Machine Learning Street Talk (MLST) #041 - Biologically Plausible Neural Networks - Dr. Simon Stringer
Feb 3, 2021
Dr. Simon Stringer, a Senior Research Fellow at Oxford University, discusses the intricate relationship between brain function and artificial intelligence. He dives into hierarchical feature binding, revealing how biologically inspired neural networks can enhance visual perception. The conversation covers the challenges of replicating human cognitive behaviors using AI and the importance of self-organization and temporal dynamics in learning. Stringer also sheds light on how insights from neuroscience can refine AI models to handle complex tasks more effectively.
AI Snips
Chapters
Transcript
Episode notes
Hierarchical Feature Binding
- Feature binding in the brain is hierarchical, representing relations between features at different scales.
- Traditional rate-coded networks fail to capture this crucial information, hindering true understanding of complex scenes.
Three-Neuron Binding Circuits
- Spiking neural networks incorporate axonal delays, unlike earlier synchronous models.
- Three-neuron binding circuits emerge, encoding specific feature relationships through precise spike timings.
Invariant Representations
- The brain learns invariant representations through unsupervised learning, like continuous transformation and trace learning.
- These mechanisms allow one-shot object recognition by building an alphabet of invariant visual primitives.