This is me - Subhodeep Moitra

    I am a computer science researcher and AI specialist with expertise in artificial intelligence, machine learning, deep learning, computer vision, adversarial robustness, and cognitive science. My research focus on developing advanced AI frameworks that integrate human-like visual perception with brain-inspired cognitive models to create intelligent systems capable of human-level reasoning and adaptability. My research bridges the gap between sensory processing and higher-order cognitive functions, emphasizing resilient computer vision systems and robust cognitive modeling. Committed to pushing the boundaries of AI innovation, my aim is to design artificial intelligence solutions that allow machines to learn from real-world data, interpret complex nuances, and make decisions with the depth, flexibility, and contextual understanding of the human brain. My work contributes to the future of AI by advancing unified frameworks for machine perception, cognition, and ...

This is kind of personal but here I am going to share some experiences that I got from training different kinds of Neural Networks (NNs). Also I will not be very technical in this post, but just a kind of overview that might give you some insights or get to you some ideas regarding deep fundamental research. This is kind of personal, so many might disagree with me. So first of all I am a AI Researcher working at the fundamentals of Deep Neural Networks (DNNs). While training DNNs for my research on various domain data, I many a times faced many situations that many a times questions about how we understand a DNN or in general NN. I am not saying that what we know is wrong or has some flaws, but if we try to look into the training fundamentals through the various epochs and the general training behavior then many a times we might see strange things that are many a times not properly understandable and we define those instances with terms like “overfitting” or “underfitting”...