Sentiment Analysis of IMDb dataset using neural networks

Implemented sentiment analysis on IMDb movie review data using feed-forward neural networks and SGD optimizer. Achieved maximum validation accuracy of 88.5% in 10 epochs.

• Created feature matrices using BOW approach
• Reduced the vocabulary size to save memory
• Used minibatch gradient descent
• Used Nesterov momentum

Training and validation loss per epoch

Validation accuracy per epoch

Loss per 1000 random batches: