What Should a Computational Theory of Cortex Explain?
summary
Argues a concrete cortical model with three resource parameters (n, d, K) and local neuron programs, formalizes operations: hierarchical memorization, association, and inductive learning, then stress-tests them in large simulations. With n ≈ 10^8, d ≈ 8,000 and K = 16 a weak-synapse regime needs about 360,000 neurons per item and yields ~3,000 stable associations; a strong-Boolean regime using ~100 neurons per concept supports on the order of 250,000 successive operations. Proposes hippocampus as a stable memory-allocation mechanism.
AI-written summary · about
chapters (8)
- 0:00Intro
- 1:36Nobel Prize Winner
- 2:27Diseases of the Will
- 3:59Computation
- 6:08David Marr
- 44:48Summary
- 48:04Hippocampus
- 48:35Memory Allocation
related talks
47mCapacity of Neural Networks for Lifelong Learning of Composable Tasks
56mDoes the Neocortex Use Grid Cell-Like Mechanisms to Learn the Structure of Objects?
01h 03mPlenary Discussion: What Is Missing in Current Theories of Brain Computation?
01h 14mTissue vs silicone musings on the future of deep learning hardware and software
43mCan Cognitive Neuroscience Provide a Theory of Deep Learning...
41mBrain-inspired computational primitives for reliable neuromorphic intelligence
01h 13mComputation in Networks of Neurons in the Brain II
58mStrong Generalization from Small Brains and No Training Data