Autoplay
Autocomplete
Previous Lecture
Complete and Continue
Refactored: Evolve Beyond a Glorified Curve-Fitter
Introduction
Course overview
Model-based Machine Learning
Overcoming Cognitive Bias with Bespoke Models (7:41)
Generative Probability Models as (Falsifiable) Explanations of Data (6:30)
Myths, Models, Programs, Simulations (8:19)
Building Myths as Causal Bayesian Networks (5:31)
Probabilistic Programming and Deep Probabilistic Programming (24:59)
A Shannon Information & Communication Approach to ML
Shannon's Model of Communication (5:44)
The Map is Not the Territory: Separating the Receiver and Transmitter (6:13)
Transmitter-Receiver Case Study (4:09)
Communicating with Friends, Strangers, and Adversaries
Information Theory for Model Evaluation and Selection
Rethinking Practical ML with Bayes
Cultivating Bayesian-thinking (3:04)
Revisiting Training Machine Learning Algorithms with Bayes
Revisiting Machine Learning In Production with Bayes
Bayesian Risk: Connect Data to Decisions
Bayesian Risk Respects Business Nuance
Thinking Clearly about Inductive Bias
The Problem of Induction (7:57)
Defining Inductive Bias in Machine Learning
Inductive Bias and the Structure of Natural Language
Inductive Bias in Deep Language Models
Inductive Bias in Computer Vision
The Bayesian CogSci Approach to Induction
Conclusion
Wrapping Up
Resources for Further Learning
Generative Probability Models as (Falsifiable) Explanations of Data
Lecture content locked
If you're already enrolled,
you'll need to login
.
Enroll in Course to Unlock