Current PhD Project - Language Mapping of Brain Vision using Deep Neural Network
Brain vision data is recorded from left and right hemispheres using conventional and tripolar electrodes of size 10mm, 6mmm ,and 4mm. Subjects were shown pictures for 3500 ms each, with 2500 ms inter-stimulus interval during which the computer monitor was grey. There were two conditions: overt and covert. In the overt condition, participants were instructed to verbalize aloud each presented image. In the covert condition, participants silently named the image without verbalizing aloud. Deep Learning techniques are applied on this data to find which part of the brain is giving good result for overt and covert. Also deep learning is used to find out which electrodes are performing best during subjects' recording.
Masters' Project - Implementation of Artificial Neural Network for Multiphase Flow Simulation
Reservoir simulation model contains large number of different data, number of injection/production wells and millions of grid-blocks with complex geological structure. This complexity leads recovery process changes from : 1) natural depletion, 2) followed by water- flooding ,and finally 3) enhanced oil recovery. Numerical models takes large number of simulation runs and time to calculate the final response. Therefore, machine learning techniques such as neural networks were used as surrogate reservoir models to estimate flow characteristics of oil, gas and water in the reservoir in very less amount of time.
Target and Non-Target classification of P300 Event Related Potentials
The P300 wave data is collected from Brain Computer Interface (BCI), Colorado State University (CSU) website. In which the data is collected from unimpaired subjects. Data consists of single letter display on the screen, not the grid data. The conventional EEG was used to record the signals from these channels: 'F3', 'F4', 'C3', 'C4', 'P3', 'P4', 'O1', 'O2' according to 10-20 system. Machine Learning techniques and Convolutional Neural Network (CNN) was used to predict the target and non-target Event Related Potentials.
Predicting Pediatric Bone Age from X-rays Using Convolutional Neural Network
Deep neural networks were used for the prediction of bone age from X-rays.
Hyperlink-Induced Topic Search (HITS) over Wikipedia Articles using Apache Spark
HITS algorithm was designed to find the key pages for specific web communities.
Detecting Phishing Websites using MachineLearning Techniques
Machine Learning was used to differentiate between the legitimate and phishing websites.
Yelp Review Star Rating Prediction
state-of-the-art machine learning algorithms were used to find the useful reviews based on the number of stars that are given to reviews.