July 13-31, 2020
Objectives: Introduce traditional and emerging computational neuroscience tools, their complementarity, and what they can tell us about the brain. A main focus is on modeling choices, model creation, model evaluation and understanding how they relate to biological questions.
Tutorial microstructure: ~10min talk, ~20min tutorial (repeated)
Day structure: Opening keynote, 3h lecture/tutorial modules, 1h interpretation (what did we learn today, what does it mean, underlying philosophy, 1h professional development/ meta-science, evening group projects (for interactive track). There will also be many networking activities!
Prerequisites: See here
Description Introduction of datasets (spikes, EEG, fMRI + behavior), and questions about them. These questions will foreshadow the whole summer school.
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | NMA organization, expectations, code of conduct, modeling vs. data |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Data intro, preprocessing |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Link of neural data to behavior |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Tuning (RFs, motor, STA) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | What it means to "understand" (signal detection) |
| 4:35 - 5:30 | Recap, Q&A | Outlook on school |
| 5:30 - 6:00 | Professional development | Being a good NMA participant |
Description Introduce different example model types (Marr 1-3, what/how/why) and the kinds of questions they can answer. MRealize how different models map onto different datasets.
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Model classifications |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Marr 1 |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Marr 2-3 |
| 2:10 - 2:45 | Lecture & Tutorial 3 | "What" |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | "How"/"Why" |
| 4:35 - 5:30 | Recap, Q & A | The role of models in discovery |
| 5:30 - 6:00 | Professional development | How-to-model guide 1 |
Description Fit models to data, quantify uncertainty, compare models
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Why and how to fit models |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Fit a model 1 (linear regression) |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Get error bars |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Compare models, cross-validation, hyperparameters |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Fit a model 2 (nonlinear models) |
| 4:35 - 5:30 | Recap, Q & A | Critical evaluation of model fitting |
| 5:30 - 6:00 | Professional development | How-to-model guide 2 |
Description Introduction to machine learning. The commonly used approaches, how to avoid false positives, how to do it well
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | We want to predict (scikit learn) |
| 0:50 - 1:25 | Lecture & Tutorial 1 | GLMs (temporal filtering models) |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Linear classifier (SVM) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Regularization (L1, L2) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Shallow nonlinear classifier (SVM with RBF kernel) |
| 4:35 - 5:30 | Recap, Q & A | Promises and pitfalls of ML |
| 5:30 - 6:00 | Professional development | How-to-model guide 3 |
Description Concept of dimensionality reduction, ways of doing it, what it means
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Manifolds to understand |
| 0:50 - 1:25 | Lecture & Tutorial 1 | PCA 1 |
| 1:30 - 2:05 | Lecture & Tutorial 2 | PCA 2 (+CCA/clustering) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Signal vs. Noise Manifolds |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Visualizing high-D nonlinear manifolds (e.g. t-SNE) |
| 4:35 - 5:30 | Recap, Q & A | The link between high-dimensional brain signals and low-dimensional behavior |
| 5:30 - 6:00 | Professional development | Efficient collaborations |
Description Bayesian statistics, modeling of behavior, modeling of neural data, quantifying information
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Uncertainty |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Bayes rule I (product rule: cue combination) |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Bayes rule II (Marginalization and nuisance variables) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Causal inference & structural models (use as example for marginalization) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Bayesian decision theory |
| 4:35 - 5:30 | Recap, Q & A | Advanced Bayesian methods |
| 5:30 - 6:00 | Professional development | Productivity tools for science |
Description How to make estimates over time, how the brain does it
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | World has time |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Linear systems theory I (ND deterministic) |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Linear systems theory II (1D stochastic = OU process; ND stocastic = AR(1)) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Markov process |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | State space model |
| 4:35 - 5:30 | Recap, Q & A | Linear systems rule the world |
| 5:30 - 6:00 | Professional development | Open source ecosystem, data management & sharing |
Description How we can make decisions when information comes in over time
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | We need to decide stuff |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Information theory |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Sequential Probability Ratio Test (SPRT) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Hidden Markov Model inference (DDM) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Kalman filter |
| 4:35 - 5:30 | Recap, Q & A | Decisions, decisions, decisions ... |
| 5:30 - 6:00 | Professional development | Open science (general), replicability & reproducibility |
Description We need to move gain info and reach goals
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | We want to control our actions... |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Expected utility / Cost |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Markov decision process (MDP) |
| 2:10 - 2:45 | Lecture & Tutorial 3 | LQG control (MDP for linear systems) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Motor control (signal-dependent noise, time cost, ...) |
| 4:35 - 5:30 | Recap, Q & A | Advanced motor control |
| 5:30 - 6:00 | Professional development | Networking at Conferences |
Description The setting of reinforcement learning and how it approximates the real world, behavior, and potential brain implementations
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Problem formulations: actor-critic |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Critic: reward prediction error |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Exploration (POMDP) vs Exploitation |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Model-based vs model-free RL |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Multi-arm bandits: foraging |
| 4:35 - 5:30 | Recap, Q & A | RL and the brain |
| 5:30 - 6:00 | Professional development | Writing Papers & Grants |
Description The things neurons are made of, channels, morphologies, neuromodulators, and plasticity
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Real neurons ftw |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Channels, HH |
| 1:30 - 2:05 | Lecture & Tutorial 2 | LIF neuron |
| 2:10 - 2:45 | Lecture & Tutorial 3 | LNP (loses fine timing info) |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Hebbian plasticity & STDP |
| 4:35 - 5:30 | Recap, Q & A | A variety of neuron models |
| 5:30 - 6:00 | Professional development | How to find a postdoc |
Description How single neurons create population dynamics
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Mechanistic models of different types of brain actvivity. |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Spikes to rates. |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Wilson-Cowen model (coarse-grained), oscillations & synchrony |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Attractors & local linearization around fixed points |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Chaos in rate networks (stimulus dependent chaotic attractor) |
| 4:35 - 5:30 | Recap, Q & A | A theory of the whole brain |
| 5:30 - 6:00 | Professional development | Early career panel - academia (how to advance through career steps) |
Description Ways of discovering causal relations, ways of estimating networks, what we can do with networks
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | Causality - different views |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Pitfalls of Granger Caausality |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Centrality |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Instrumental variables |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Interventions |
| 4:35 - 5:30 | Recap, Q & A | Latters of causality |
| 5:30 - 6:00 | Professional development | Computational neuroscience in industry - career panel |
Description The concept of ANNs, how to train them,what they are made out of, convnets, and how to fit them to brains
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | DL = crucial tool |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Pytorch intro & model components |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Training it & inductive bias |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Convolutional Neural Network |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Fit to brain (RSA - represenatational similarity analysis) |
| 4:35 - 5:30 | Recap, Q & A | Digging deep |
| 5:30 - 6:00 | Professional development | Job fair (FRL) |
Description Deep learning in more advanced settings. Autoencoders for structure discovery, RNNs, and fitting them to brains
| Time (Hour) | Lecture | Details |
|---|---|---|
| 0:00 - 0:50 | Intro / keynote & tutorial setup | DL for structure |
| 0:50 - 1:25 | Lecture & Tutorial 1 | Autoencoders |
| 1:30 - 2:05 | Lecture & Tutorial 2 | Recurrent Neural Network |
| 2:10 - 2:45 | Lecture & Tutorial 3 | Transfer learning / generalization |
| 2:45 - 3:30 | Break | 45 minutes break |
| 3:30 - 4:05 | Lecture & Tutorial 4 | Causality |
| 4:35 - 5:30 | Recap, Q & A | Digging deeper |
| 5:30 - 6:00 | Professional development | NMA wrap-up |
- Meet a prof about your group's project
- Meet a prof about your career
- Meet a prof about your own project
- Meet other participants interested in similar topics
- Meet a group of likeminded people
- Meet people that are local to you (same city, country)
TBA