Theoretical Neuroscience Podcast

Today’s AI is largely based on supervised learning of neural networks using the backpropagation-of-error synaptic learning rule. This learning rule relies on differentiation of continuous activation functions and is thus not directly applicable to spiking neurons. Today’s guest has developed the algorithm SuperSpike to address the problem. He has also recently developed a biologically more plausible learning rule based on self-supervised learning. We talk about both.  

Over the last ten years or so, the MindScope project at the Allen Institute in Seattle has pursued an industrylab-like approach to study the mouse visual cortex in unprecedented detail using electrophysiology, optophysiology, optical imaging and electron microscopy.  Together with collaborators at Allen, today’s guest has worked to integrate of these data into large-scale neural network, and in the podcast he talks about their ambitious endeavor.

Delving into the origins of computational neuroscience and AI, the podcast explores the transition from rule-based to learning-based AI approaches. It highlights the unreasonable effectiveness of math in deep learning and the evolution of reinforcement learning in neural structures. The synergy of AI and neuroscience in medical diagnostics, advancements in self-driving technology, and the transformative impact of AI on society are also discussed.

Challenges in traditional neuroscience methods, focus on reverse engineering C.elegans, parallels between transistors and neurons, pitfalls of statistical analysis in biology, mechanistic understanding in neuroscience, neural complexity of C.elegans, error recalibration in neural modeling, activation functions in machine learning, optimization challenges in bio-physical models, variability in neural networks

Over the last decade topological analysis has been established as a new tool for analysis of spiking data. Today’s guest has been a pioneer in adapting this mathematical technique for use in our field and explains concepts and example applications.  We also also talk about so-called threshold-linear network model, a generalization of Hopfield networks exhibiting a much richer dynamics, where Carina has done some exciting mathematical explorations

Not all interesting network activity occurs in cortex. Networks in the spinal cord, the long thin tubular structure extending downwards from the neck, is responsible for setting up rhythmic motor activity needed for moving around. How do these so-called central pattern generators work? Today’s guest has, together with colleagues in Copenhagen, developed a neuron-based network theory for how these rhythmic oscillations may arise even without pace-maker neurons driving the collective. 

The guest, Li Zhaoping, discusses the theory of V1 as a 'saliency detector' in vision processing, directing gaze to important objects. The podcast explores the progression of feature detectors in visual processing, the limitations of human vision, and the role of attention selection in humans and animals. It also delves into the connection between visual movement in birds and mice and sensory systems, and interdisciplinary advancements in visual neuroscience.

In this engaging conversation, Sacha van Albada, a leading expert in theoretical neuroanatomy, shares insights on developing advanced multi-area cortex models for macaques and humans. He discusses the challenges of linking single-neuron activity to larger systems, the intricacies of simulating neural dynamics on supercomputers, and the importance of physiological data for accuracy. Sacha also highlights the significance of collaboration in improving models and explores concepts like predictive coding, making complex neuroscience more accessible for researchers.

Arvind Kumar, a guest who has thought about coding questions throughout his career, discusses the neural code and its interpretations. Topics include neural representation of visual features, decoding motor systems with spike count codes and tuning curves, population coding and correlation, neural code and connectivity, diversity in tuning curves and encoding information, comparing visual neurons and entorhinal cortex neurons, catastrophic errors in AI and grid cells, collaboration to solve a problem involving time and stimuli, adversarial attacks and grid cells' representation of navigation, and efficient coding and sparse coding in the brain.

Christof Koch, author of 'Biophysics of Computation', discusses the intersection of physics, neuroscience, and philosophy. He explores homeostasis in neurons, locust movement detection, brain mapping to logical operations, degeneracy and redundancy in biological systems, and the limitations of the mouse v1 model. Koch also introduces the concept of integrated information theory of consciousness and discusses challenges in evaluating complex systems.