Stock Trading System

Performance of Trading system on test data

I set up an LSTM-based stock trading system for high-frequency trading and modified it to use multiple stock prices and day of the week as inputs to predict a single stock using this Dataset

Generative AI

I implemented a Siamese Network using a metric learning scheme (Cosine Similarity; paired with Crossentropy) on labeled faces in the wild (LFW) dataset. Following this, I proceeded to train a generative model for creating facial images, initially employing a Generative Adversarial Network (GAN).

Subsequently, I further advanced this model to function as a conditional GAN (CGAN), wherein the condition for image generation relies on a real face image of a specific individual. To achieve this, I incorporated a Siamese Discriminator, which was initially implemented in the primary task. This enhanced CGAN successfully generates additional images portraying the same individual based on these specified conditions.