In this note book we implement a single layer LSTM based Recurrent Neural Network (RNN) classifier, for MNIST database of handwritten digits.
The note book organization is as follows.
- LSTM Based RNNs with TensorFlow
- Basics of LSTM Based RNN
- LSTM Based RNN in TensorFlow - A Closer Look into Details 3.1. Static VS Dynamic RNN 3.2. Important Notes on the shape of the input and outputs for RNN (tf.nn.dynamic_rnn) 3.3. Using Dynam RNN Requires Caution if we have inputs with different length 3.4. The Choice of LSTM Cell
- MNIST Dataset Overview
- RNN Implementation and The Code
If you are just interested to see the code and dont care about details, jump to Section IV of the notebook, or directly use this.