pnmatrix 1.2.4
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/ext/nmatrix/binary_format.txt +53 -0
- data/ext/nmatrix/data/complex.h +388 -0
- data/ext/nmatrix/data/data.cpp +274 -0
- data/ext/nmatrix/data/data.h +651 -0
- data/ext/nmatrix/data/meta.h +64 -0
- data/ext/nmatrix/data/ruby_object.h +386 -0
- data/ext/nmatrix/extconf.rb +70 -0
- data/ext/nmatrix/math/asum.h +99 -0
- data/ext/nmatrix/math/cblas_enums.h +36 -0
- data/ext/nmatrix/math/cblas_templates_core.h +507 -0
- data/ext/nmatrix/math/gemm.h +241 -0
- data/ext/nmatrix/math/gemv.h +178 -0
- data/ext/nmatrix/math/getrf.h +255 -0
- data/ext/nmatrix/math/getrs.h +121 -0
- data/ext/nmatrix/math/imax.h +82 -0
- data/ext/nmatrix/math/laswp.h +165 -0
- data/ext/nmatrix/math/long_dtype.h +62 -0
- data/ext/nmatrix/math/magnitude.h +54 -0
- data/ext/nmatrix/math/math.h +751 -0
- data/ext/nmatrix/math/nrm2.h +165 -0
- data/ext/nmatrix/math/rot.h +117 -0
- data/ext/nmatrix/math/rotg.h +106 -0
- data/ext/nmatrix/math/scal.h +71 -0
- data/ext/nmatrix/math/trsm.h +336 -0
- data/ext/nmatrix/math/util.h +162 -0
- data/ext/nmatrix/math.cpp +1368 -0
- data/ext/nmatrix/nm_memory.h +60 -0
- data/ext/nmatrix/nmatrix.cpp +285 -0
- data/ext/nmatrix/nmatrix.h +476 -0
- data/ext/nmatrix/ruby_constants.cpp +151 -0
- data/ext/nmatrix/ruby_constants.h +106 -0
- data/ext/nmatrix/ruby_nmatrix.c +3130 -0
- data/ext/nmatrix/storage/common.cpp +77 -0
- data/ext/nmatrix/storage/common.h +183 -0
- data/ext/nmatrix/storage/dense/dense.cpp +1096 -0
- data/ext/nmatrix/storage/dense/dense.h +129 -0
- data/ext/nmatrix/storage/list/list.cpp +1628 -0
- data/ext/nmatrix/storage/list/list.h +138 -0
- data/ext/nmatrix/storage/storage.cpp +730 -0
- data/ext/nmatrix/storage/storage.h +99 -0
- data/ext/nmatrix/storage/yale/class.h +1139 -0
- data/ext/nmatrix/storage/yale/iterators/base.h +143 -0
- data/ext/nmatrix/storage/yale/iterators/iterator.h +131 -0
- data/ext/nmatrix/storage/yale/iterators/row.h +450 -0
- data/ext/nmatrix/storage/yale/iterators/row_stored.h +140 -0
- data/ext/nmatrix/storage/yale/iterators/row_stored_nd.h +169 -0
- data/ext/nmatrix/storage/yale/iterators/stored_diagonal.h +124 -0
- data/ext/nmatrix/storage/yale/math/transpose.h +110 -0
- data/ext/nmatrix/storage/yale/yale.cpp +2074 -0
- data/ext/nmatrix/storage/yale/yale.h +203 -0
- data/ext/nmatrix/types.h +55 -0
- data/ext/nmatrix/util/io.cpp +279 -0
- data/ext/nmatrix/util/io.h +115 -0
- data/ext/nmatrix/util/sl_list.cpp +627 -0
- data/ext/nmatrix/util/sl_list.h +144 -0
- data/ext/nmatrix/util/util.h +78 -0
- data/lib/nmatrix/blas.rb +378 -0
- data/lib/nmatrix/cruby/math.rb +744 -0
- data/lib/nmatrix/enumerate.rb +253 -0
- data/lib/nmatrix/homogeneous.rb +241 -0
- data/lib/nmatrix/io/fortran_format.rb +138 -0
- data/lib/nmatrix/io/harwell_boeing.rb +221 -0
- data/lib/nmatrix/io/market.rb +263 -0
- data/lib/nmatrix/io/point_cloud.rb +189 -0
- data/lib/nmatrix/jruby/decomposition.rb +24 -0
- data/lib/nmatrix/jruby/enumerable.rb +13 -0
- data/lib/nmatrix/jruby/error.rb +4 -0
- data/lib/nmatrix/jruby/math.rb +501 -0
- data/lib/nmatrix/jruby/nmatrix_java.rb +840 -0
- data/lib/nmatrix/jruby/operators.rb +283 -0
- data/lib/nmatrix/jruby/slice.rb +264 -0
- data/lib/nmatrix/lapack_core.rb +181 -0
- data/lib/nmatrix/lapack_plugin.rb +44 -0
- data/lib/nmatrix/math.rb +953 -0
- data/lib/nmatrix/mkmf.rb +100 -0
- data/lib/nmatrix/monkeys.rb +137 -0
- data/lib/nmatrix/nmatrix.rb +1172 -0
- data/lib/nmatrix/rspec.rb +75 -0
- data/lib/nmatrix/shortcuts.rb +1163 -0
- data/lib/nmatrix/version.rb +39 -0
- data/lib/nmatrix/yale_functions.rb +118 -0
- data/lib/nmatrix.rb +28 -0
- data/spec/00_nmatrix_spec.rb +892 -0
- data/spec/01_enum_spec.rb +196 -0
- data/spec/02_slice_spec.rb +407 -0
- data/spec/03_nmatrix_monkeys_spec.rb +80 -0
- data/spec/2x2_dense_double.mat +0 -0
- data/spec/4x4_sparse.mat +0 -0
- data/spec/4x5_dense.mat +0 -0
- data/spec/blas_spec.rb +215 -0
- data/spec/elementwise_spec.rb +311 -0
- data/spec/homogeneous_spec.rb +100 -0
- data/spec/io/fortran_format_spec.rb +88 -0
- data/spec/io/harwell_boeing_spec.rb +98 -0
- data/spec/io/test.rua +9 -0
- data/spec/io_spec.rb +159 -0
- data/spec/lapack_core_spec.rb +482 -0
- data/spec/leakcheck.rb +16 -0
- data/spec/math_spec.rb +1363 -0
- data/spec/nmatrix_yale_resize_test_associations.yaml +2802 -0
- data/spec/nmatrix_yale_spec.rb +286 -0
- data/spec/rspec_monkeys.rb +56 -0
- data/spec/rspec_spec.rb +35 -0
- data/spec/shortcuts_spec.rb +474 -0
- data/spec/slice_set_spec.rb +162 -0
- data/spec/spec_helper.rb +172 -0
- data/spec/stat_spec.rb +214 -0
- data/spec/test.pcd +20 -0
- data/spec/utm5940.mtx +83844 -0
- metadata +295 -0
data/lib/nmatrix/math.rb
ADDED
@@ -0,0 +1,953 @@
|
|
1
|
+
#--
|
2
|
+
# = NMatrix
|
3
|
+
#
|
4
|
+
# A linear algebra library for scientific computation in Ruby.
|
5
|
+
# NMatrix is part of SciRuby.
|
6
|
+
#
|
7
|
+
# NMatrix was originally inspired by and derived from NArray, by
|
8
|
+
# Masahiro Tanaka: http://narray.rubyforge.org
|
9
|
+
#
|
10
|
+
# == Copyright Information
|
11
|
+
#
|
12
|
+
# SciRuby is Copyright (c) 2010 - 2014, Ruby Science Foundation
|
13
|
+
# NMatrix is Copyright (c) 2012 - 2014, John Woods and the Ruby Science Foundation
|
14
|
+
#
|
15
|
+
# Please see LICENSE.txt for additional copyright notices.
|
16
|
+
#
|
17
|
+
# == Contributing
|
18
|
+
#
|
19
|
+
# By contributing source code to SciRuby, you agree to be bound by
|
20
|
+
# our Contributor Agreement:
|
21
|
+
#
|
22
|
+
# * https://github.com/SciRuby/sciruby/wiki/Contributor-Agreement
|
23
|
+
#
|
24
|
+
# == math.rb
|
25
|
+
#
|
26
|
+
# Math functionality for NMatrix, along with any NMatrix instance
|
27
|
+
# methods that correspond to ATLAS/BLAS/LAPACK functions (e.g.,
|
28
|
+
# laswp).
|
29
|
+
#++
|
30
|
+
|
31
|
+
class NMatrix
|
32
|
+
|
33
|
+
module NMMath #:nodoc:
|
34
|
+
METHODS_ARITY_2 = [:atan2, :ldexp, :hypot]
|
35
|
+
METHODS_ARITY_1 = [:cos, :sin, :tan, :acos, :asin, :atan, :cosh, :sinh, :tanh, :acosh,
|
36
|
+
:asinh, :atanh, :exp, :log2, :log10, :sqrt, :cbrt, :erf, :erfc, :gamma, :-@]
|
37
|
+
end
|
38
|
+
|
39
|
+
# Methods for generating permutation matrix from LU factorization results.
|
40
|
+
module FactorizeLUMethods
|
41
|
+
class << self
|
42
|
+
def permutation_matrix_from(pivot_array)
|
43
|
+
perm_arry = permutation_array_for(pivot_array)
|
44
|
+
n = NMatrix.zeros(perm_arry.size, dtype: :byte)
|
45
|
+
|
46
|
+
perm_arry.each_with_index { |e, i| n[e,i] = 1 }
|
47
|
+
|
48
|
+
n
|
49
|
+
end
|
50
|
+
|
51
|
+
def permutation_array_for(pivot_array)
|
52
|
+
perm_arry = Array.new(pivot_array.size) { |i| i }
|
53
|
+
perm_arry.each_index do |i|
|
54
|
+
#the pivot indices returned by LAPACK getrf are indexed starting
|
55
|
+
#from 1, so we need to subtract 1 here
|
56
|
+
perm_arry[i], perm_arry[pivot_array[i]-1] = perm_arry[pivot_array[i]-1], perm_arry[i]
|
57
|
+
end
|
58
|
+
|
59
|
+
perm_arry
|
60
|
+
end
|
61
|
+
end
|
62
|
+
end
|
63
|
+
|
64
|
+
#
|
65
|
+
# call-seq:
|
66
|
+
# invert! -> NMatrix
|
67
|
+
#
|
68
|
+
# Use LAPACK to calculate the inverse of the matrix (in-place) if available.
|
69
|
+
# Only works on dense matrices. Alternatively uses in-place Gauss-Jordan
|
70
|
+
# elimination.
|
71
|
+
#
|
72
|
+
# * *Raises* :
|
73
|
+
# - +StorageTypeError+ -> only implemented on dense matrices.
|
74
|
+
# - +ShapeError+ -> matrix must be square.
|
75
|
+
# - +DataTypeError+ -> cannot invert an integer matrix in-place.
|
76
|
+
#
|
77
|
+
def invert!
|
78
|
+
raise(StorageTypeError, "invert only works on dense matrices currently") unless self.dense?
|
79
|
+
raise(ShapeError, "Cannot invert non-square matrix") unless self.dim == 2 && self.shape[0] == self.shape[1]
|
80
|
+
raise(DataTypeError, "Cannot invert an integer matrix in-place") if self.integer_dtype?
|
81
|
+
|
82
|
+
#No internal implementation of getri, so use this other function
|
83
|
+
__inverse__(self, true)
|
84
|
+
end
|
85
|
+
|
86
|
+
#
|
87
|
+
# call-seq:
|
88
|
+
# invert -> NMatrix
|
89
|
+
#
|
90
|
+
# Make a copy of the matrix, then invert using Gauss-Jordan elimination.
|
91
|
+
# Works without LAPACK.
|
92
|
+
#
|
93
|
+
# * *Returns* :
|
94
|
+
# - A dense NMatrix. Will be the same type as the input NMatrix,
|
95
|
+
# except if the input is an integral dtype, in which case it will be a
|
96
|
+
# :float64 NMatrix.
|
97
|
+
#
|
98
|
+
# * *Raises* :
|
99
|
+
# - +StorageTypeError+ -> only implemented on dense matrices.
|
100
|
+
# - +ShapeError+ -> matrix must be square.
|
101
|
+
#
|
102
|
+
def invert
|
103
|
+
#write this in terms of invert! so plugins will only have to overwrite
|
104
|
+
#invert! and not invert
|
105
|
+
if self.integer_dtype?
|
106
|
+
cloned = self.cast(dtype: :float64)
|
107
|
+
cloned.invert!
|
108
|
+
else
|
109
|
+
cloned = self.clone
|
110
|
+
cloned.invert!
|
111
|
+
end
|
112
|
+
end
|
113
|
+
alias :inverse :invert
|
114
|
+
|
115
|
+
# call-seq:
|
116
|
+
# exact_inverse! -> NMatrix
|
117
|
+
#
|
118
|
+
# Calulates inverse_exact of a matrix of size 2 or 3.
|
119
|
+
# Only works on dense matrices.
|
120
|
+
#
|
121
|
+
# * *Raises* :
|
122
|
+
# - +DataTypeError+ -> cannot invert an integer matrix in-place.
|
123
|
+
# - +NotImplementedError+ -> cannot find exact inverse of matrix with size greater than 3 #
|
124
|
+
def exact_inverse!
|
125
|
+
raise(ShapeError, "Cannot invert non-square matrix") unless self.dim == 2 && self.shape[0] == self.shape[1]
|
126
|
+
raise(DataTypeError, "Cannot invert an integer matrix in-place") if self.integer_dtype?
|
127
|
+
#No internal implementation of getri, so use this other function
|
128
|
+
n = self.shape[0]
|
129
|
+
if n>3
|
130
|
+
raise(NotImplementedError, "Cannot find exact inverse of matrix of size greater than 3")
|
131
|
+
else
|
132
|
+
clond=self.clone
|
133
|
+
__inverse_exact__(clond, n, n)
|
134
|
+
end
|
135
|
+
end
|
136
|
+
|
137
|
+
#
|
138
|
+
# call-seq:
|
139
|
+
# exact_inverse -> NMatrix
|
140
|
+
#
|
141
|
+
# Make a copy of the matrix, then invert using exact_inverse
|
142
|
+
#
|
143
|
+
# * *Returns* :
|
144
|
+
# - A dense NMatrix. Will be the same type as the input NMatrix,
|
145
|
+
# except if the input is an integral dtype, in which case it will be a
|
146
|
+
# :float64 NMatrix.
|
147
|
+
#
|
148
|
+
# * *Raises* :
|
149
|
+
# - +StorageTypeError+ -> only implemented on dense matrices.
|
150
|
+
# - +ShapeError+ -> matrix must be square.
|
151
|
+
# - +NotImplementedError+ -> cannot find exact inverse of matrix with size greater than 3
|
152
|
+
#
|
153
|
+
def exact_inverse
|
154
|
+
#write this in terms of exact_inverse! so plugins will only have to overwrite
|
155
|
+
#exact_inverse! and not exact_inverse
|
156
|
+
if self.integer_dtype?
|
157
|
+
cloned = self.cast(dtype: :float64)
|
158
|
+
cloned.exact_inverse!
|
159
|
+
else
|
160
|
+
cloned = self.clone
|
161
|
+
cloned.exact_inverse!
|
162
|
+
end
|
163
|
+
end
|
164
|
+
alias :invert_exactly :exact_inverse
|
165
|
+
|
166
|
+
|
167
|
+
|
168
|
+
#
|
169
|
+
# call-seq:
|
170
|
+
# pinv -> NMatrix
|
171
|
+
#
|
172
|
+
# Compute the Moore-Penrose pseudo-inverse of a matrix using its
|
173
|
+
# singular value decomposition (SVD).
|
174
|
+
#
|
175
|
+
# This function requires the nmatrix-atlas gem installed.
|
176
|
+
#
|
177
|
+
# * *Arguments* :
|
178
|
+
# - +tolerance(optional)+ -> Cutoff for small singular values.
|
179
|
+
#
|
180
|
+
# * *Returns* :
|
181
|
+
# - Pseudo-inverse matrix.
|
182
|
+
#
|
183
|
+
# * *Raises* :
|
184
|
+
# - +NotImplementedError+ -> If called without nmatrix-atlas or nmatrix-lapacke gem.
|
185
|
+
# - +TypeError+ -> If called without float or complex data type.
|
186
|
+
#
|
187
|
+
# * *Examples* :
|
188
|
+
#
|
189
|
+
# a = NMatrix.new([2,2],[1,2,
|
190
|
+
# 3,4], dtype: :float64)
|
191
|
+
# a.pinv # => [ [-2.0000000000000018, 1.0000000000000007]
|
192
|
+
# [1.5000000000000016, -0.5000000000000008] ]
|
193
|
+
#
|
194
|
+
# b = NMatrix.new([4,1],[1,2,3,4], dtype: :float64)
|
195
|
+
# b.pinv # => [ [ 0.03333333, 0.06666667, 0.99999999, 0.13333333] ]
|
196
|
+
#
|
197
|
+
# == References
|
198
|
+
#
|
199
|
+
# * https://en.wikipedia.org/wiki/Moore%E2%80%93Penrose_pseudoinverse
|
200
|
+
# * G. Strang, Linear Algebra and Its Applications, 2nd Ed., Orlando, FL, Academic Press
|
201
|
+
#
|
202
|
+
def pinv(tolerance = 1e-15)
|
203
|
+
raise DataTypeError, "pinv works only with matrices of float or complex data type" unless
|
204
|
+
[:float32, :float64, :complex64, :complex128].include?(dtype)
|
205
|
+
if self.complex_dtype?
|
206
|
+
u, s, vt = self.complex_conjugate.gesvd # singular value decomposition
|
207
|
+
else
|
208
|
+
u, s, vt = self.gesvd
|
209
|
+
end
|
210
|
+
rows = self.shape[0]
|
211
|
+
cols = self.shape[1]
|
212
|
+
if rows < cols
|
213
|
+
u_reduced = u
|
214
|
+
vt_reduced = vt[0..rows - 1, 0..cols - 1].transpose
|
215
|
+
else
|
216
|
+
u_reduced = u[0..rows - 1, 0..cols - 1]
|
217
|
+
vt_reduced = vt.transpose
|
218
|
+
end
|
219
|
+
largest_singular_value = s.max.to_f
|
220
|
+
cutoff = tolerance * largest_singular_value
|
221
|
+
(0...[rows, cols].min).each do |i|
|
222
|
+
s[i] = 1 / s[i] if s[i] > cutoff
|
223
|
+
s[i] = 0 if s[i] <= cutoff
|
224
|
+
end
|
225
|
+
multiplier = u_reduced.dot(NMatrix.diagonal(s.to_a)).transpose
|
226
|
+
vt_reduced.dot(multiplier)
|
227
|
+
end
|
228
|
+
alias :pseudo_inverse :pinv
|
229
|
+
alias :pseudoinverse :pinv
|
230
|
+
|
231
|
+
|
232
|
+
#
|
233
|
+
# call-seq:
|
234
|
+
# adjugate! -> NMatrix
|
235
|
+
#
|
236
|
+
# Calculate the adjugate of the matrix (in-place).
|
237
|
+
# Only works on dense matrices.
|
238
|
+
#
|
239
|
+
# * *Raises* :
|
240
|
+
# - +StorageTypeError+ -> only implemented on dense matrices.
|
241
|
+
# - +ShapeError+ -> matrix must be square.
|
242
|
+
# - +DataTypeError+ -> cannot calculate adjugate of an integer matrix in-place.
|
243
|
+
#
|
244
|
+
def adjugate!
|
245
|
+
raise(StorageTypeError, "adjugate only works on dense matrices currently") unless self.dense?
|
246
|
+
raise(ShapeError, "Cannot calculate adjugate of a non-square matrix") unless self.dim == 2 && self.shape[0] == self.shape[1]
|
247
|
+
raise(DataTypeError, "Cannot calculate adjugate of an integer matrix in-place") if self.integer_dtype?
|
248
|
+
d = self.det
|
249
|
+
self.invert!
|
250
|
+
self.map! { |e| e * d }
|
251
|
+
self
|
252
|
+
end
|
253
|
+
alias :adjoint! :adjugate!
|
254
|
+
|
255
|
+
#
|
256
|
+
# call-seq:
|
257
|
+
# adjugate -> NMatrix
|
258
|
+
#
|
259
|
+
# Make a copy of the matrix and calculate the adjugate of the matrix.
|
260
|
+
# Only works on dense matrices.
|
261
|
+
#
|
262
|
+
# * *Returns* :
|
263
|
+
# - A dense NMatrix. Will be the same type as the input NMatrix,
|
264
|
+
# except if the input is an integral dtype, in which case it will be a
|
265
|
+
# :float64 NMatrix.
|
266
|
+
#
|
267
|
+
# * *Raises* :
|
268
|
+
# - +StorageTypeError+ -> only implemented on dense matrices.
|
269
|
+
# - +ShapeError+ -> matrix must be square.
|
270
|
+
#
|
271
|
+
def adjugate
|
272
|
+
raise(StorageTypeError, "adjugate only works on dense matrices currently") unless self.dense?
|
273
|
+
raise(ShapeError, "Cannot calculate adjugate of a non-square matrix") unless self.dim == 2 && self.shape[0] == self.shape[1]
|
274
|
+
d = self.det
|
275
|
+
mat = self.invert
|
276
|
+
mat.map! { |e| e * d }
|
277
|
+
mat
|
278
|
+
end
|
279
|
+
alias :adjoint :adjugate
|
280
|
+
|
281
|
+
# Reduce self to upper hessenberg form using householder transforms.
|
282
|
+
#
|
283
|
+
# == References
|
284
|
+
#
|
285
|
+
# * http://en.wikipedia.org/wiki/Hessenberg_matrix
|
286
|
+
# * http://www.mymathlib.com/c_source/matrices/eigen/hessenberg_orthog.c
|
287
|
+
def hessenberg
|
288
|
+
clone.hessenberg!
|
289
|
+
end
|
290
|
+
|
291
|
+
# Destructive version of #hessenberg
|
292
|
+
def hessenberg!
|
293
|
+
raise ShapeError, "Trying to reduce non 2D matrix to hessenberg form" if
|
294
|
+
shape.size != 2
|
295
|
+
raise ShapeError, "Trying to reduce non-square matrix to hessenberg form" if
|
296
|
+
shape[0] != shape[1]
|
297
|
+
raise StorageTypeError, "Matrix must be dense" if stype != :dense
|
298
|
+
raise TypeError, "Works with float matrices only" unless
|
299
|
+
[:float64,:float32].include?(dtype)
|
300
|
+
|
301
|
+
__hessenberg__(self)
|
302
|
+
self
|
303
|
+
end
|
304
|
+
|
305
|
+
|
306
|
+
# call-seq:
|
307
|
+
# matrix_norm -> Numeric
|
308
|
+
#
|
309
|
+
# Calculates the selected norm (defaults to 2-norm) of a 2D matrix.
|
310
|
+
#
|
311
|
+
# This should be used for small or medium sized matrices.
|
312
|
+
# For greater matrices, there should be a separate implementation where
|
313
|
+
# the norm is estimated rather than computed, for the sake of computation speed.
|
314
|
+
#
|
315
|
+
# Currently implemented norms are 1-norm, 2-norm, Frobenius, Infinity.
|
316
|
+
# A minus on the 1, 2 and inf norms returns the minimum instead of the maximum value.
|
317
|
+
#
|
318
|
+
# Tested mainly with dense matrices. Further checks and modifications might
|
319
|
+
# be necessary for sparse matrices.
|
320
|
+
#
|
321
|
+
# * *Returns* :
|
322
|
+
# - The selected norm of the matrix.
|
323
|
+
# * *Raises* :
|
324
|
+
# - +NotImplementedError+ -> norm can be calculated only for 2D matrices
|
325
|
+
# - +ArgumentError+ -> unrecognized norm
|
326
|
+
#
|
327
|
+
def matrix_norm type = 2
|
328
|
+
raise(NotImplementedError, "norm can be calculated only for 2D matrices") unless self.dim == 2
|
329
|
+
raise(NotImplementedError, "norm only implemented for dense storage") unless self.stype == :dense
|
330
|
+
raise(ArgumentError, "norm not defined for byte dtype")if self.dtype == :byte
|
331
|
+
case type
|
332
|
+
when nil, 2, -2
|
333
|
+
return self.two_matrix_norm (type == -2)
|
334
|
+
when 1, -1
|
335
|
+
return self.one_matrix_norm (type == -1)
|
336
|
+
when :frobenius, :fro
|
337
|
+
return self.fro_matrix_norm
|
338
|
+
when :infinity, :inf, :'-inf', :'-infinity'
|
339
|
+
return self.inf_matrix_norm (type == :'-inf' || type == :'-infinity')
|
340
|
+
else
|
341
|
+
raise ArgumentError.new("argument must be a valid integer or symbol")
|
342
|
+
end
|
343
|
+
end
|
344
|
+
|
345
|
+
# Calculate the variance co-variance matrix
|
346
|
+
#
|
347
|
+
# == Options
|
348
|
+
#
|
349
|
+
# * +:for_sample_data+ - Default true. If set to false will consider the denominator for
|
350
|
+
# population data (i.e. N, as opposed to N-1 for sample data).
|
351
|
+
#
|
352
|
+
# == References
|
353
|
+
#
|
354
|
+
# * http://stattrek.com/matrix-algebra/covariance-matrix.aspx
|
355
|
+
def cov(opts={})
|
356
|
+
raise TypeError, "Only works for non-integer dtypes" if integer_dtype?
|
357
|
+
opts = {
|
358
|
+
for_sample_data: true
|
359
|
+
}.merge(opts)
|
360
|
+
|
361
|
+
denominator = opts[:for_sample_data] ? rows - 1 : rows
|
362
|
+
ones = NMatrix.ones [rows,1]
|
363
|
+
deviation_scores = self - ones.dot(ones.transpose).dot(self) / rows
|
364
|
+
deviation_scores.transpose.dot(deviation_scores) / denominator
|
365
|
+
end
|
366
|
+
|
367
|
+
# Calculate the correlation matrix.
|
368
|
+
def corr
|
369
|
+
raise NotImplementedError, "Does not work for complex dtypes" if complex_dtype?
|
370
|
+
standard_deviation = std
|
371
|
+
cov / (standard_deviation.transpose.dot(standard_deviation))
|
372
|
+
end
|
373
|
+
|
374
|
+
# Raise a square matrix to a power. Be careful of numeric overflows!
|
375
|
+
# In case *n* is 0, an identity matrix of the same dimension is returned. In case
|
376
|
+
# of negative *n*, the matrix is inverted and the absolute value of *n* taken
|
377
|
+
# for computing the power.
|
378
|
+
#
|
379
|
+
# == Arguments
|
380
|
+
#
|
381
|
+
# * +n+ - Integer to which self is to be raised.
|
382
|
+
#
|
383
|
+
# == References
|
384
|
+
#
|
385
|
+
# * R.G Dromey - How to Solve it by Computer. Link -
|
386
|
+
# http://www.amazon.com/Solve-Computer-Prentice-Hall-International-Science/dp/0134340019/ref=sr_1_1?ie=UTF8&qid=1422605572&sr=8-1&keywords=how+to+solve+it+by+computer
|
387
|
+
def pow n
|
388
|
+
raise ShapeError, "Only works with 2D square matrices." if
|
389
|
+
shape[0] != shape[1] or shape.size != 2
|
390
|
+
raise TypeError, "Only works with integer powers" unless n.is_a?(Integer)
|
391
|
+
|
392
|
+
sequence = (integer_dtype? ? self.cast(dtype: :int64) : self).clone
|
393
|
+
product = NMatrix.eye shape[0], dtype: sequence.dtype, stype: sequence.stype
|
394
|
+
|
395
|
+
if n == 0
|
396
|
+
return NMatrix.eye(shape, dtype: dtype, stype: stype)
|
397
|
+
elsif n == 1
|
398
|
+
return sequence
|
399
|
+
elsif n < 0
|
400
|
+
n = n.abs
|
401
|
+
sequence.invert!
|
402
|
+
product = NMatrix.eye shape[0], dtype: sequence.dtype, stype: sequence.stype
|
403
|
+
end
|
404
|
+
|
405
|
+
# Decompose n to reduce the number of multiplications.
|
406
|
+
while n > 0
|
407
|
+
product = product.dot(sequence) if n % 2 == 1
|
408
|
+
n = n / 2
|
409
|
+
sequence = sequence.dot(sequence)
|
410
|
+
end
|
411
|
+
|
412
|
+
product
|
413
|
+
end
|
414
|
+
|
415
|
+
# Compute the Kronecker product of +self+ and other NMatrix
|
416
|
+
#
|
417
|
+
# === Arguments
|
418
|
+
#
|
419
|
+
# * +mat+ - A 2D NMatrix object
|
420
|
+
#
|
421
|
+
# === Usage
|
422
|
+
#
|
423
|
+
# a = NMatrix.new([2,2],[1,2,
|
424
|
+
# 3,4])
|
425
|
+
# b = NMatrix.new([2,3],[1,1,1,
|
426
|
+
# 1,1,1], dtype: :float64)
|
427
|
+
# a.kron_prod(b) # => [ [1.0, 1.0, 1.0, 2.0, 2.0, 2.0]
|
428
|
+
# [1.0, 1.0, 1.0, 2.0, 2.0, 2.0]
|
429
|
+
# [3.0, 3.0, 3.0, 4.0, 4.0, 4.0]
|
430
|
+
# [3.0, 3.0, 3.0, 4.0, 4.0, 4.0] ]
|
431
|
+
#
|
432
|
+
def kron_prod(mat)
|
433
|
+
unless self.dimensions==2 and mat.dimensions==2
|
434
|
+
raise ShapeError, "Implemented for 2D NMatrix objects only."
|
435
|
+
end
|
436
|
+
|
437
|
+
# compute the shape [n,m] of the product matrix
|
438
|
+
n, m = self.shape[0]*mat.shape[0], self.shape[1]*mat.shape[1]
|
439
|
+
# compute the entries of the product matrix
|
440
|
+
kron_prod_array = []
|
441
|
+
if self.yale?
|
442
|
+
# +:yale+ requires to get the row by copy in order to apply +#transpose+ to it
|
443
|
+
self.each_row(getby=:copy) do |selfr|
|
444
|
+
mat.each_row do |matr|
|
445
|
+
kron_prod_array += (selfr.transpose.dot matr).to_flat_a
|
446
|
+
end
|
447
|
+
end
|
448
|
+
else
|
449
|
+
self.each_row do |selfr|
|
450
|
+
mat.each_row do |matr|
|
451
|
+
kron_prod_array += (selfr.transpose.dot matr).to_flat_a
|
452
|
+
end
|
453
|
+
end
|
454
|
+
end
|
455
|
+
|
456
|
+
NMatrix.new([n,m], kron_prod_array)
|
457
|
+
end
|
458
|
+
|
459
|
+
#
|
460
|
+
# call-seq:
|
461
|
+
# trace -> Numeric
|
462
|
+
#
|
463
|
+
# Calculates the trace of an nxn matrix.
|
464
|
+
#
|
465
|
+
# * *Raises* :
|
466
|
+
# - +ShapeError+ -> Expected square matrix
|
467
|
+
#
|
468
|
+
# * *Returns* :
|
469
|
+
# - The trace of the matrix (a numeric value)
|
470
|
+
#
|
471
|
+
def trace
|
472
|
+
raise(ShapeError, "Expected square matrix") unless self.shape[0] == self.shape[1] && self.dim == 2
|
473
|
+
|
474
|
+
(0...self.shape[0]).inject(0) do |total,i|
|
475
|
+
total + self[i,i]
|
476
|
+
end
|
477
|
+
end
|
478
|
+
|
479
|
+
##
|
480
|
+
# call-seq:
|
481
|
+
# mean() -> NMatrix
|
482
|
+
# mean(dimen) -> NMatrix
|
483
|
+
#
|
484
|
+
# Calculates the mean along the specified dimension.
|
485
|
+
#
|
486
|
+
# This will force integer types to float64 dtype.
|
487
|
+
#
|
488
|
+
# @see #inject_rank
|
489
|
+
#
|
490
|
+
def mean(dimen=0)
|
491
|
+
reduce_dtype = nil
|
492
|
+
if integer_dtype? then
|
493
|
+
reduce_dtype = :float64
|
494
|
+
end
|
495
|
+
inject_rank(dimen, 0.0, reduce_dtype) do |mean, sub_mat|
|
496
|
+
mean + sub_mat
|
497
|
+
end / shape[dimen]
|
498
|
+
end
|
499
|
+
|
500
|
+
##
|
501
|
+
# call-seq:
|
502
|
+
# sum() -> NMatrix
|
503
|
+
# cumsum() -> NMatrix
|
504
|
+
# sum(dimen) -> NMatrix
|
505
|
+
# cumsum(dimen) -> NMatrix
|
506
|
+
#
|
507
|
+
# Calculates the sum along the specified dimension.
|
508
|
+
#
|
509
|
+
# @see #inject_rank
|
510
|
+
def sum(dimen=0)
|
511
|
+
inject_rank(dimen, 0.0) do |sum, sub_mat|
|
512
|
+
sum + sub_mat
|
513
|
+
end
|
514
|
+
end
|
515
|
+
alias :cumsum :sum
|
516
|
+
|
517
|
+
##
|
518
|
+
# call-seq:
|
519
|
+
# min() -> NMatrix
|
520
|
+
# min(dimen) -> NMatrix
|
521
|
+
#
|
522
|
+
# Calculates the minimum along the specified dimension.
|
523
|
+
#
|
524
|
+
# @see #inject_rank
|
525
|
+
#
|
526
|
+
def min(dimen=0)
|
527
|
+
inject_rank(dimen) do |min, sub_mat|
|
528
|
+
if min.is_a? NMatrix then
|
529
|
+
min * (min <= sub_mat).cast(self.stype, self.dtype) + ((min)*0.0 + (min > sub_mat).cast(self.stype, self.dtype)) * sub_mat
|
530
|
+
else
|
531
|
+
min <= sub_mat ? min : sub_mat
|
532
|
+
end
|
533
|
+
end
|
534
|
+
end
|
535
|
+
|
536
|
+
##
|
537
|
+
# call-seq:
|
538
|
+
# max() -> NMatrix
|
539
|
+
# max(dimen) -> NMatrix
|
540
|
+
#
|
541
|
+
# Calculates the maximum along the specified dimension.
|
542
|
+
#
|
543
|
+
# @see #inject_rank
|
544
|
+
#
|
545
|
+
def max(dimen=0)
|
546
|
+
inject_rank(dimen) do |max, sub_mat|
|
547
|
+
if max.is_a? NMatrix then
|
548
|
+
max * (max >= sub_mat).cast(self.stype, self.dtype) + ((max)*0.0 + (max < sub_mat).cast(self.stype, self.dtype)) * sub_mat
|
549
|
+
else
|
550
|
+
max >= sub_mat ? max : sub_mat
|
551
|
+
end
|
552
|
+
end
|
553
|
+
end
|
554
|
+
|
555
|
+
|
556
|
+
##
|
557
|
+
# call-seq:
|
558
|
+
# variance() -> NMatrix
|
559
|
+
# variance(dimen) -> NMatrix
|
560
|
+
#
|
561
|
+
# Calculates the sample variance along the specified dimension.
|
562
|
+
#
|
563
|
+
# This will force integer types to float64 dtype.
|
564
|
+
#
|
565
|
+
# @see #inject_rank
|
566
|
+
#
|
567
|
+
def variance(dimen=0)
|
568
|
+
reduce_dtype = nil
|
569
|
+
if integer_dtype? then
|
570
|
+
reduce_dtype = :float64
|
571
|
+
end
|
572
|
+
m = mean(dimen)
|
573
|
+
inject_rank(dimen, 0.0, reduce_dtype) do |var, sub_mat|
|
574
|
+
var + (m - sub_mat)*(m - sub_mat)/(shape[dimen]-1)
|
575
|
+
end
|
576
|
+
end
|
577
|
+
|
578
|
+
##
|
579
|
+
# call-seq:
|
580
|
+
# std() -> NMatrix
|
581
|
+
# std(dimen) -> NMatrix
|
582
|
+
#
|
583
|
+
#
|
584
|
+
# Calculates the sample standard deviation along the specified dimension.
|
585
|
+
#
|
586
|
+
# This will force integer types to float64 dtype.
|
587
|
+
#
|
588
|
+
# @see #inject_rank
|
589
|
+
#
|
590
|
+
def std(dimen=0)
|
591
|
+
variance(dimen).sqrt
|
592
|
+
end
|
593
|
+
|
594
|
+
|
595
|
+
#
|
596
|
+
# call-seq:
|
597
|
+
# abs_dtype -> Symbol
|
598
|
+
#
|
599
|
+
# Returns the dtype of the result of a call to #abs. In most cases, this is the same as dtype; it should only differ
|
600
|
+
# for :complex64 (where it's :float32) and :complex128 (:float64).
|
601
|
+
def abs_dtype
|
602
|
+
if self.dtype == :complex64
|
603
|
+
:float32
|
604
|
+
elsif self.dtype == :complex128
|
605
|
+
:float64
|
606
|
+
else
|
607
|
+
self.dtype
|
608
|
+
end
|
609
|
+
end
|
610
|
+
|
611
|
+
|
612
|
+
#
|
613
|
+
# call-seq:
|
614
|
+
# abs -> NMatrix
|
615
|
+
#
|
616
|
+
# Maps all values in a matrix to their absolute values.
|
617
|
+
def abs
|
618
|
+
if stype == :dense
|
619
|
+
self.__dense_map__ { |v| v.abs }
|
620
|
+
elsif stype == :list
|
621
|
+
# FIXME: Need __list_map_stored__, but this will do for now.
|
622
|
+
self.__list_map_merged_stored__(nil, nil) { |v,dummy| v.abs }
|
623
|
+
else
|
624
|
+
self.__yale_map_stored__ { |v| v.abs }
|
625
|
+
end.cast(self.stype, abs_dtype)
|
626
|
+
end
|
627
|
+
|
628
|
+
# Norm calculation methods
|
629
|
+
# Frobenius norm: the Euclidean norm of the matrix, treated as if it were a vector
|
630
|
+
def fro_matrix_norm
|
631
|
+
#float64 has to be used in any case, since nrm2 will not yield correct result for float32
|
632
|
+
self_cast = self.cast(:dtype => :float64)
|
633
|
+
|
634
|
+
column_vector = self_cast.reshape([self.size, 1])
|
635
|
+
|
636
|
+
return column_vector.nrm2
|
637
|
+
end
|
638
|
+
|
639
|
+
# 2-norm: the largest/smallest singular value of the matrix
|
640
|
+
def two_matrix_norm minus = false
|
641
|
+
|
642
|
+
self_cast = self.cast(:dtype => :float64)
|
643
|
+
|
644
|
+
#TODO: confirm if this is the desired svd calculation
|
645
|
+
svd = self_cast.gesvd
|
646
|
+
return svd[1][0, 0] unless minus
|
647
|
+
return svd[1][svd[1].rows-1, svd[1].cols-1]
|
648
|
+
end
|
649
|
+
|
650
|
+
# 1-norm: the maximum/minimum absolute column sum of the matrix
|
651
|
+
def one_matrix_norm minus = false
|
652
|
+
#TODO: change traversing method for sparse matrices
|
653
|
+
number_of_columns = self.cols
|
654
|
+
col_sums = []
|
655
|
+
|
656
|
+
number_of_columns.times do |i|
|
657
|
+
col_sums << self.col(i).inject(0) { |sum, number| sum += number.abs}
|
658
|
+
end
|
659
|
+
|
660
|
+
return col_sums.max unless minus
|
661
|
+
return col_sums.min
|
662
|
+
end
|
663
|
+
|
664
|
+
# Infinity norm: the maximum/minimum absolute row sum of the matrix
|
665
|
+
def inf_matrix_norm minus = false
|
666
|
+
number_of_rows = self.rows
|
667
|
+
row_sums = []
|
668
|
+
|
669
|
+
number_of_rows.times do |i|
|
670
|
+
row_sums << self.row(i).inject(0) { |sum, number| sum += number.abs}
|
671
|
+
end
|
672
|
+
|
673
|
+
return row_sums.max unless minus
|
674
|
+
return row_sums.min
|
675
|
+
end
|
676
|
+
|
677
|
+
#
|
678
|
+
# call-seq:
|
679
|
+
# positive_definite? -> boolean
|
680
|
+
#
|
681
|
+
# A matrix is positive definite if it’s symmetric and all its eigenvalues are positive
|
682
|
+
#
|
683
|
+
# * *Returns* :
|
684
|
+
# - A boolean value telling if the NMatrix is positive definite or not.
|
685
|
+
# * *Raises* :
|
686
|
+
# - +ShapeError+ -> Must be used on square matrices.
|
687
|
+
#
|
688
|
+
def positive_definite?
|
689
|
+
raise(ShapeError, "positive definite calculated only for square matrices") unless
|
690
|
+
self.dim == 2 && self.shape[0] == self.shape[1]
|
691
|
+
cond = 0
|
692
|
+
while cond != self.cols
|
693
|
+
if self[0..cond, 0..cond].det <= 0
|
694
|
+
return false
|
695
|
+
end
|
696
|
+
cond += 1
|
697
|
+
end
|
698
|
+
true
|
699
|
+
end
|
700
|
+
|
701
|
+
#
|
702
|
+
# call-seq:
|
703
|
+
# svd_rank() -> int
|
704
|
+
# svd_rank(tolerence) ->int
|
705
|
+
# Gives rank of the matrix based on the singular value decomposition.
|
706
|
+
# The rank of a matrix is computed as the number of diagonal elements in Sigma that are larger than a tolerance
|
707
|
+
#
|
708
|
+
#* *Returns* :
|
709
|
+
# - An integer equal to the rank of the matrix
|
710
|
+
#* *Raises* :
|
711
|
+
# - +ShapeError+ -> Is only computable on 2-D matrices
|
712
|
+
#
|
713
|
+
def svd_rank(tolerence="default")
|
714
|
+
raise(ShapeError, "rank calculated only for 2-D matrices") unless
|
715
|
+
self.dim == 2
|
716
|
+
|
717
|
+
sigmas = self.gesvd[1].to_a.flatten
|
718
|
+
eps = NMatrix::FLOAT64_EPSILON
|
719
|
+
|
720
|
+
# epsilon depends on the width of the number
|
721
|
+
if (self.dtype == :float32 || self.dtype == :complex64)
|
722
|
+
eps = NMatrix::FLOAT32_EPSILON
|
723
|
+
end
|
724
|
+
case tolerence
|
725
|
+
when "default"
|
726
|
+
tolerence = self.shape.max * sigmas.max * eps # tolerence of a Matrix A is max(size(A))*eps(norm(A)). norm(A) is nearly equal to max(sigma of A)
|
727
|
+
end
|
728
|
+
return sigmas.map { |x| x > tolerence ? 1 : 0 }.reduce(:+)
|
729
|
+
end
|
730
|
+
|
731
|
+
|
732
|
+
|
733
|
+
protected
|
734
|
+
# Define the element-wise operations for lists. Note that the __list_map_merged_stored__ iterator returns a Ruby Object
|
735
|
+
# matrix, which we then cast back to the appropriate type. If you don't want that, you can redefine these functions in
|
736
|
+
# your own code.
|
737
|
+
{add: :+, sub: :-, mul: :*, div: :/, pow: :**, mod: :%}.each_pair do |ewop, op|
|
738
|
+
define_method("__list_elementwise_#{ewop}__") do |rhs|
|
739
|
+
self.__list_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }.cast(stype, NMatrix.upcast(dtype, rhs.dtype))
|
740
|
+
end
|
741
|
+
define_method("__dense_elementwise_#{ewop}__") do |rhs|
|
742
|
+
self.__dense_map_pair__(rhs) { |l,r| l.send(op,r) }.cast(stype, NMatrix.upcast(dtype, rhs.dtype))
|
743
|
+
end
|
744
|
+
define_method("__yale_elementwise_#{ewop}__") do |rhs|
|
745
|
+
self.__yale_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }.cast(stype, NMatrix.upcast(dtype, rhs.dtype))
|
746
|
+
end
|
747
|
+
define_method("__list_scalar_#{ewop}__") do |rhs|
|
748
|
+
self.__list_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }.cast(stype, NMatrix.upcast(dtype, NMatrix.min_dtype(rhs)))
|
749
|
+
end
|
750
|
+
define_method("__yale_scalar_#{ewop}__") do |rhs|
|
751
|
+
self.__yale_map_stored__ { |l| l.send(op,rhs) }.cast(stype, NMatrix.upcast(dtype, NMatrix.min_dtype(rhs)))
|
752
|
+
end
|
753
|
+
define_method("__dense_scalar_#{ewop}__") do |rhs|
|
754
|
+
self.__dense_map__ { |l| l.send(op,rhs) }.cast(stype, NMatrix.upcast(dtype, NMatrix.min_dtype(rhs)))
|
755
|
+
end
|
756
|
+
end
|
757
|
+
|
758
|
+
# These don't actually take an argument -- they're called reverse-polish style on the matrix.
|
759
|
+
# This group always gets casted to float64.
|
760
|
+
[:log, :log2, :log10, :sqrt, :sin, :cos, :tan, :acos, :asin, :atan, :cosh, :sinh, :tanh, :acosh,
|
761
|
+
:asinh, :atanh, :exp, :erf, :erfc, :gamma, :cbrt, :round].each do |ewop|
|
762
|
+
define_method("__list_unary_#{ewop}__") do
|
763
|
+
self.__list_map_stored__(nil) { |l| Math.send(ewop, l) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
764
|
+
end
|
765
|
+
define_method("__yale_unary_#{ewop}__") do
|
766
|
+
self.__yale_map_stored__ { |l| Math.send(ewop, l) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
767
|
+
end
|
768
|
+
define_method("__dense_unary_#{ewop}__") do
|
769
|
+
self.__dense_map__ { |l| Math.send(ewop, l) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
770
|
+
end
|
771
|
+
end
|
772
|
+
|
773
|
+
#:stopdoc:
|
774
|
+
# log takes an optional single argument, the base. Default to natural log.
|
775
|
+
def __list_unary_log__(base)
|
776
|
+
self.__list_map_stored__(nil) { |l| Math.log(l, base) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
777
|
+
end
|
778
|
+
|
779
|
+
def __yale_unary_log__(base)
|
780
|
+
self.__yale_map_stored__ { |l| Math.log(l, base) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
781
|
+
end
|
782
|
+
|
783
|
+
def __dense_unary_log__(base)
|
784
|
+
self.__dense_map__ { |l| Math.log(l, base) }.cast(stype, NMatrix.upcast(dtype, :float64))
|
785
|
+
end
|
786
|
+
|
787
|
+
# These are for negating matrix contents using -@
|
788
|
+
def __list_unary_negate__
|
789
|
+
self.__list_map_stored__(nil) { |l| -l }.cast(stype, dtype)
|
790
|
+
end
|
791
|
+
|
792
|
+
def __yale_unary_negate__
|
793
|
+
self.__yale_map_stored__ { |l| -l }.cast(stype, dtype)
|
794
|
+
end
|
795
|
+
|
796
|
+
def __dense_unary_negate__
|
797
|
+
self.__dense_map__ { |l| -l }.cast(stype, dtype)
|
798
|
+
end
|
799
|
+
#:startdoc:
|
800
|
+
|
801
|
+
# These are for rounding each value of a matrix. Takes an optional argument
|
802
|
+
def __list_unary_round__(precision)
|
803
|
+
if self.complex_dtype?
|
804
|
+
self.__list_map_stored__(nil) { |l| Complex(l.real.round(precision), l.imag.round(precision)) }
|
805
|
+
.cast(stype, dtype)
|
806
|
+
else
|
807
|
+
self.__list_map_stored__(nil) { |l| l.round(precision) }.cast(stype, dtype)
|
808
|
+
end
|
809
|
+
end
|
810
|
+
|
811
|
+
def __yale_unary_round__(precision)
|
812
|
+
if self.complex_dtype?
|
813
|
+
self.__yale_map_stored__ { |l| Complex(l.real.round(precision), l.imag.round(precision)) }
|
814
|
+
.cast(stype, dtype)
|
815
|
+
else
|
816
|
+
self.__yale_map_stored__ { |l| l.round(precision) }.cast(stype, dtype)
|
817
|
+
end
|
818
|
+
end
|
819
|
+
|
820
|
+
def __dense_unary_round__(precision)
|
821
|
+
if self.complex_dtype?
|
822
|
+
self.__dense_map__ { |l| Complex(l.real.round(precision), l.imag.round(precision)) }
|
823
|
+
.cast(stype, dtype)
|
824
|
+
else
|
825
|
+
self.__dense_map__ { |l| l.round(precision) }.cast(stype, dtype)
|
826
|
+
end
|
827
|
+
end
|
828
|
+
|
829
|
+
# These are for calculating the floor or ceil of matrix
|
830
|
+
def dtype_for_floor_or_ceil
|
831
|
+
if self.integer_dtype? or [:complex64, :complex128, :object].include?(self.dtype)
|
832
|
+
return_dtype = dtype
|
833
|
+
elsif [:float32, :float64].include?(self.dtype)
|
834
|
+
return_dtype = :int64
|
835
|
+
end
|
836
|
+
|
837
|
+
return_dtype
|
838
|
+
end
|
839
|
+
|
840
|
+
[:floor, :ceil].each do |meth|
|
841
|
+
define_method("__list_unary_#{meth}__") do
|
842
|
+
return_dtype = dtype_for_floor_or_ceil
|
843
|
+
|
844
|
+
if [:complex64, :complex128].include?(self.dtype)
|
845
|
+
self.__list_map_stored__(nil) { |l| Complex(l.real.send(meth), l.imag.send(meth)) }.cast(stype, return_dtype)
|
846
|
+
else
|
847
|
+
self.__list_map_stored__(nil) { |l| l.send(meth) }.cast(stype, return_dtype)
|
848
|
+
end
|
849
|
+
end
|
850
|
+
|
851
|
+
define_method("__yale_unary_#{meth}__") do
|
852
|
+
return_dtype = dtype_for_floor_or_ceil
|
853
|
+
|
854
|
+
if [:complex64, :complex128].include?(self.dtype)
|
855
|
+
self.__yale_map_stored__ { |l| Complex(l.real.send(meth), l.imag.send(meth)) }.cast(stype, return_dtype)
|
856
|
+
else
|
857
|
+
self.__yale_map_stored__ { |l| l.send(meth) }.cast(stype, return_dtype)
|
858
|
+
end
|
859
|
+
end
|
860
|
+
|
861
|
+
define_method("__dense_unary_#{meth}__") do
|
862
|
+
return_dtype = dtype_for_floor_or_ceil
|
863
|
+
|
864
|
+
if [:complex64, :complex128].include?(self.dtype)
|
865
|
+
self.__dense_map__ { |l| Complex(l.real.send(meth), l.imag.send(meth)) }.cast(stype, return_dtype)
|
866
|
+
else
|
867
|
+
self.__dense_map__ { |l| l.send(meth) }.cast(stype, return_dtype)
|
868
|
+
end
|
869
|
+
end
|
870
|
+
end
|
871
|
+
|
872
|
+
# These take two arguments. One might be a matrix, and one might be a scalar.
|
873
|
+
# See also monkeys.rb, which contains Math module patches to let the first
|
874
|
+
# arg be a scalar
|
875
|
+
[:atan2, :ldexp, :hypot].each do |ewop|
|
876
|
+
define_method("__list_elementwise_#{ewop}__") do |rhs,order|
|
877
|
+
if order then
|
878
|
+
self.__list_map_merged_stored__(rhs, nil) { |r,l| Math.send(ewop,l,r) }
|
879
|
+
else
|
880
|
+
self.__list_map_merged_stored__(rhs, nil) { |l,r| Math.send(ewop,l,r) }
|
881
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
882
|
+
end
|
883
|
+
|
884
|
+
define_method("__dense_elementwise_#{ewop}__") do |rhs, order|
|
885
|
+
if order then
|
886
|
+
self.__dense_map_pair__(rhs) { |r,l| Math.send(ewop,l,r) }
|
887
|
+
else
|
888
|
+
self.__dense_map_pair__(rhs) { |l,r| Math.send(ewop,l,r) }
|
889
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
890
|
+
end
|
891
|
+
|
892
|
+
define_method("__yale_elementwise_#{ewop}__") do |rhs, order|
|
893
|
+
if order then
|
894
|
+
self.__yale_map_merged_stored__(rhs, nil) { |r,l| Math.send(ewop,l,r) }
|
895
|
+
else
|
896
|
+
self.__yale_map_merged_stored__(rhs, nil) { |l,r| Math.send(ewop,l,r) }
|
897
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
898
|
+
end
|
899
|
+
|
900
|
+
define_method("__list_scalar_#{ewop}__") do |rhs,order|
|
901
|
+
if order then
|
902
|
+
self.__list_map_stored__(nil) { |l| Math.send(ewop, rhs, l) }
|
903
|
+
else
|
904
|
+
self.__list_map_stored__(nil) { |l| Math.send(ewop, l, rhs) }
|
905
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
906
|
+
end
|
907
|
+
|
908
|
+
define_method("__yale_scalar_#{ewop}__") do |rhs,order|
|
909
|
+
if order then
|
910
|
+
self.__yale_map_stored__ { |l| Math.send(ewop, rhs, l) }
|
911
|
+
else
|
912
|
+
self.__yale_map_stored__ { |l| Math.send(ewop, l, rhs) }
|
913
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
914
|
+
end
|
915
|
+
|
916
|
+
define_method("__dense_scalar_#{ewop}__") do |rhs,order|
|
917
|
+
if order
|
918
|
+
self.__dense_map__ { |l| Math.send(ewop, rhs, l) }
|
919
|
+
else
|
920
|
+
self.__dense_map__ { |l| Math.send(ewop, l, rhs) }
|
921
|
+
end.cast(stype, NMatrix.upcast(dtype, :float64))
|
922
|
+
end
|
923
|
+
end
|
924
|
+
|
925
|
+
# Equality operators do not involve a cast. We want to get back matrices of TrueClass and FalseClass.
|
926
|
+
{eqeq: :==, neq: :!=, lt: :<, gt: :>, leq: :<=, geq: :>=}.each_pair do |ewop, op|
|
927
|
+
define_method("__list_elementwise_#{ewop}__") do |rhs|
|
928
|
+
self.__list_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }
|
929
|
+
end
|
930
|
+
define_method("__dense_elementwise_#{ewop}__") do |rhs|
|
931
|
+
self.__dense_map_pair__(rhs) { |l,r| l.send(op,r) }
|
932
|
+
end
|
933
|
+
define_method("__yale_elementwise_#{ewop}__") do |rhs|
|
934
|
+
self.__yale_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }
|
935
|
+
end
|
936
|
+
|
937
|
+
define_method("__list_scalar_#{ewop}__") do |rhs|
|
938
|
+
self.__list_map_merged_stored__(rhs, nil) { |l,r| l.send(op,r) }
|
939
|
+
end
|
940
|
+
define_method("__yale_scalar_#{ewop}__") do |rhs|
|
941
|
+
self.__yale_map_stored__ { |l| l.send(op,rhs) }
|
942
|
+
end
|
943
|
+
define_method("__dense_scalar_#{ewop}__") do |rhs|
|
944
|
+
self.__dense_map__ { |l| l.send(op,rhs) }
|
945
|
+
end
|
946
|
+
end
|
947
|
+
end
|
948
|
+
|
949
|
+
if jruby?
|
950
|
+
require_relative "./jruby/math.rb"
|
951
|
+
else
|
952
|
+
require_relative "./cruby/math.rb"
|
953
|
+
end
|