DNN with Just Tensor Mathematics and Optimiser

At low-level, TensorFlow is used just like a mathematics library with matrix multiplication, plus, minus, square, sum, etc.

The magical things in machine learning with TensorFlow are the optimisers. They change the values of variables in the network to optimise it, ie. miminise the final loss.

Hereby model is trained (fit weights) to do XOR regression:

#core
import sys;

#libs
import tensorflow        as tf;
import numpy             as np;
import matplotlib.pyplot as pyplot;

'''
\brief Log
'''
def log(*Args):
  Str = "";
  for I in range(len(Args)):
    Str += str(Args[I]);
    if (I<len(Args)-1): Str += "\x20";
      
  tf.logging.info(Str);
#end def

'''
\brief Log a separator
'''
def log_sep():
  tf.logging.info("\n"+"="*80);
#end def

#PROGRAMME ENTRY POINT==========================================================
tf.logging.set_verbosity(tf.logging.INFO);
log("TensorFlow version:",tf.__version__);

#data
X = [[0,0],[0,1],[1,0],[1,1]];
Y = [[0],  [1],  [1],  [0]  ];
X = np.array(X, dtype=np.float32);
Y = np.array(Y, dtype=np.float32);

#dnn
Batch_Size = 4;

Input     = tf.placeholder(tf.float32, shape=[Batch_Size,2]);
Expected  = tf.placeholder(tf.float32, shape=[Batch_Size,1]);

Weight1   = tf.Variable(tf.random_uniform([2,20], -1,1), name="weight1");
Bias1     = tf.Variable(tf.random_uniform([  20], -1,1), name="bias1");
Hidden1   = tf.nn.relu(tf.matmul(Input,Weight1) + Bias1);

Weight2   = tf.Variable(tf.random_uniform([20,1], -1,1), name="weight2");
Bias2     = tf.Variable(tf.random_uniform([   1], -1,1), name="bias2");
Output    = tf.sigmoid(tf.matmul(Hidden1,Weight2) + Bias2);

Loss      = tf.reduce_sum(tf.square(Expected-Output));
Optimiser = tf.train.GradientDescentOptimizer(1e-1);
Training  = Optimiser.minimize(Loss);

#training
Sess = tf.Session();
Init = tf.global_variables_initializer();
Sess.run(Init);

Losses = [];
for I in range(10000):
  if (I%100==0):
    Lossvalue = Sess.run(Loss, feed_dict={Input:X, Expected:Y});
    print("Loss:",Lossvalue);
    Losses += [Lossvalue];
  #end if
  
  Sess.run(Training, feed_dict={Input:X, Expected:Y});
#end for

print("Eval:",Sess.run(Output, feed_dict={Input:X, Expected:Y}));
print("Loss curve:");
pyplot.plot(Losses);
#eof

Colab link:
https://colab.research.google.com/drive/1trM0syMQzNTEMXPztDIYjKfU9cFdW1lt