MAPS, A Metacognitive Architecture for Improved Learning
Tags: Metacognition Reinforcement Learning Artificial Intelligence Presented at the AAAI 2025 conference (TOM4AI workshop), and submitted to the Reinforcement Learning Conference (RLC) 2025. This study introduces the Metacognitive Architecture for Perceptual and Social Learning (MAPS), which integrates a second-order (metacognitive) network into AI systems to enhance social and continuous learning. Documentation: https://drive.google.com/file/d/1Muyolvj6fD2IwnQXNde1kSpChpXFQja2/view?usp=sharing
Apr 1, 2025
MAPLE, Modular Attention for Interpretable and Prosocial Multi-Agent Reinforcement Learning
Tags: Reinforcement Learning Neuro-AI Multi-Agent Systems Modular Attention Submitted to RLC 2025, and presented to NeurIPS 2023 (Meltingpot Challenge Workshop) and the Workshop of Advances in Neuro AI 2023. MAPLE introduces a novel approach to enhancing interpretability and performance in multi-agent reinforcement learning (MARL) through modular architecture and representation learning. Documentation: https://drive.google.com/file/d/1aEcKU-kzjo8WxM_sjoJr9HGxAzQRVw4g/view?usp=sharing
Feb 1, 2025
Hybrid AI Financial Advisor Chatbot
Tags: Deep learning Risk profiling Financial modeling Hybrid chatbot that uses financial modeling to transform economic and deep learning generated predictions into insights for investing and optimizing a consumer’s financial portfolio (investments, savings, spending, and debt) according to risk profile. The first version of this powers the insights with generative AI, while a second version includes automated text answers that compile deep learning predictions using fixed financial formulas to interpret results. This work was presented in a demo day in pitch format at Mila.
Aug 1, 2024