Research Projects
Here I’ve listed the projects I had the chance to be part of, with a brief description about each and a link to a webpage that contains more details, related publications and code, as well as some pointers to the related work in the literature.
Machine Learning in Computational Chemistry (2018-present)
Discovering new molecules has been a major challenge in chemical and pharmaceutical applications. while computational methods have been widely used to speed up this process, we focus on developing machine learning approaches to speed this process even further up and save both time and money in the process of discovering and synthesising new materials.
Machine Learning Algorithm in Supply-Demand Management and Pricing in Logistics (2017-present)
In two-sided marketplaces, such as the one in logistics where on one side we have shippers and on the other we have drivers, efficiently matching supply to demand and accurately predict the price of shipping goods is very important to increase the fulfilment rate of orders as well as maximizing the “welfare” of both sides. This is what we have been working on as a part of our business, Ubaar.
Neural Network Connectome Mapping (2com014-Present)
Our goal is to identify (effective) synaptic connections in a network of neurons only by observing spike times of neurons; We obtain theoretical performance guarantees in presence of hidden neurons. Extensions to cases with spike erasure and jitter are also considered.
Exponential-capacity Associative Memory (2010-2014)
We exploit redundancy in data to increase the pattern retrieval capacity of neural associative memories (especially from linear to exponential if patterns form subspaces). The effect of noisy neurons is also considered on the perfomance of the algorithms.
Information and Coding Theory (2009-2014)
The project was aimed at studying and improving performance of error correcting codes. At the later stages, we used ideas from noisy neuronal networks to improve short-length performance of some codes on graphs.
Bio-inspired Networking (2007-2009)
Bio-inspired algorithms can be of a great help and inspiration in communication systems. Here we focused on using swarm intelligence, especially ant colony optimization methods, to design rate allocation methods for content distribution over P2P networks.
Quality of Service Network Coding (2006-2009)
It has been shown that network coding can be used to improve achievable rates in some communication channels. Given that content has become ever more important in communication systems, assuring QoS has become a priority and in this project we try to develop network coding algorithms with integrated QoS assurance schemes.