BIOPOL- Biologically Derived Principles for Online Learning
Abstract
The BIOPOL project aims to derive principles leading to the efficiency and adaptability of the biological brains. Such principles will be essential for replicating online learning mechanisms with machine learning methods. Existing machine learning methods, especially deep neural networks, require immense computational power and energy, which both leads to unsustainable scaling and makes them unsuitable for resource-constrained devices. In contrast, biological systems like the human brain operate with minimal power while solving complex tasks. BIOPOL proposes to explore neuro-inspired computational frameworks that mimic these efficient learning processes, reducing computing costs while maintaining advanced capabilities. By integrating biological insights with computational principles, the project seeks to create sustainable, more computationally efficient machine learning methods.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 06, 2025
- Source ID
- FA86552517007
Entities
People
- Denis Kleyko
Organizations
- Air Force Office of Scientific Research
- United States Air Force