|
13
|
+
def initialize dim, model = :basis
|
14
|
+
@dim = dim
|
15
|
+
@model = model
|
16
|
+
end
|
17
|
+
|
18
|
+
# Train with a supervised data
|
19
|
+
#
|
20
|
+
# @param [Hash] data supervised input data (mapping from array to integer)
|
21
|
+
# @param [Integer] iteration the desired iteration number
|
22
|
+
def train! data, iteration = 1000
|
23
|
+
self.current_vector = Matrix.column_vector(Array.new(@dim + 1, 0))
|
24
|
+
iteration.times do |i|
|
25
|
+
v = calc_v i
|
26
|
+
eta = calc_eta data, v
|
27
|
+
self.current_vector += eta * v
|
28
|
+
end
|
29
|
+
end
|
30
|
+
|
31
|
+
private
|
32
|
+
|
33
|
+
def calc_eta data, v
|
34
|
+
v_t = v.transpose
|
35
|
+
w_t = self.current_vector.transpose
|
36
|
+
train = {}
|
37
|
+
|
38
|
+
for xn, yn in data
|
39
|
+
x_n = Matrix.column_vector(xn)
|
40
|
+
dot = (v_t * x_n)[0,0] * yn
|
41
|
+
thr = (w_t * x_n)[0,0] * (-yn) / dot
|
42
|
+
|
43
|
+
next if dot == 0
|
44
|
+
if dot > 0
|
45
|
+
train[[thr]] = 1
|
46
|
+
else
|
47
|
+
train[[thr]] = -1
|
48
|
+
end
|
49
|
+
end
|
50
|
+
|
51
|
+
learner = DecisionStumpLearner.new(1)
|
52
|
+
learner.train! train
|
53
|
+
learner.hypothesis[2]
|
54
|
+
end
|
55
|
+
|
56
|
+
def calc_v iteration
|
57
|
+
v = Array.new(@dim + 1, 0)
|
58
|
+
if @model == :basis
|
59
|
+
v[iteration % @dim] = 1
|
60
|
+
else
|
61
|
+
v[iteration % @dim] = Util.normal_distribution 0,1
|
62
|
+
end
|
63
|
+
Matrix.column_vector(v)
|
64
|
+
end
|
65
|
+
end
|
66
|
+
end
|
67
|
+
end
|
@@ -2,6 +2,7 @@ module ML
|
|
2
2
|
module Learner
|
3
3
|
# Implementation of decision stump learning
|
4
4
|
class DecisionStumpLearner
|
5
|
+
include Toolbox
|
5
6
|
# Initialize a decision stump learner
|
6
7
|
#
|
7
8
|
# @param [Integer] dim dimension
|
@@ -14,15 +15,12 @@ module ML
|
|
14
15
|
# Train with a supervised data
|
15
16
|
#
|
16
17
|
# @param [Hash] data supervised input data (mapping from array to integer)
|
17
|
-
# @return [Hash] {error} error of the training data
|
18
18
|
def train! data
|
19
19
|
for i in 0...@dim
|
20
20
|
hypo, error = search data, i
|
21
21
|
update_hypo hypo, error
|
22
22
|
@error_vector[i] = error
|
23
23
|
end
|
24
|
-
|
25
|
-
{:error => @min_error}
|
26
24
|
end
|
27
25
|
|
28
26
|
# Predict certain data
|
@@ -67,8 +65,9 @@ module ML
|
|
67
65
|
def search data, dim
|
68
66
|
pool = data.to_a.sort_by {|line| line[0][dim]}
|
69
67
|
max_diff, index = 0, nil
|
70
|
-
pcount, ncount = 0, 0
|
71
68
|
|
69
|
+
# in order search
|
70
|
+
pcount, ncount = 0, 0
|
72
71
|
pool.each_with_index do |dat, i|
|
73
72
|
if dat[1] == 1
|
74
73
|
pcount += 1
|
@@ -82,10 +81,33 @@ module ML
|
|
82
81
|
end
|
83
82
|
end
|
84
83
|
|
85
|
-
|
86
|
-
|
84
|
+
# reverse search
|
85
|
+
pcount, ncount = 0, 0
|
86
|
+
pool.reverse.each_with_index do |dat, i|
|
87
|
+
if dat[1] == 1
|
88
|
+
pcount += 1
|
89
|
+
else
|
90
|
+
ncount += 1
|
91
|
+
end
|
92
|
+
|
93
|
+
if (ncount - pcount).abs > max_diff.abs
|
94
|
+
max_diff = ncount - pcount
|
95
|
+
index = pool.size - i - 1
|
96
|
+
end
|
97
|
+
end
|
98
|
+
|
99
|
+
thres = if max_diff > 0
|
100
|
+
if index == pool.size - 1
|
101
|
+
pool[-1][0][dim] + 0.01
|
102
|
+
else
|
103
|
+
(pool[index][0][dim] + pool[index+1][0][dim]) / 2.0
|
104
|
+
end
|
87
105
|
else
|
88
|
-
|
106
|
+
if index == 0
|
107
|
+
pool[0][0][dim] - 0.01
|
108
|
+
else
|
109
|
+
(pool[index][0][dim] + pool[index-1][0][dim]) / 2.0
|
110
|
+
end
|
89
111
|
end
|
90
112
|
hypo = if max_diff > 0
|
91
113
|
[-1, dim, thres]
|
@@ -93,10 +115,10 @@ module ML
|
|
93
115
|
[1, dim, thres]
|
94
116
|
end
|
95
117
|
|
96
|
-
[hypo,
|
118
|
+
[hypo, hypo_error(pool, hypo)]
|
97
119
|
end
|
98
120
|
|
99
|
-
def
|
121
|
+
def hypo_error data, hypo
|
100
122
|
error = 0
|
101
123
|
for dat, result in data
|
102
124
|
error += 1 unless classify(dat, hypo) == result
|
@@ -0,0 +1,34 @@
|
|
1
|
+
require 'matrix'
|
2
|
+
|
3
|
+
module ML
|
4
|
+
module Learner
|
5
|
+
# Implementation of linear regression
|
6
|
+
class LinearRegressionLearner
|
7
|
+
include Toolbox
|
8
|
+
include LinearToolbox
|
9
|
+
|
10
|
+
# Intialize linear regression
|
11
|
+
#
|
12
|
+
# @param [Integer] dim the input dimension
|
13
|
+
def initialize dim
|
14
|
+
@dim = dim
|
15
|
+
end
|
16
|
+
|
17
|
+
# Train with supervised data
|
18
|
+
#
|
19
|
+
# @param [Hash] data supervised input data (mapping from array to integer)
|
20
|
+
def train! data
|
21
|
+
x = Matrix.rows(data.keys)
|
22
|
+
ary_y = []
|
23
|
+
for k in data.keys
|
24
|
+
ary_y << data[k]
|
25
|
+
end
|
26
|
+
y = Matrix.column_vector(ary_y)
|
27
|
+
|
28
|
+
x_t = x.transpose
|
29
|
+
x_dag = (x_t * x).inverse * x_t
|
30
|
+
self.current_vector = x_dag * y
|
31
|
+
end
|
32
|
+
end
|
33
|
+
end
|
34
|
+
end
|
@@ -0,0 +1,50 @@
|
|
1
|
+
module ML
|
2
|
+
module Learner
|
3
|
+
# Implementing logistic regression
|
4
|
+
class LogisticRegressionLearner
|
5
|
+
include Toolbox
|
6
|
+
include LinearToolbox
|
7
|
+
|
8
|
+
# Intialize logistic regression
|
9
|
+
#
|
10
|
+
# @param [Integer] dim the input dimension
|
11
|
+
# @param [Numeric] eta the eta parameter
|
12
|
+
# @param [Symbol] model the learning model, :variate for variating
|
13
|
+
# learning rate and :fixed for fixed learning rate
|
14
|
+
def initialize dim, eta = 0.01, model = :variate
|
15
|
+
@dim = dim
|
16
|
+
@eta = eta
|
17
|
+
@model = model
|
18
|
+
end
|
19
|
+
|
20
|
+
# Train with supervised data
|
21
|
+
#
|
22
|
+
# @param [Hash] data supervised input data (mapping from array to integer)
|
23
|
+
# @param [Integer] iteration the number of the iterations
|
24
|
+
def train! data, iteration = 1000
|
25
|
+
self.current_vector = Matrix.column_vector(Array.new(@dim + 1, 0))
|
26
|
+
|
27
|
+
iteration.times do
|
28
|
+
if @model == :variate
|
29
|
+
n = (rand * data.size).to_i
|
30
|
+
key = data.keys[n]
|
31
|
+
self.current_vector -= gradiant(key, data[key]).map {|e| e * @eta }
|
32
|
+
else
|
33
|
+
sum = Matrix.column_vector(Array.new(@dim + 1, 0))
|
34
|
+
for key, value in data
|
35
|
+
sum += gradiant key, value
|
36
|
+
end
|
37
|
+
self.current_vector -= sum.map {|e| e * @eta / data.size }
|
38
|
+
end
|
39
|
+
end
|
40
|
+
end
|
41
|
+
|
42
|
+
protected
|
43
|
+
def gradiant x, y
|
44
|
+
exp = Math.exp(-y * 2 * (self.current_vector.transpose * Matrix.column_vector(x))[0,0])
|
45
|
+
coef = exp * (-2 * y) / (1 + exp)
|
46
|
+
Matrix.column_vector(x).map {|e| e * coef}
|
47
|
+
end
|
48
|
+
end
|
49
|
+
end
|
50
|
+
end
|
data/lib/method/perceptron.rb
CHANGED
@@ -4,12 +4,15 @@ module ML
|
|
4
4
|
module Learner
|
5
5
|
# Implementation of Perceptron Learning Algorithm
|
6
6
|
class PerceptronLearner
|
7
|
+
include Toolbox
|
8
|
+
include LinearToolbox
|
9
|
+
|
7
10
|
# Initialize a perceptron learner
|
8
11
|
#
|
9
12
|
# @param [Integer] dim the number of dimension
|
10
13
|
def initialize dim
|
11
14
|
@dim = dim
|
12
|
-
|
15
|
+
self.current_vector = Matrix.column_vector(Array.new(dim + 1, 0))
|
13
16
|
end
|
14
17
|
|
15
18
|
# Train with supervised data
|
@@ -41,54 +44,15 @@ module ML
|
|
41
44
|
|
42
45
|
break unless misclassified
|
43
46
|
end
|
44
|
-
|
45
|
-
# check out errors
|
46
|
-
error = if update >= threshold
|
47
|
-
classify_error pool
|
48
|
-
else
|
49
|
-
0
|
50
|
-
end
|
51
|
-
|
52
|
-
{:error => error, :update_count => update}
|
53
|
-
end
|
54
|
-
|
55
|
-
# The final coefficient of the line
|
56
|
-
#
|
57
|
-
# @return [Array] [a,b,c] for ax+by+c=0
|
58
|
-
def line
|
59
|
-
@w.column(0).to_a
|
60
|
-
end
|
61
|
-
|
62
|
-
# Predict certain data
|
63
|
-
#
|
64
|
-
# @param [Array] data data in question
|
65
|
-
# @return [Integer] prediction
|
66
|
-
def predict data
|
67
|
-
classify(Matrix.column_vector(data + [1.0])) <=> 0
|
68
47
|
end
|
69
48
|
|
70
49
|
protected
|
71
|
-
def classify data
|
72
|
-
(@w.transpose * data)[0,0]
|
73
|
-
end
|
74
|
-
|
75
50
|
def wrongly_classify x, y
|
76
|
-
|
51
|
+
classify_inner(x) * y <= 0
|
77
52
|
end
|
78
53
|
|
79
54
|
def update_vector x, y
|
80
|
-
|
81
|
-
end
|
82
|
-
|
83
|
-
def classify_error supervised_data
|
84
|
-
error = 0
|
85
|
-
|
86
|
-
for data, result in supervised_data
|
87
|
-
classified_result = (classify(Matrix.column_vector(data)) <=> 0)
|
88
|
-
error += 1 unless result == classified_result
|
89
|
-
end
|
90
|
-
|
91
|
-
error
|
55
|
+
self.current_vector += y * x
|
92
56
|
end
|
93
57
|
end
|
94
58
|
end
|
data/lib/method/pocket.rb
CHANGED
@@ -6,7 +6,7 @@ module ML
|
|
6
6
|
#
|
7
7
|
# @param [Hash] data supervised input data (mapping from array to integer)
|
8
8
|
# @param [Integer] iteration the number of the iterations
|
9
|
-
def train! data, iteration
|
9
|
+
def train! data, iteration = 1000
|
10
10
|
pool = data.to_a
|
11
11
|
best_error, pocket = 1.0/0, nil
|
12
12
|
|
@@ -15,7 +15,7 @@ module ML
|
|
15
15
|
error = classify_error pool
|
16
16
|
if error < best_error
|
17
17
|
error = best_error
|
18
|
-
pocket =
|
18
|
+
pocket = current_vector.dup
|
19
19
|
end
|
20
20
|
break if best_error == 0
|
21
21
|
|
@@ -0,0 +1,73 @@
|
|
1
|
+
require 'matrix'
|
2
|
+
|
3
|
+
module ML
|
4
|
+
module Learner
|
5
|
+
|
6
|
+
# General toolbox for learning methods
|
7
|
+
module Toolbox
|
8
|
+
# Predict a single data with current prediction
|
9
|
+
#
|
10
|
+
# @param [Array] data input vector array
|
11
|
+
# @return [Integer] classified data
|
12
|
+
def predict data
|
13
|
+
raise "Cannot predict"
|
14
|
+
end
|
15
|
+
|
16
|
+
# Calculating model error
|
17
|
+
#
|
18
|
+
# @param [Hash] data
|
19
|
+
# supervised input data (mapping from array to integer)
|
20
|
+
def classify_error supervised_data
|
21
|
+
error = 0
|
22
|
+
|
23
|
+
for data, result in supervised_data
|
24
|
+
classified_result = predict(data)
|
25
|
+
error += 1 unless result == classified_result
|
26
|
+
end
|
27
|
+
|
28
|
+
error.to_f / supervised_data.size
|
29
|
+
end
|
30
|
+
end
|
31
|
+
|
32
|
+
# Learner toolbox for linear model
|
33
|
+
#
|
34
|
+
# The prediction is a vector of dimension d+1 where d is the
|
35
|
+
# dimension of the input data
|
36
|
+
module LinearToolbox
|
37
|
+
# Current prediction vector
|
38
|
+
attr_accessor :current_vector
|
39
|
+
|
40
|
+
# Predict certain data
|
41
|
+
#
|
42
|
+
# @param [Array] data data in question
|
43
|
+
# @return [Integer] prediction
|
44
|
+
def predict data
|
45
|
+
classify_bool Matrix.column_vector(data)
|
46
|
+
end
|
47
|
+
|
48
|
+
# The final coefficient of the line
|
49
|
+
#
|
50
|
+
# @return [Array] [a,b,c] for ax+by+c=0 for 2-d case
|
51
|
+
def line
|
52
|
+
current_vector.column(0).to_a
|
53
|
+
end
|
54
|
+
|
55
|
+
protected
|
56
|
+
# Classify with single data with 0/1 error
|
57
|
+
#
|
58
|
+
# @param [Matrix] data input column vector
|
59
|
+
# @return [Integer] classified data
|
60
|
+
def classify_bool data
|
61
|
+
classify_inner(data) <=> 0
|
62
|
+
end
|
63
|
+
|
64
|
+
# Classify with single data with inner product
|
65
|
+
#
|
66
|
+
# @param [Matrix] data input column vector
|
67
|
+
# @return [Integer] classified data
|
68
|
+
def classify_inner data
|
69
|
+
(current_vector.transpose * data)[0,0]
|
70
|
+
end
|
71
|
+
end
|
72
|
+
end
|
73
|
+
end
|
data/lib/ml.rb
CHANGED
@@ -1,14 +1,21 @@
|
|
1
1
|
require 'rubygems'
|
2
2
|
require 'bundler/setup'
|
3
3
|
|
4
|
+
require 'util/util'
|
5
|
+
|
4
6
|
require 'data/plotter'
|
5
7
|
require 'data/generator'
|
6
8
|
require 'data/parser'
|
7
9
|
|
10
|
+
require 'method/toolbox'
|
11
|
+
|
8
12
|
require 'method/perceptron'
|
9
13
|
require 'method/adaptive_perceptron'
|
10
14
|
require 'method/pocket'
|
11
15
|
require 'method/decision_stump'
|
16
|
+
require 'method/linear_regression'
|
17
|
+
require 'method/logistic_regression'
|
18
|
+
require 'method/cyclic_descent'
|
12
19
|
|
13
20
|
# Top namespace for machine learning algorithms
|
14
21
|
module ML
|
data/lib/util/util.rb
ADDED
@@ -0,0 +1,20 @@
|
|
1
|
+
module ML
|
2
|
+
module Util
|
3
|
+
module_function
|
4
|
+
|
5
|
+
def normal_distribution mean, variance
|
6
|
+
if @y
|
7
|
+
y = @y
|
8
|
+
@y = nil
|
9
|
+
return y
|
10
|
+
end
|
11
|
+
|
12
|
+
theta = 2 * Math::PI * rand
|
13
|
+
rho = Math.sqrt(-2 * Math.log(1 - rand))
|
14
|
+
scale = variance * rho
|
15
|
+
x = mean + scale * Math.cos(theta)
|
16
|
+
@y = mean + scale * Math.sin(theta)
|
17
|
+
x
|
18
|
+
end
|
19
|
+
end
|
20
|
+
end
|
data/ml.gemspec
CHANGED
@@ -5,11 +5,11 @@
|
|
5
5
|
|
6
6
|
Gem::Specification.new do |s|
|
7
7
|
s.name = "ml"
|
8
|
-
s.version = "0.
|
8
|
+
s.version = "0.4.0"
|
9
9
|
|
10
10
|
s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
|
11
11
|
s.authors = ["Andrew Liu"]
|
12
|
-
s.date = "2011-
|
12
|
+
s.date = "2011-12-13"
|
13
13
|
s.description = "Machine learning library in Ruby"
|
14
14
|
s.email = "andrewliu33@gmail.com"
|
15
15
|
s.extra_rdoc_files = [
|
@@ -27,10 +27,15 @@ Gem::Specification.new do |s|
|
|
27
27
|
"lib/data/parser.rb",
|
28
28
|
"lib/data/plotter.rb",
|
29
29
|
"lib/method/adaptive_perceptron.rb",
|
30
|
+
"lib/method/cyclic_descent.rb",
|
30
31
|
"lib/method/decision_stump.rb",
|
32
|
+
"lib/method/linear_regression.rb",
|
33
|
+
"lib/method/logistic_regression.rb",
|
31
34
|
"lib/method/perceptron.rb",
|
32
35
|
"lib/method/pocket.rb",
|
36
|
+
"lib/method/toolbox.rb",
|
33
37
|
"lib/ml.rb",
|
38
|
+
"lib/util/util.rb",
|
34
39
|
"ml.gemspec",
|
35
40
|
"spec/data_spec.rb",
|
36
41
|
"spec/learner_spec.rb",
|
data/spec/learner_spec.rb
CHANGED
@@ -1,111 +1,32 @@
|
|
1
1
|
require 'spec_helper'
|
2
2
|
|
3
3
|
describe "Learner" do
|
4
|
-
|
5
|
-
|
6
|
-
|
4
|
+
[ML::Learner::PerceptronLearner, ML::Learner::DecisionStumpLearner,
|
5
|
+
ML::Learner::AdaptivePerceptronLearner, ML::Learner::PocketLearner,
|
6
|
+
ML::Learner::LinearRegressionLearner, ML::Learner::CyclicDescentLearner,
|
7
|
+
ML::Learner::LogisticRegressionLearner].each do |method|
|
8
|
+
describe method.to_s do
|
9
|
+
it "should run #{method.to_s} in 2d" do
|
10
|
+
learner = method.new(2)
|
7
11
|
|
8
|
-
|
9
|
-
|
12
|
+
generator = ML::Data::Generator2D.new
|
13
|
+
data = generator.points_2d(100)
|
10
14
|
|
11
|
-
|
12
|
-
error = response[:error]
|
13
|
-
update_count = response[:update_count]
|
15
|
+
learner.train! data
|
14
16
|
|
15
|
-
|
16
|
-
|
17
|
-
line.size.should == 3
|
17
|
+
learner.classify_error(data).should < 0.5
|
18
|
+
end
|
18
19
|
|
19
|
-
|
20
|
-
|
21
|
-
|
22
|
-
it "should run perceptron learning in hyperspace" do
|
23
|
-
learner = ML::Learner::PerceptronLearner.new(4)
|
24
|
-
|
25
|
-
generator = ML::Data::Generator.new(4)
|
26
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
27
|
-
|
28
|
-
response = learner.train! data
|
29
|
-
error = response[:error]
|
30
|
-
update_count = response[:update_count]
|
31
|
-
|
32
|
-
line = learner.line
|
33
|
-
line.should.kind_of?(Array).should == true
|
34
|
-
line.size.should == 5
|
35
|
-
|
36
|
-
update_count.should > 0
|
37
|
-
end
|
38
|
-
end
|
39
|
-
|
40
|
-
describe "Adpative Perceptron Learner" do
|
41
|
-
it "should run adaptive perceptron learning in hyperspace" do
|
42
|
-
learner = ML::Learner::AdaptivePerceptronLearner.new(4, 0.1)
|
43
|
-
|
44
|
-
generator = ML::Data::Generator.new(4, 100)
|
45
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
46
|
-
|
47
|
-
response = learner.train! data, 1000
|
48
|
-
error = response[:error]
|
49
|
-
update_count = response[:update_count]
|
50
|
-
|
51
|
-
line = learner.line
|
52
|
-
line.should.kind_of?(Array).should == true
|
53
|
-
line.size.should == 5
|
54
|
-
|
55
|
-
update_count.should > 0
|
56
|
-
update_count.should < 1000
|
57
|
-
end
|
58
|
-
end
|
59
|
-
|
60
|
-
describe "Pocket Learner" do
|
61
|
-
it "should run pocket perceptron learning in hyperspace" do
|
62
|
-
learner = ML::Learner::PocketLearner.new(4)
|
63
|
-
|
64
|
-
generator = ML::Data::Generator.new(4)
|
65
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
66
|
-
|
67
|
-
learner.train! data, 1000
|
68
|
-
|
69
|
-
line = learner.line
|
70
|
-
line.should.kind_of?(Array).should == true
|
71
|
-
line.size.should == 5
|
72
|
-
end
|
73
|
-
|
74
|
-
it "should run pocket perceptron learning in noisy data" do
|
75
|
-
learner = ML::Learner::PocketLearner.new(4)
|
76
|
-
|
77
|
-
generator = ML::Data::Generator.new(4, 1, 0.1)
|
78
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
79
|
-
|
80
|
-
learner.train! data, 1000
|
81
|
-
|
82
|
-
line = learner.line
|
83
|
-
line.should.kind_of?(Array).should == true
|
84
|
-
line.size.should == 5
|
85
|
-
end
|
86
|
-
end
|
87
|
-
|
88
|
-
describe "Decision Stump Learner" do
|
89
|
-
it "should run decision stump learning in hyperspace" do
|
90
|
-
learner = ML::Learner::DecisionStumpLearner.new(4)
|
91
|
-
|
92
|
-
generator = ML::Data::Generator.new(4)
|
93
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
94
|
-
|
95
|
-
learner.train! data
|
96
|
-
vector = learner.error_vector
|
97
|
-
vector.size.should == 4
|
98
|
-
end
|
20
|
+
it "should run #{method.to_s} in hyperspace" do
|
21
|
+
learner = method.new(4)
|
99
22
|
|
100
|
-
|
101
|
-
|
23
|
+
generator = ML::Data::Generator.new(4)
|
24
|
+
data = generator.points(100, ML::Data::Generator.generate_vector(4))
|
102
25
|
|
103
|
-
|
104
|
-
data = generator.points(10, ML::Data::Generator.generate_vector(4))
|
26
|
+
learner.train! data
|
105
27
|
|
106
|
-
|
107
|
-
|
108
|
-
vector.size.should == 4
|
28
|
+
learner.classify_error(data).should < 0.5
|
29
|
+
end
|
109
30
|
end
|
110
31
|
end
|
111
32
|
end
|
metadata
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
2
2
|
name: ml
|
3
3
|
version: !ruby/object:Gem::Version
|
4
|
-
version: 0.
|
4
|
+
version: 0.4.0
|
5
5
|
prerelease:
|
6
6
|
platform: ruby
|
7
7
|
authors:
|
@@ -9,11 +9,11 @@ authors:
|
|
9
9
|
autorequire:
|
10
10
|
bindir: bin
|
11
11
|
cert_chain: []
|
12
|
-
date: 2011-
|
12
|
+
date: 2011-12-13 00:00:00.000000000Z
|
13
13
|
dependencies:
|
14
14
|
- !ruby/object:Gem::Dependency
|
15
15
|
name: rubyvis
|
16
|
-
requirement: &
|
16
|
+
requirement: &2153078140 !ruby/object:Gem::Requirement
|
17
17
|
none: false
|
18
18
|
requirements:
|
19
19
|
- - ! '>='
|
@@ -21,10 +21,10 @@ dependencies:
|
|
21
21
|
version: '0'
|
22
22
|
type: :runtime
|
23
23
|
prerelease: false
|
24
|
-
version_requirements: *
|
24
|
+
version_requirements: *2153078140
|
25
25
|
- !ruby/object:Gem::Dependency
|
26
26
|
name: nokogiri
|
27
|
-
requirement: &
|
27
|
+
requirement: &2153076840 !ruby/object:Gem::Requirement
|
28
28
|
none: false
|
29
29
|
requirements:
|
30
30
|
- - ! '>='
|
@@ -32,10 +32,10 @@ dependencies:
|
|
32
32
|
version: '0'
|
33
33
|
type: :runtime
|
34
34
|
prerelease: false
|
35
|
-
version_requirements: *
|
35
|
+
version_requirements: *2153076840
|
36
36
|
- !ruby/object:Gem::Dependency
|
37
37
|
name: bacon
|
38
|
-
requirement: &
|
38
|
+
requirement: &2153075100 !ruby/object:Gem::Requirement
|
39
39
|
none: false
|
40
40
|
requirements:
|
41
41
|
- - ! '>='
|
@@ -43,10 +43,10 @@ dependencies:
|
|
43
43
|
version: '0'
|
44
44
|
type: :development
|
45
45
|
prerelease: false
|
46
|
-
version_requirements: *
|
46
|
+
version_requirements: *2153075100
|
47
47
|
- !ruby/object:Gem::Dependency
|
48
48
|
name: yard
|
49
|
-
requirement: &
|
49
|
+
requirement: &2153060800 !ruby/object:Gem::Requirement
|
50
50
|
none: false
|
51
51
|
requirements:
|
52
52
|
- - ~>
|
@@ -54,10 +54,10 @@ dependencies:
|
|
54
54
|
version: 0.6.0
|
55
55
|
type: :development
|
56
56
|
prerelease: false
|
57
|
-
version_requirements: *
|
57
|
+
version_requirements: *2153060800
|
58
58
|
- !ruby/object:Gem::Dependency
|
59
59
|
name: bundler
|
60
|
-
requirement: &
|
60
|
+
requirement: &2153059640 !ruby/object:Gem::Requirement
|
61
61
|
none: false
|
62
62
|
requirements:
|
63
63
|
- - ~>
|
@@ -65,10 +65,10 @@ dependencies:
|
|
65
65
|
version: 1.0.0
|
66
66
|
type: :development
|
67
67
|
prerelease: false
|
68
|
-
version_requirements: *
|
68
|
+
version_requirements: *2153059640
|
69
69
|
- !ruby/object:Gem::Dependency
|
70
70
|
name: jeweler
|
71
|
-
requirement: &
|
71
|
+
requirement: &2153058400 !ruby/object:Gem::Requirement
|
72
72
|
none: false
|
73
73
|
requirements:
|
74
74
|
- - ~>
|
@@ -76,10 +76,10 @@ dependencies:
|
|
76
76
|
version: 1.6.4
|
77
77
|
type: :development
|
78
78
|
prerelease: false
|
79
|
-
version_requirements: *
|
79
|
+
version_requirements: *2153058400
|
80
80
|
- !ruby/object:Gem::Dependency
|
81
81
|
name: rcov
|
82
|
-
requirement: &
|
82
|
+
requirement: &2153056540 !ruby/object:Gem::Requirement
|
83
83
|
none: false
|
84
84
|
requirements:
|
85
85
|
- - ! '>='
|
@@ -87,7 +87,7 @@ dependencies:
|
|
87
87
|
version: '0'
|
88
88
|
type: :development
|
89
89
|
prerelease: false
|
90
|
-
version_requirements: *
|
90
|
+
version_requirements: *2153056540
|
91
91
|
description: Machine learning library in Ruby
|
92
92
|
email: andrewliu33@gmail.com
|
93
93
|
executables: []
|
@@ -106,10 +106,15 @@ files:
|
|
106
106
|
- lib/data/parser.rb
|
107
107
|
- lib/data/plotter.rb
|
108
108
|
- lib/method/adaptive_perceptron.rb
|
109
|
+
- lib/method/cyclic_descent.rb
|
109
110
|
- lib/method/decision_stump.rb
|
111
|
+
- lib/method/linear_regression.rb
|
112
|
+
- lib/method/logistic_regression.rb
|
110
113
|
- lib/method/perceptron.rb
|
111
114
|
- lib/method/pocket.rb
|
115
|
+
- lib/method/toolbox.rb
|
112
116
|
- lib/ml.rb
|
117
|
+
- lib/util/util.rb
|
113
118
|
- ml.gemspec
|
114
119
|
- spec/data_spec.rb
|
115
120
|
- spec/learner_spec.rb
|
@@ -129,7 +134,7 @@ required_ruby_version: !ruby/object:Gem::Requirement
|
|
129
134
|
version: '0'
|
130
135
|
segments:
|
131
136
|
- 0
|
132
|
-
hash:
|
137
|
+
hash: -1343753565689468056
|
133
138
|
required_rubygems_version: !ruby/object:Gem::Requirement
|
134
139
|
none: false
|
135
140
|
requirements:
|