Class: MachineLearningWorkbench::Compressor::VectorQuantization

Inherits:
Object
  • Object
show all
Defined in:
lib/machine_learning_workbench/compressor/vector_quantization.rb

Constant Summary collapse

Verification = 
MachineLearningWorkbench::Tools::Verification

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(ncentrs:, dims:, vrange:, dtype:, lrate:, rseed: Random.new_seed) ⇒ VectorQuantization

Returns a new instance of VectorQuantization.



6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 6

def initialize ncentrs:, dims:, vrange:, dtype:, lrate:, rseed: Random.new_seed
  @rng = Random.new rseed
  @ncentrs = ncentrs
  @dtype = dtype
  @dims = dims
  @lrate = lrate
  @vrange = case vrange
  when Array
    raise ArgumentError, "vrange size not 2: #{vrange}" unless vrange.size == 2
    vrange.map &method(:Float)
  when Range
    [vrange.first, vrange.last].map &method(:Float)
  else
    raise ArgumentError, "vrange: unrecognized type: #{vrange.class}"
  end
  @centrs = ncentrs.times.map { new_centr }
end

Instance Attribute Details

#centrs ⇒ Object (readonly)

Returns the value of attribute centrs.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def centrs
  @centrs
end

#dims ⇒ Object (readonly)

Returns the value of attribute dims.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def dims
  @dims
end

#dtype ⇒ Object (readonly)

Returns the value of attribute dtype.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def dtype
  @dtype
end

#lrate ⇒ Object (readonly)

Returns the value of attribute lrate.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def lrate
  @lrate
end

#ncentrs ⇒ Object (readonly)

Returns the value of attribute ncentrs.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def ncentrs
  @ncentrs
end

#rng ⇒ Object (readonly)

Returns the value of attribute rng.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def rng
  @rng
end

#vrange ⇒ Object (readonly)

Returns the value of attribute vrange.



3
4
5
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 3

def vrange
  @vrange
end

Instance Method Details

#most_similar_centr(img) ⇒ Object

Returns index and similitude of most similar centroid to image



40
41
42
43
44
45
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 40

def most_similar_centr img
  simils = similarities img
  max_simil = simils.max
  max_idx = simils.index max_simil
  [max_idx, max_simil]
end

#new_centr ⇒ Object

Creates a new (random) centroid



25
26
27
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 25

def new_centr
  NMatrix.new(dims, dtype: dtype) { rng.rand Range.new *vrange }
end

#reconstr_error(img) ⇒ Object

Per-pixel errors in reconstructing image



53
54
55
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 53

def reconstr_error img
  reconstruction(img) - img
end

#reconstruction(img) ⇒ Object

Reconstruct image as its most similar centroid



48
49
50
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 48

def reconstruction img
  centrs[most_similar_centr(img).first]
end

#similarities(img) ⇒ Object Also known as: encode

Computes similarities between image and all centroids

Raises:

  • (NotImplementedError) 


30
31
32
33
34
35
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 30

def similarities img
  raise NotImplementedError if img.shape.size > 1
  # centrs.map { |c| c.dot(img).first }
  require 'parallel'
  Parallel.map(centrs) { |c| c.dot(img).first }
end

#train(img_lst, debug: false) ⇒ Object

Train on image list



66
67
68
69
70
71
72
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 66

def train img_lst, debug: false
  # Two ways here:
  # - Batch: canonical, centrs updated with each img
  # - Parallel: could be parallel either on simils or on training (?)
  # Unsure on the correctness of either Parallel, let's stick with Batch
  img_lst.each { |img| train_one img; print '.' if debug }
end

#train_one(img, simils: nil) ⇒ Object

Train on one image



58
59
60
61
62
63
 
# File 'lib/machine_learning_workbench/compressor/vector_quantization.rb', line 58

def train_one img, simils: nil
  trg_idx, _simil = simils || most_similar_centr(img)
  centrs[trg_idx] = centrs[trg_idx] * (1-lrate) + img * lrate
  Verification.in_range! centrs[trg_idx], vrange
  centrs[trg_idx]
end