IMDB Movie reviews sentiment classification
- Import required libraries and the dataset
- Visualise a sample of dataset
- Data preprocessing
- Sequential Model Building with Activation for each layer
- Compile with binary CE loss and metric accuracy
- Train Model with cross validation, with total time taken shown for 20 epochs
- Visualise Loss and Accuracy history
- Show Confusion Matrix for validation dataset
- Network Depth : 1 Hidden Layer of 64 units 2 Hidden Layer of 64-32 units
- Regularisations :
- Dropout of ratio 0.25
- Dropout of ratio 0.25 with L2 regulariser with factor 1e-04
- Activations :
- ReLU for every hidden layer
- Sigmoid for every hidden layer (sigmoid for final layer)
- Batch Size :
- 64
- 256
- Optimisers :
- Adam with lr 1e-03
- SGD with lr 1e-02 and momentum 0.95