Skip to content
lectern.tv
Editorial path·AI & ML·graduate·2 institutions

Deep Representation Learning & Approximation Theory

Neural approximation theorems, optimization dynamics, and relational query engines

Pairs MIT 6.7960 Deep Learning (Isola & Beery) with CMU 15-799 Query Optimization (Pavlo) to contrast learned continuous representations with algebraic cost-based execution engines.

Start Step 15 steps · 6h total
Lecture 01: Introduction to Deep Learning (6.7960)
Step 1 of 5

Curriculum sequence

5
Step 1·MIT·Sara Beery·6.7960· 1:00:51

Lecture 01: Introduction to Deep Learning (6.7960)

Frames representation learning as hierarchical composition of parameterized differentiable transformations.

Step 2·MIT·Sara Beery·6.7960· 1:19:34

Lecture 02: How to Train a Neural Net (6.7960)

Analyzes stochastic gradient descent, loss landscapes, and adaptive preconditioning in deep networks.

Step 3·MIT·Jeremy Bernstein·6.7960· 1:22:42

Lecture 03: Approximation Theory (6.7960)

Proves universal approximation properties and contrasts width versus depth expressivity.

Step 4·Carnegie Mellon University·Andy Pavlo·15-799· 53:57

Lecture 1: SQL Optimization Background (15-799)

Establishes declarative query optimization as search over equivalent relational algebra plans.

Step 5·Carnegie Mellon University·Andy Pavlo·15-799· 1:10:40

Lecture 2: History of Query Optimizers ft. IBM System R (15-799)

Dissects IBM System R's bottom-up dynamic programming optimizer and Selinger selectivity estimation.