Flashcards
One idea per card. Learn it, flip it, test yourself.
15
decks
370
cards
45
free cards
Gradient descent
Step downhill along the negative gradient, a little at a time.
Which change most often makes training diverge?
Core ideas behind how neural networks learn
From raw pixels to classical filters to modern deep vision
The core building blocks behind efficient code
Noise, score, and sampling — the machinery behind image generation
MDPs, value functions and policy gradients
From tokens and TF-IDF to attention and decoding
Scaling, alignment, prompting, retrieval and serving
ReAct, planning, memory and the failures that bite
Cameras, geometry, NeRF and Gaussian Splatting
CLIP, bridges, instruction tuning and what makes a VLM see
Python, NumPy, linear algebra and probability — the toolkit under ML
Regression, trees, margins and clusters — the classical ML toolkit
Linux, networking, containers, orchestration and serving models
GPU architecture, memory and writing fast kernels in Triton
Threads, memory and parallel patterns — GPU programming in Python