Ankit K. Singh

Projects And Repos

Abductive Reasoning to Interpret EEG Dataset

Abductive Reasoning model to interpret the EEG time series data for Epileptic Seizure Onset detection. Implemented framework and different hypothesis grammar for extracting explanations for classification.

Answer selection using Text-GCN and attention mechanism

Designed an architecture using Graph Convolutional Network for Answer selection task. Implemented above framework using TensorFlow and attained and performance improvement of 2%(accuracy).

Improvement on Perturbative Neural Network

Implemented Perturbative Neural Network architecture using TensorFlow and proposed tweaks to improve model performance. Compared the model with baselines for Image Classification Task on CIFAR dataset.

Parallel Accelerated Gradient Descent

Implemented serial and parallel version of GD, Heavy Ball, Nesterov and FISTA algorithm using CUDA framework. Analyzed the parallel code performance wrt. the serial implementation.

Technothlon App

Made a scalable multi-platform app using Cordova for Technothlon (All India competition taken by around 70000 students annually). Included features viz. Infinite Scroll, Dynamic Notifications etc.

Smart Health monitoring System

Built a smart health monitoring system using Arduino(MCU) and related sensors performance. Used django based central server hosted on Raspberry-Pi(MPU) with wifi adapter and Xbee for communication.