Unsupervised Thinking
a podcast about neuroscience, artificial intelligence and science more broadly

Sunday, October 28, 2018

Episode 38: Reinforcement Learning - Biological and Artificial

Reinforcement learning is important for understanding behavior because it tells us how actions are guided by reward. But the topic also has a broader significance---as an example of the happy marriage that can come from blending computer science, psychology and neuroscience. In this way, RL is a poster child for what's known as Marr's levels analysis, an approach to understanding computation that essentially asks why, how, and where. On this episode we first define some of the basic terms of reinforcement learning (action, state, environment, policy, value). Then we break it down according to Marr's three levels: what is the goal of RL? How can we (or an artificial intelligence) learn better behavior through rewards? and where in the brain is this carried out? Also we get into the relationship between reinforcement learning and evolution, discuss what counts as a reward, and try to improvise some relatable examples involving cake, cigarettes, chess, and tomatoes.

We read:
Reinforcement Learning with Marr
Reinforcement learning: Computational theory and biological mechanisms

And mentioned:
Understanding dopamine and reinforcement learning: The dopamine reward prediction error hypothesis - by Paul Glimcher
Training a glider with RL

To listen to (or download) this episode, (right) click here or use the player below

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Tuesday, September 25, 2018

Episode 37: What is an Explanation? - Part 2

In part two of our conversation on what counts as an explanation in science, we pickup with special guest David Barack giving his thoughts on the "model–mechanism–mapping" criteria for explanation. This leads us into a lengthy discussion on explanatory versus phenomenological (or "descriptive") models. We ask if there truly is a distinction between these model classes or if a sufficiently good description will end up being explanatory. We illustrate these points with examples such as the Nernst equation, the Hodgkin-Huxley model of the action potential, and multiple uses of Difference of Gaussians in neuroscience. Throughout, we ask such burning questions as: can a model be explanatory if the people who made it thought it wasn't? are diagrams explanations? and, is gravity descriptive or mechanistic?

Listen to Part One first!

We read:
Explanation and description in computational neuroscience
(which was a recommendation from this Twitter thread on the philosophy of mathematical modeling)

And David mentioned: Craver's Explaining the Brain

To listen to (or download) this episode, (right) click here

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Wednesday, August 29, 2018

Episode 36: What is an Explanation? - Part 1

As scientists, we throw around words like "explanation" a lot. We assume explaining stuff is part of what we're doing when we make and synthesize discoveries. But what does it actually take for something to be an explanation? Can a theory or model be successful without truly being one? How do these questions play out in computational neuroscience specifically? We bring in philosopher-neuroscientist David Barack to tackle this big topic. In part one of the conversation, David describes the historical trajectory of the concept of "explanation" in philosophy. We then take some time to try to define computational neuroscience, and discuss "computational chauvinism": the (extremist) view that the mind could be understood and explained independently of the brain. We end this first half of the conversation by defining the "3M" model of explanation and giving our initial reactions to it.

We read:
Explanation and description in computational neuroscience
(which was a recommendation from this Twitter thread on the philosophy of mathematical modeling)

Previous episodes that may be of interest:
Episode 8: Neuroscience vs. Psychology
Episode 13: The Unreasonable Effectiveness of Mathematics
Episode 21: Understanding fMRI

To listen to (or download) this episode, (right) click here

UPDATE: Part 2 is now posted here

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Tuesday, July 31, 2018

Episode 35: Generative Models

Machine learning has been making big strides in a lot of straightforward tasks, such as taking an image and labeling the objects in it. But what if you want an algorithm that can, for example, generate an image of an object? That's a much vaguer and more difficult request. And it's where generative models come in! We discuss the motivation for making generative models (in addition to making cool images) and how they help us understand the core components of our data. We also get into the specific types of generative models and how they can be trained to create images, text, sound and more. We then move onto the practical concerns that would arise in a world with good generative models: fake videos of politicians, AI assistants making our phone calls, and  computer-generated novels. Finally, we connect these ideas to neuroscience, asking both how can neuroscientists make use of these and is the brain a generative model?

We read:
OpenAI blog
MIT Tech Review Article 
(accidentally referred to as Wired article...oops!)

And skimmed/mentioned:
Episode 4 - Deep Learning
Another good overview of some generative models
Google Assistant making phone call
Uses of generative models in neuroscience 
Episode 33 - Predictive Coding

And our special guest was Yann Sweeney!

To listen to (or download) this episode, (right) click here

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Thursday, June 28, 2018

Episode 34: The Gut-Brain Connection

Because of the sheer number of neurons in the gut, the enteric nervous system is sometimes called the second brain. What're all those neurons doing down there? And what, or who, is controlling them? Science has recently revealed that the incredibly large population of microorganisms in the gut have a lot to say to the brain, by acting on these neurons and other mechanisms, and can impact everything from stress to obesity to autism. In this episode, we give the basic stats and facts about the enteric nervous system (and argue about whether it really is a "second brain") and cover how the gut can alter the brain via nerves, hormones, and the immune system. We then talk about what happens when mice are raised without gut microbes (weird) and whether yogurt has any chance of curing things like anxiety. Throughout, we marvel at how intuitive all this seems despite being incredibly difficult to actually study. All that plus: obscure literary references, Josh's hilariously extreme fear of snakes, multiple misuses of the word "species," and DIY feces transplants! 

We read:
Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour

And mentioned:
Nancy's previous episode Episode 31: Consuming Science
Episode 23: What Can Neuroscience Say About Consciousness?

And our special guest was Nancy Padilla!

To listen to (or download) this episode, (right) click here
[apologies for minor audio issues on this one!]

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Wednesday, May 30, 2018

Episode 33: Predictive Coding

You may have heard of predictive coding; it's a theory that gets around. In fact, it's been used to understand everything from the retina to consciousness. So, before we get into the details, we start this episode by describing our impressions of predictive coding. Where have we encountered it? Has it influenced our work? Why do philosophers like it? And, finally, what does it actually mean? Eventually we settle on a two-tiered definition: "hard" predictive coding refers to a very specific hypothesis about how the brain calculates errors, and "soft" predictive coding refers to the general idea that the brain predicts things. We then get into how predictive coding relates to other theories, like Bayesian modeling. But like Bayesian models, which we've covered on a previous episode, predictive coding is prone to "just-so" stories. So we discuss what concrete predictions predictive coding can make, and whether the data supports them. Finally, Grace tries to describe the free energy principle, which extends predictive coding into a grand unified theory of the brain and beyond.

We read:
Whatever next? Predictive brains, situated agents, and the future of cognitive science

And mentioned:
Episode 15: "Just-so" stories in Bayesian modeling in psychology
Kok and Lange book chapter on predictive coding
NYU panel/debate on predictive coding

And our special guest was Alex Cayco-Gajic!

To listen to (or download) this episode, (right) click here

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com) 

Tuesday, May 1, 2018

Episode 32: How Do We Study Behavior?

There is a tension when it comes to the study of behavior in neuroscience. On the one hand, we would love to understand animals as they behave in the wild---with the full complexity of the stimuli they take in and the actions they emit. On the other hand, such complexity is almost antithetical to the scientific endeavor, where control over inputs and precise measurement of outputs is required. Throw in the constraints that come when trying to record from and manipulate neurons and you've got a real mess. In this episode, we discuss these tensions and the modern attempts to resolve them.

First, we take the example of decision-making in rodents to showcase what behavior looks like in neuroscience experiments (and how strangely we use the term "decision-making"). In these studies, using more natural stimuli can help with training and lead to better neural responses. But does going natural make the analysis of the data more difficult? We then talk about how machine learning can be used to automate the analysis of behavior, and potentially remove harmful human biases. Throughout, we provide multiple definitions of "behavior", Grace relates animal training to parenting, and our special guest Adam Calhoun uses his encyclopedic knowledge of this area to provide many insightful examples!

We read:
Decision making behaviors: weighing ethology complexity and sensorimotor compatibility
Computational Analysis of Behavior - Annual Review of Neuroscience [$]

And mentioned:
Episode 18: Does Neuroscience Need More Behavior?
Mala Murthy (fly courtship work)
Low dimensionality of C. elegans (worm) behavior
How sensory neural responses are heavily influenced by behavior

For more, check out this list of behavioral papers Adam made on Twitter!

To listen to (or download) this episode, (right) click here.

As always, our jazzy theme music "Quirky Dog" is courtesy of Kevin MacLeod (incompetech.com)