Ramashish Gaurav

rgaurav@vt.edu

Degree Objective: Ph.D.

Research Interests:

Spiking Networks
Neuromorphic Computing
Computational Neuroscience

Education:

Dual Degree (B.Tech. + M.Tech), CSE, Indian Institute of Technology - BHU, Varanasi (2012-17)
Master of Applied Science, SYDE, University of Waterloo (2020-2021)

Research Experiences:

Neuroscience and Neuroimaging Research Experience

Resting-State Functional Connectivity analysis of Autistic Individuals

This project involves a study of the alterations in the resting-state functional connectivity of autistic patients. Data-driven approaches (first level and group level GLM) are applied on ABIDE-I and ABIDE-II data-sets to discover functionally altered links along with consolidating, reproducing, and validating existing results.

Neuroscience and Neuroimaging related courses

Advanced Functional Brain Imaging: A course at IIT-Delhi which teaches basic neuroanatomy, MRI physics, fMRI processing, related statistical concepts, GLM, ISC, and MVPA analysis. [Assignments] ◦
Coursera courses: Computational Neuroscience, Principles of fMRI – part 1, Principles of fMRI – part 2

Machine Learning Research Experience

Estimation of train delays at railway stations in India

A delay prediction algorithm inspired from N-Order Markov Processes was formulated which leveraged Random Forest Regressors and Ridge Regressors to predict delays at in-line stations

Algorithms for Subspace Learning

The thesis involved developing algorithms for learning latent subspaces from visual features of images for image classification. Two different problem settings were addressed, briefed in following sub-projects.

Traditional image classification with training and test images drawn from the same database
- A novel algorithm was developed for achieving early fusion of information (modals) via supervised Matrix Factorization which added intelligence to the obtained latent subspace from all modals
A novel image classification challenge where training and test images’ classes are disjoint
- Novel approaches to transfer knowledge from training classes to zero-shot test classes via high-level features were developed which achieved state-of-the-art results and outperformed few existing ones

Content-based image classification via multi-modal fusion of visual features

A Matrix Factorization based framework for multi-modal fusion of n different modals of image data-sets was designed where a latent subspace was learned with the help of simple gradient descent additive update rules.

Work Experiences:

Nutanix Technologies, Bangalore, India (2017-2019): Member of Technical Staff - 3 (Details in linked CV)

Publications:

Gaurav, R., & Srivastava, B. (2018, November). Estimating train delays in a large rail network using a zero shot Markov model. In 2018 21st International Conference on Intelligent Transportation Systems (ITSC) (pp. 1221-1226). IEEE.
Gaurav, R., Verma, M., & Shukla, K. K. (2016, December). Informed multimodal latent subspace learning via supervised matrix factorization. In Proceedings of the Tenth Indian Conference on Computer Vision, Graphics and Image Processing (pp. 1-8).
Gaurav, R., Vallecha, A., Verma, M., & Shukla, K. K. (2015, December). Multimodal subspace learning on Flickr images. In 2015 IEEE UP Section Conference on Electrical Computer and Electronics (UPCON) (pp. 1-6). IEEE.