Deep learning model’s capability fascinates me. As part of my academic work or the applied research via Dataperformers, deep neural networks was a useful approaches for solving complex problems related to Video content understanding, Object identification or pattern detection.
Reinforcement Learning and Distributed intelligence:
Reinforcement learning is one of my favorite topics. It represent a very good representation of natural intelligence where systems interact with their environments to learn and improve themselves. In my research, I intend to couple this learning approaches with some distributed intelligence principles where many agents are considered as a sub-systems forming a larger distributed intelligence that could better learn from its environment.
Mehdi Merai on OpenAI Gym: https://gym.openai.com/users/mehdimerai
Deep RNN and Smart energy
With my Ph.D supervisor Jia Yuan Yu, we worked on a patented technology that improves global energy demand prediction. Joint electricity predictor and controller (JEPAC) is a system that allows energy suppliers to better predict its electricity grid activity and then, optimize its energy production, management and distribution. In fact, more the prediction is accurate, less negative its economic and environmental impacts will be. Once JEPAC system is installed in the energy consumer place, it will collect indoor ambient parameters and energy usage, then, thanks to an embed machine learning algorithm it will predict the individual future consumption. This prediction will be recurrently transmitted to the energy supplier as a formatted commit-ment then, in a second time, the same device will try to respect this commitment by adjusting wisely the user appliances and HVAC. As a result, energy supplier will then crowd-source the global energy demand by aggregating highly detailed individual consumption commitments.