Brain Inspired

In this second part of my discussion with Wolfgang (check out the first part), we talk about spiking neural networks in general, principles of brain computation he finds promising for implementing better network models, and we quickly overview some of his recent work on using these principles to build models with biologically plausible learning mechanisms, a spiking network analog of the well-known LSTM recurrent network, and meta-learning using reservoir computing.

• Wolfgang's website.
• Advice To a Young Investigator (has the quote at the beginning of the episode) by Santiago Ramon y Cajal.
• Papers we discuss or mention:
  • Searching for principles of brain computation.
  • Brain Computation: A Computer Science Perspective.
  • Long short-term memory and learning-to-learn in networks of spiking neurons.
  • A solution to the learning dilemma for recurrent networks of spiking neurons.
  • Reservoirs learn to learn.
• Talks that cover some of these topics:
  • Computation in Networks of Neurons in the Brain I.
  • Computation in Networks of Neurons in the Brain II.

In this first part of our conversation (here's the second part), Wolfgang and I discuss the state of theoretical and computational neuroscience, and how experimental results in neuroscience should guide theories and models to understand and explain how brains compute. We also discuss brain-machine interfaces, neuromorphics, and more. In the next part (here), we discuss principles of brain processing to inform and constrain theories of computations, and we briefly talk about some of his most recent work making spiking neural networks that incorporate some of these brain processing principles.

• Wolfgang's website.
• The book Wolfgang recommends:
  • The Brain from Inside Out by György Buzsáki.
• Papers we discuss or mention:
  • Searching for principles of brain computation.
  • Brain Computation: A Computer Science Perspective.
  • Long short-term memory and learning-to-learn in networks of spiking neurons.
  • A solution to the learning dilemma for recurrent networks of spiking neurons.
  • Reservoirs learn to learn.
• Talks that cover some of these topics:
  • Computation in Networks of Neurons in the Brain I.
  • Computation in Networks of Neurons in the Brain II.

Nicole and I discuss how a signature for visual memory can be coded among the same population of neurons known to encode object identity, how the same coding scheme arises in convolutional neural networks trained to identify objects, and how neuroscience and machine learning (reinforcement learning) can join forces to understand how curiosity and novelty drive efficient learning.

• Check out Nicole’s Visual Memory Laboratory website.
• Follow her on twitter: @VisualMemoryLab
• The papers we discuss or mention:
  • Single-exposure visual memory judgments are reflected in inferotemporal cortex.
  • Population response magnitude variation in inferotemporal cortex predicts image memorability.
  • Visual novelty, curiosity, and intrinsic reward in machine learning and the brain.
• The work by Dan Yamins’s group that Nicole mentions:
  • Local Aggregation for Unsupervised Learning of Visual Embeddings

I speak with Tom Griffiths about his “resource-rational framework”, inspired by Herb Simon's bounded rationality and Stuart Russel’s bounded optimality concepts. The resource-rational framework illuminates how the constraints of optimizing our available cognition can help us understand what algorithms our brains use to get things done, and can serve as a bridge between Marr’s computational, algorithmic, and implementation levels of understanding. We also talk cognitive prostheses, artificial general intelligence, consciousness, and more.

• Visit Tom's Computational Cognitive Science Lab.
• Check out his book with Brian Christian, Algorithms To Live By.
• Some of the papers we discuss or mention:
  • Rational Use of Cognitive Resources: Levels of Analysis Between the Computational and the Algorithmic.
  • Resource-rational analysis: understanding human cognition as the optimal use of limited computational resources.
• Data on the mind - the data repository we discussed briefly
• A paper that discusses it: Finding the traces of behavioral and cognitive processes in big data and naturally occurring datasets.

Thomas and I talk about what happens in the brain’s visual system when you see something versus imagine it. He uses generative encoding and decoding models and brain signals like fMRI and EEG to test the nature of mental imagery. We also discuss the huge fMRI dataset of natural images he’s collected to infer models of the entire visual system, how we’ve still not tapped the potential of fMRI, and more.

• Thomas's lab website. 
• Papers we discuss or mention:
  • Resolving Ambiguities of MVPA Using Explicit Models of Representation.
  • Human brain activity during mental imagery exhibits signatures of inference in a hierarchical generative model.

Support the Podcast

Kanaka and I discuss a few different ways she uses recurrent neural networks to understand how brains give rise to behaviors. We talk about her work showing how neural circuits transition from active to passive coping behavior in zebrafish, and how RNNs could be used to understand how we switch tasks in general and how we multi-task. Plus the usual fun speculation, advice, and more.

• Kanaka’s google scholar profile.
• Follow her on twitter: @rajankdr.
• Papers we discuss:
  • Neuronal Dynamics Regulating Brain and Behavioral State Transitions.
  • How to study the neural mechanisms of multiple tasks.
• Gilbert Strang's linear algebra video lectures Kanaka suggested.

Support the Podcast

Jon and I discuss understanding the syntax and semantics of language in our brains. He uses linguistic knowledge at the level of sentence and words, neuro-computational models, and neural data like EEG and fMRI to figure out how we process and understand language while listening to the natural language found in everyday conversations and stories. I also get his take on the current state of natural language processing and other AI advances, and how linguistics, neurolinguistics, and AI can contribute to each other. 

• Jon's Computational Neurolinguistics Lab.
• His personal website.
• The papers we discuss or mention:
  • Hierarchical structure guides rapid linguistic predictions during naturalistic listening.
  • Finding syntax in human encephalography with beam search.

Support the Podcast

Andrew and I discuss his work exploring how various facets of deep networks contribute to their function, i.e. deep network theory. We talk about what he’s learned by studying linear deep networks and asking how depth and initial weights affect learning dynamics, when replay is appropriate (and when it’s not), how semantics develop, and what it all might tell us about deep learning in brains.

Show notes:

• Visit Andrew's website.
• The papers we discuss or mention:
  • Are Efficient Deep Representations Learnable?
  • A theory of memory replay and generalization performance in neural networks.
  • A mathematical theory of semantic development in deep neural networks.
• A good talk: 
  • High-Dimensional Dynamics Of Generalization Errors.

A few recommended texts to dive deeper:

• Introduction To The Theory Of Neural Computation.
• Statistical Mechanics of Learning.
• Theoretical Neuroscience.

Support the Podcast

Jess and I discuss construction using graph neural networks. She makes AI agents that build structures to solve tasks in a simulated blocks and glue world using graph neural networks and deep reinforcement learning. We also discuss her work modeling mental simulation in humans and how it could be implemented in machines, and plenty more.

Show notes:

• Jess’s website.
• Follow her on twitter: @jhamrick
• The papers we discuss or mention:
  • Analogues of mental simulation and imagination in deeplearning.
  • Structured agents for physical construction.
  • Relational inductive biases, deep learning, and graph networks.
    • Build your own graph networks: Open source graph network library.

Kyle and I talk about his work modeling the basal ganglia and its circuitry to control whether we take an action and how we select among alternative actions. We also reflect on his experiences in academia, the larger picture of what it’s like in graduate school and after - at least in a computational neuroscience program - why he left, what he’s doing now, and how it all fits together.

Show notes:

• Kyle’s website.
• Follow him on twitter: @dunovank
• Examples of his work on basal ganglia and decision-making and control:
  • Believer-Skeptic Meets Actor-Critic: Rethinking the Role of Basal Ganglia Pathways during Decision-Making and Reinforcement Learning.
  • Reward-driven changes in striatal pathway competition shape evidence evaluation in decision-making.
  • Errors in Action Timing and Inhibition Facilitate Learning by Tuning Distinct Mechanisms in the Underlying Decision Process.
• Mark Humphries’ article on Medium: Academia is the Alternative Career Path.
• For fun, a bit about the “free will” experiments of Benjamin Libet.