Algorithmic Research

Domain: Machine learning, Deep learning.

Reinforcement Learning is the science of learning what to do and how to do it by introducing a reward-penalty system. The reward function is applied whenever an ‘agent’ wins the ‘game’ over a certain number of rounds. The agent is a program that processes a sequence of decisions in an unexplored environment to achieve some goals. With trial and error, the agent’s objective is to learn to make decisions that will result in a maximum reward.

Silicon Orchard Research and Analytics team has extended its research on Reinforcement Learning to devise an intelligent way to capture human interactions with the Internet. The team introduces a novel approach to utilize reinforcement learning algorithms that understand human preferences and can devise recommendation systems without the need of collecting any identifiable user data.

We are proud to present our research Leveraging Reinforcement Learning to build a Recommendation System for Incognito mode Users has been published in the ICML 2021 Workshop on Representation Learning for Finance and e-Commerce Applications, part of the prestigious A* conference, International Conference on Machine Learning (ICML).

Abstract: Code review is one of the crucial steps in the software development process. Despite having many experts, assigning the appropriate one is often challenging, time-consuming, and inefficient for industrial developers and researchers who demand instant solutions. An automated code review system can serve as a proficient and alternative opportunity for those necessities. This paper aims to identify appropriate reviewers for a selected task based on data analysis using Natural Language Processing (NLP) techniques. Appropriate Developer for Code Review (ADCR) is proposed taking into account a set of data that comprises reviewers’ information-responsiveness, experience , and acquaintanceship-benefits of the proposed methods including unbiased review accountability and the early feedback opportunity for the developers. Additionally, a tool is developed to process the automated review and speed up the development cycles.