This repository contains my implementations of Deep Neural Network based architectures for specific usecases using PyTorch. Each implementation comes with a detailed problem specification and a link to the solution.

• Given the sp500 stock market tickers dataset
• Modelling the time series using LSTM, and perform various experiments involving techniques like Normalization, Feature Engineering
• Assessing the profitability of the algorithmic trading module under various conditions like buy-ask spread, commissions .etc
• Detailed problem specification can be found here
• My solution for the same can be found here

• Given the Labeled Faces in the Wild dataset
• Using a Transfer Learned ResNet based Siamese Network for Similarity Metric Learning for Identification
• Performing experiments using techniques like Regularization, Learning Rate Scheduling, Dropout, Variation of Optimizers to identify best performing model
• Training a DCGAN to generate new faces from input gaussian noise
• Modifying the DCGAN to become a Conditional GAN, by using the Siamese Network trained earlier, i.e, given an input image generate an unseen image of the same person
• Detailed problem specification can be found here
• My solution for the same can be found here

Input Image	Conditionally Generated Image