How to train and evaluate a bidirectional recurrent neural network (BRNN) for sentiment analysis. This example is based on LSTMs, for better handling of long term dependencies, achieving an accuracy of around 97% on unseen data. An example of using this model client side in a chat app
How to train a feedforward DNN (deep neural network) for binary classification with normalization included in the model.
CNN (convolutional neural network) recognizing vegetables
How to fit a model with early stopping callback. Visualization of the loss reveal a local minimum early on, which are bypassed by increasing the patience param in the callback. In the end there's an explicit calculation of a forward pass using only the weights from keras.