Sha256: 2c4c27c2e520c11d26de5ca221a22e8f033ed815e918ca9cd8ca3b550c47aa90
Contents?: true
Size: 588 Bytes
Versions: 9
Compression:
Stored size: 588 Bytes
Contents
module TensorStream
module Train
# High Level implementation of the gradient descent algorithm
class GradientDescentOptimizer
attr_accessor :learning_rate
def initialize(learning_rate, _options = {})
@learning_rate = learning_rate
end
def minimize(cost)
trainable_vars = TensorStream.trainable_variables
derivatives = TensorStream.gradients(cost, trainable_vars)
trainable_vars.each_with_index.collect do |var, index|
var.assign_sub(derivatives[index] * @learning_rate)
end
end
end
end
end
Version data entries
9 entries across 9 versions & 1 rubygems