RNNs propagate hidden states through time; gradient products can shrink or grow. LSTM gates control cell-state retention and updates, improving long-range signal flow without guaranteeing stability. Self-attention shortens paths between positions, while dense attention still has quadratic sequence-length cost.
Representation learning compresses raw data into a 'useful' vector: autoencoders force a meaningful latent space via reconstruction, VQ discretizes it into a codebook, and self-supervision turns 'mask-and-reconstruct' into free supervision.
Similarity-based representation learning does not reconstruct input; it directly shapes latent geometry: pull same-class representations together, push different ones apart. InfoNCE turns this into 'spot the positive among negatives', and alignment / uniformity give it interpretable metrics.
Take a net to infinite width and its random-init output becomes a Gaussian process (NN–GP); its training dynamics freeze into the Neural Tangent Kernel (NTK). This theory analyzes nets and, in reverse, guides us to design the 'right inductive bias'.
Generative models learn the data distribution p(x). Density models model probability directly, energy models use an unnormalized potential + sampler, GANs let a discriminator force realistic samples, autoregressive predicts the next token step by step, and diffusion dodges tricky maximum-likelihood via 'add noise then learn to denoise'.
The core of VAE is ELBO + reparameterization: log p(x) is replaced with E_q[log p(x|z)] − KL(q(z|x)‖p(z)); the encoder outputs μ/σ and z = μ + σ⊙ε (ε ~ N(0,1)) makes sampling differentiable. Training = reconstruction + KL in tension, which gives rise to β-VAE, posterior collapse, VQ-VAE, and related fixes.
The key to conditional generation is 'feed y into the model': cGAN concatenates y into G/D; cVAE passes y to both encoder and decoder; in diffusion, Classifier Guidance uses gradients from an external classifier to push samples toward a class, while Classifier-Free Guidance trains conditional + unconditional together and linearly combines them at inference — the latter is the standard weapon behind Stable Diffusion and Imagen.
OOD failure is not a bug, it's the i.i.d. assumption breaking: covariate shift (image style changes), label shift (class proportions change), concept shift (a word's meaning changes) each need different responses; the most common cause is the model latching onto spurious correlations (using grass as a cue for cows); IRM and domain randomization try to fix this in training data structure, test-time adaptation fixes it at inference.
Transfer learning's core insight is 'features learned on big data are good general-purpose representations': freeze the backbone and train only a linear head when downstream data is tiny; full fine-tune when data is plentiful; reach for LoRA / adapter when compute is tight. SimCLR and MAE removed the need for upstream labels and pushed downstream quality another notch.
ResNet's skip connections solve degradation, enabling 100+ layer nets; EfficientNet compound scales depth/width/resolution; ConvNeXt absorbs Transformer design to reclaim CV crown.
HW0 checks CS181 prerequisites in four problems — y=Xw solvability, optimizing an objective, reasoning about randomness, and OLS in Python. The problem that slows you down most is the gap to patch before HW1.
HW1 uses an 800,000-year ice-core temperature dataset across four problems: kNN and kernel regression, a geometric proof of least squares, basis-function regression, and a probabilistic derivation of ridge and LASSO, ending with a coordinate-descent LASSO implementation.
CS181 2026 is A3 with hw0–6 as the weekly clock (no public recording links listed for the corresponding lectures); 2025 adds a practical, 2024 has two midterms, 2023 was taught by Weiwei Pan. Start with HW0, then follow hw1→hw6.
Week 0 is about search: DFS and BFS differ only in whether the frontier is a stack or a queue; A* adds the steps already taken to a heuristic estimate and guarantees the shortest path when the heuristic never overestimates; Minimax has the two players take the maximum and the minimum in turn, and alpha-beta pruning cuts branches that cannot change the result. With time links for every section, plus the approach, common mistakes and ways to check your work for the Degrees and Tic-Tac-Toe projects; the reference implementations are folded at the end of each project.
When CLAUDE.md rules are ignored, hooks never fire, or an MCP server shows no tools, the file usually didn't load, loaded from an unexpected location, or got overridden. This guide covers what the diagnostic entries (/context, /memory, /skills, /doctor, /mcp, and more) actually show, safe-mode bisection, and a table of six high-frequency error messages with fixes.
Work through install and login failures in five steps: verify PATH on your OS, confirm there is only one installation, test downloads.claude.ai for a 200, recover failed OAuth callbacks by pasting the login code or using claude auth login, then finish with claude doctor.
Five classes of runtime fixes: diagnose high memory with /compact plus /heapdump; recover a hung session with Ctrl+C and claude --resume; write large tables to files instead of forcing terminal output; escape autocompact thrashing by reading files in chunks or running /compact with a focus; fix broken search by installing system ripgrep and setting USE_BUILTIN_RIPGREP=0.
No budget needed: Tailscale plus built-in Screen Sharing is free and the most reliable stack for daily remote work on Mac — upgrade to Jump Desktop only if you want it smoother. This guide compares four options and fixes lid-close and sleep pitfalls in 5 minutes.
The key question in self-hosted inference isn't how fast the engine is — it's your GPU utilization. A fully saturated A100 costs ~$0.70 per million output tokens; at 10% utilization that becomes $7, more than most cloud APIs. This overview maps seven tools across three layers to help you decide which layer you need.
You cannot take model vendors' self-reported scores at face value. This guide covers the most important independent evaluation platforms, domain benchmarks, adoption indicators, and official sources in 2026: what each measures, how to read it, where it is biased, and which figures matter for different use cases.