Video course · 15 chapters · 106 min · certificate
Understanding Deep Learning Research Tutorial - Theory, Code and Math
A method for understanding deep learning research: a 7-step paper-reading process, decoding formulas (QHAdam example), learning maths efficiently, reading large codebases, and a deep dive into Meta's Segment Anything Model from theory to image encoder, prompt encoder, mask decoder, data engine and limitations.
What you'll learn
- Read research papers with a structured process
- Translate formulas into intuition
- Learn the maths you need efficiently
- Navigate deep learning codebases
- Explain the Segment Anything Model architecture
- Assess zero-shot results and limitations
Chapters
15 chapters · 106:13-
3:49
01Intro Members
How to read a research paper
Not so intimidating.
-
4:40
02Reading Members
External context and first reads
Steps 1-5.
-
4:38
03Method Members
Method walk and reading formulas
Steps 6-7.
-
23:50
04Formulas Members
Translating symbols into meaning
QHAdam.
-
7:34
05Intuition Members
Intuition and learning maths
Make it make sense.
-
3:38
06Maths Members
Choosing maths resources
Sub-fields and exercises.
-
5:07
07Code Members
Study theory and reading codebases
Fix weak spots.
-
3:31
08Structure Members
Mapping codebase structure
Architecture.
-
6:26
09Components Members
Elucidating components
Abstraction layers.
-
5:37
10SAM Members
Notes and the SAM deep dive
Depth-first.
-
8:24
11Theory Members
SAM testing and theory
Try it first.
-
6:32
12Overview Members
SAM code overview
Three components.
-
4:47
13Encoders Members
Image and prompt encoder code
ViT backbone.
-
11:48
14Decoder Members
Mask decoder code
The messy part.
-
5:52
15Wrap-up Members
Data engine, results and limitations
Zero-shot.
Jobs this course suits
Our AI checked this course against 500 jobs; these get the most out of it. Each job links to its learning path.