Computational foundations for Neuroscience

Psych 178

Fall 2023, Tor Wager

Assignments: https://canvas.dartmouth.edu/courses/60516

Preliminary background

• Some resources
• Choosing a statistical test
• A menu of topics
• Fundamentals of distance, covariance, and correlation
• Stats in Matlab

General linear model (GLM)

• General Linear Model (GLM)
• Parametric inference in the GLM
• Sample dataset and analysis 1: Ashar 2022
• Parameterizing models in a GLM
• Generalized Least Squares (GLS)
• A hand-coded GLM in MATLAB

Mixed effects

• Mixed effects: Introduction
• Mixed effects: Examples
• Mixed effects algebra
• Mixed effects simulations

Multivariate statistics

• Singular Value Decomposition
• Interpreting SVD of a design matrix and using it for dimensionality reduction

Bayesian models

• Active inference

Reinforcement Learning

• Reinforcement Learning

Natural Language Processing (NLP)

• An introduction to natural language processing in Python
• What’s covered in this tutorial?
• What is natural language processing?
• Early tasks: part of speech tagging
• Early tasks: stemming/lemmatization
• Early tasks: automatic summarization
• But how can we get at the meaning of natural language?
• Text embedding models
• Latent semantic analysis
• Latent Dirichlet Allocation (LDA)
• Word2vec
• Consider the following phrases:
• Bag of words vs. context-sensitive models
• Universal Sentence Encoder
• Transformers
• Transfomers
• Transformers
• Generative Pretrained Transformer (GPT)
• NLP in practice
• NLTK
• Hugging Face
• transformers library (by Hugging Face)
• Interactive agents (ChatBots)
• langchain library
• Summary
• Models of text and language
• Text embedding models
• Text embedding example 1: Latent Dirichlet Allocation
• Text embedding example 2: deep embeddings
• Within-document “dynamics”
• Interactive agents
• Chatbot demo 2: run a chatbot locally!
• Followup things to explore/try

Neural networks

• Recurrent Neural networks

What it is, what it’s not

What this course is intended to do:

• Provide an expandable series of topics and tutorials

• Help you gain some fundamental building blocks that will enhance your self-education

• Take a “back to basics” kind of approach without (hopefully) being too redundant

• Allow flexibility so you can adapt this to your research

• Be a collaboration with you!

What this course is not:

• A finished product

• A complete course in statistics or computation

• A “how to do this fancy thing with this toolbox” kind of course.

The importance of estimating uncertainty

aka: “The importance of knowing that we don’t know”

Statistical inference is a part of virtually all data analyses efforts in neuroscience and psychology. When they are not, they usually should be.

Inference is about establishing how confident we should be that the pattern of effects we see in the data are “real”, i.e., replicable and likely to be observed again in the future.

Whether we’re observing a multivariate pattern in the data, plotting latent trajectories, testing a new alignment algorithm, fitting a learning model, or whatever, we should all be asking ourselves:

• What is the uncertainty on whatever property I’m interested in?

• How can I construct a statistical test to characterize the uncertainty and calibrate my confidence level?

This is why we’re starting with some statistical tests that are the most useful and widely used.

Schedule and links

This book: https://torwager.github.io/ComputationalFoundations

Canvas link for Assignments

OLD Dropbox link to post assignments (don’t use this)

Course Schedule and Timetable: Fall 2023 September 11, 2023 – Fall term classes begin

Week 1: Sept 14

Overview and student feedback, discussion. Explore datasets for final projects General linear model (GLM) Covariates Contrasts Model diagnostics and VIFs Basis functions and nonparametric regression (splines) Robust regression Autocorrelation and time series models No, I do not expect to get through all this!!!

Week 2: Sept 21

Multi-level linear models and mixed effects IGLS/rIGLS Simulating power and false positive rates (bias/variance, precision-recall) Inference with randomization tests (bootstrapping, permutation) Mediation

Week 3: Sept 28

Penalized regression (ridge, LASSO) Cross-validation Multivariate predictive classification and regression models (SVM, LASSO-PCR)

Week 4: Oct 5

No class Read/prep on: Bayesian fundamentals Empirical Bayes Bayes Factors

Week 5: Oct 12

Oct 12: Predictive processing and active inference (Ryan Smith) – all day workshop

Week 6: Oct 19

Reinforcement learning and sequential models (Soltani) Kalman Filter

Week 7: Oct 26

Natural language processing (Manning)

Week 8: Nov 2

Recurrent neural networks (Murray)

Week 9: Nov 9

Present final projects

Week 10: Nov 16

Present final projects November 14, 2023 – Fall term classes end

Final projects

Choose one:

1. Apply a technique to a dataset you are currently working on and hope to/have published on. Present method and findings. Create a notebook that will allow other students to access data (or a subset of it) and run.

2. Create a tutorial on a selected topic not already covered.

Present your final project the last 2 weeks of class (20-30 min per presentation)

Things you don’t need access to (for Tor’s organization): tor’s working doc

tor’s dropbox

Built with Jupyter Book