1
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 1 year has passed since last update.

wslのtensorflowでkerasのbackend その7

Posted at

概要

wslのtensorflowでkerasのbackendを試す。
練習問題、やってみた。

練習問題

kerasのbackendでxorを学習したウェイトとバイアスを使って、elixirで再現せよ。

サンプルコード

defmodule Nn do
	def mul(a, b) do
		Enum.map(a, fn i ->
			i * b
		end)
	end
	def add(a, b) do
		for x <- 0..7 do
			Enum.at(a, x) + Enum.at(b, x)
		end
	end
	def sum(a, b) do
		for x <- 0..7 do
			Enum.at(a, x) * Enum.at(b, x)
		end
		|> Enum.sum
	end
	def tanh(v) do
		Enum.map(v, fn i ->
			a = :math.exp(i)
			b = :math.exp(-1 * i)
			(a - b) / (a + b)
		end)
	end
	def sigmoid(a) do
		Enum.map(a, fn i ->
			(1 / (1 + :math.exp(-1 * i)))
		end)
	end
end

w10 = [ 1.6427742,   1.292738,    0.13679211,  0.09276549,  0.31930393,  0.18661255,   1.0932094,  -0.7801719 ]
w11 = [ 1.4525393,   1.1745324,   0.20021456,  0.497589,    0.11109181,  0.33417672,  -0.44360426, -1.1322144 ]
b1 = [-0.20589589,  0.09548836,  0.54812616,  0.60974604,  0.15801114,  0.30214155,  0.38613093,  1.4007448 ]
w2 = [ 1.6341141,  1.1821091, -0.24378781, -0.2940346, -0.22311434, -0.34191525, -0.79992425,  1.717806  ]
b2 = -0.5830822

for i <- 0..1, j <- 0..1 do
	h = Nn.mul(w10, i)
	l = Nn.mul(w11, j)
	h = Nn.add(h, l)
	h = Nn.add(h, b1)
	h = Nn.tanh(h)
	y = Nn.sum(h, w2)
	y = y + b2
	y = Nn.tanh([y])
	y = Enum.at(y, 0)
	IO.puts "#{i} #{j} #{y}"
end



実行結果

0 0 0.007566404874442216
0 1 0.9309703232706648
1 0 0.9168728513830984
1 1 0.011421405940028202


成果物

以上。

1
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
1
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?