Metropolis Monti Carlo Simulation of 2D Hydrophobic-Polar model of Proteins

(6th March, 2018 to 19th March 2018)

2D HP-model is a simple 2 dimensional lattice structure of proteins containing only two type of residues; Hydrophobic, and polar. The goal of the project was to simulate the conversion of a HP-structure of a protein to a stable form with minimum possible energy configuration. The algorithm used is named as Monti-Carlo Simulation which accepts wrong moves with some probability.

WatchYourLanguage

Final Year Project (Aug 2017 – Present)

It is an intelligent text reader, a web application that takes the input written-text and converts to its spoken form through the technique of English-text Normalization using Recurrent Neural Networks and Long Short Term Memory Neural Networks.

Text-Independent Speaker Verification

Deep Learning Cource Project (Nov 2017 – Dec 2017)

We have used CNN model to get speaker’s features that are more robust. In 1st phase of project we have built an underlying model of CNN to classify speaker at the utterance level and in 2nd phase we used the trained model to get speaker’s features the extracted features from the test set will be compared to stored speakers features to verify the claimed identity of a speaker. We used KING and NKING data set for our speaker verification model.

Simulated Car Racing Competition

Artificial Inteligence Cource Project (Oct 2017 – Nov 2017)

We designed a controller of car for racing with other cars on different tracks and wins the race. The usual performance metrics like speed, obstacle avoidance and track following were kept in mind while designing this controller. We kept this problem simple by disabling noisy sensor and car damage, and by supplying unlimited amount of fuel. The tool that we used was a TORCS framework and we were provided the complete architecture and static actuators that we had to predict later on using Random Forests on the sensor values that we get from the environment.

Supply-Demand Gap Prediction

Data Mining Cource Project (Apr 2017 – May 2017)

Supply-demand forecasting is critical to enabling Uber to maximize utilization of drivers and ensure that riders can always get a car whenever and wherever they may need a ride. Supply-demand forecasting helps to predict the volume of drivers and riders at a certain time period in a specific geographic area. The end result is higher earnings for drivers and no surge pricing for riders! The prediction of such supply-demand gap was done by using the concepts of Data Science and Machine Learning, and built in libraries of Python such as scikit-learn and tensorflow.