TensorFlow is an open source platform for machine learning from Google. It can make us to build some AI applications easily. It is a popular deep learning platform in word.

In this page, we write some tutorials and examples on how to use tensorflow, you can build some AI applications by following our tutorials and examples.

In this tutorial, we simple introduce Matrix Multiplication (Matrix Product) in Deep Learning, you can learn how to compute it by following our tutorial.

In this tutorial, we write two examples to introduce how to compute hadamard product of two tensors in tensorflow, you can learn how to compute it by following our examples.

In this tutorial, we explain why we use cross entropy as loss function in deep learning and why we use H(y, ypred) but H(ypred,y). You can follow our tutorial to learn.

tf.clip_by_value() can clip a tenser values to min and max number in tensorflow. In this tutorial, we write an example to introduce how to use this function, you can follow our example to learn it.

In this tutorial, we write an example to introduce that we may get nan or inf value when computing kl divergence value in tensorflow. To fix this problem, you can follow our tutorial and learn how to process.

KL Divergence has two equations: Forward and Reverse. Which shoud be selected in our application? In this tutorial, we discuss the difference between them, you can lean and understand how to select them by following our tutorial.

Entropy, Cross Entropy and KL Divergence are common used algorithms in deep learning. What is relation among them? In this tutorial, we will discuss the relation of them to help you to understand easily.

In this tutorial, we introduceKullback-Leibler Divergence (KL Divergence) to beginners. Kullback-Leibler divergence can measure the difference between two probability distributions p(x) and q(x). Readers can understand how to compute it by following this tutorial.

In this tutorial, we write an example to fix NaN bug when using tf.svd() function to compute singular value decomposition in tensorflow. We use np.linalg.svd() function to replace tensorflow original tf.svd() when computing. I tested this method in our train model. It works well.