/Deep-Learning

Implementations of a variety of classic deep learning architectures for problems such as object recognition, sentiment analysis, word embeddings, etc.

Primary LanguagePython

Word2Vec

Implementation Word2Vec embedding layer using the IMDB sentiment analysis dataset and gensim.

Sentiment Analysis

Implementation of an LSTM in Keras for sentiment analysis on the IMDB sentiment analysis dataset.

Implementation of Bidirectional LSTM which improved accuracy from 0.8339 baseline to 0.87432.

CIFAR-10

Implements a CNN in Keras for image classification on the CIFAR-10 dataset (https://www.cs.toronto.edu/~kriz/cifar.html). My model: 83.87% validation accuracy in 10 epochs, and 88.17% after 25 epochs.

Reinforcement Learning

Implementation of Q-learning using the OpenAI Gym environment.

Implementation of Deep Q Network for Ms Pacman using OpenAI Gym and Keras.

Transfer Learning

Uses Inception model as a pretrained feature extractor on CIFAR-10.