https://mistatle.github.io/en/ Notes on mathematics, algorithms and AI: structure, mechanism and boundary. en https://mistatle.github.io/en/notes/littlestone-dimension/ https://mistatle.github.io/en/notes/littlestone-dimension/ Sun, 11 Oct 2026 16:00:00 GMT Littlestone dimension, the optimal mistake bound, and the SOA proof, with a comparison to VC dimension. https://mistatle.github.io/en/notes/policy-gradient-optimization/ https://mistatle.github.io/en/notes/policy-gradient-optimization/ Tue, 05 May 2026 16:00:00 GMT One continuous line from the objective to gradient estimation, variance and stepsize choice. https://mistatle.github.io/en/notes/leworldmodel-sigreg/ https://mistatle.github.io/en/notes/leworldmodel-sigreg/ Tue, 24 Mar 2026 16:00:00 GMT Connecting latent prediction, representation collapse and SIGReg, and identifying the conditions the mechanism actually depends on. https://mistatle.github.io/en/notes/typical-set/ https://mistatle.github.io/en/notes/typical-set/ Tue, 20 Jan 2026 16:00:00 GMT Why does probability mass sit on a thin shell rather than at the point of highest density? https://mistatle.github.io/en/notes/scaled-dot-product-attention/ https://mistatle.github.io/en/notes/scaled-dot-product-attention/ Sat, 22 Nov 2025 16:00:00 GMT Deriving the Transformer's scaling factor from dot-product variance, softmax saturation and gradient stability, with a PyTorch implementation.