Learning in Python: Learn Data Science and Machine Learning with Modern Neural Networks composed in Python, Theano, and TensorFlow by MADHAN KUMAR

Author:MADHAN KUMAR [KUMAR, MADHAN]
Language: eng
Format: azw3, epub
Published: 2020-11-05T00:00:00+00:00

A neural network in Theano

To begin with, I will characterize my information sources, yields, and loads (the loads will be shared factors):

thX = T.matrix('X')

thT = T.matrix('T')

W1 = theano.shared(np.random.randn(D, M), 'W1')

W2 = theano.shared(np.random.randn(M, K), ‘W2')

Notice I've added a "th" prefix to the Theano factors since I will call my real information, which are Numpy clusters, X and T.

Review that M is the number of units in the shrouded layer.

Next, I characterize the feedforward activity.

thZ = T.tanh( thX.dot(W1))

thY = T.nnet.softmax( thZ.dot(W2) )

T.tanh is a non-straight capacity like the sigmoid, however, it runs between - 1 and +1.

Next, I characterize my cost work and my expectation work (this is utilized to figure the arrangement mistake later).

cost = -(thT * T.log(thY)).sum()

prediction = T.argmax(thY, axis=1)

What's more, I characterize my updated articulations. (notice how Theano has a capacity to ascertain angles!)

update_W1 = W1 - lr*T.grad(cost, W1)

update_W2 = W2 - lr*T.grad(cost, W2)

I make train work like the basic model above:

train = theano.function(

inputs=[thX, thT],

updates=[(W1, update_W1),(W2, update_W2)],

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.