https://tahabouhsine.com/blog/https://tahabouhsine.com/blog/about/https://tahabouhsine.com/blog/activations-are-bad-for-geometry/https://tahabouhsine.com/blog/ai-illiteracy-pt1/https://tahabouhsine.com/blog/attention-is-a-kernel/https://tahabouhsine.com/blog/cheap-attention-is-linear-attention/https://tahabouhsine.com/blog/convex-readout-jax-flax-nnx/https://tahabouhsine.com/blog/latent-on-the-spectrum/https://tahabouhsine.com/blog/linear-attention-jax-flax-nnx/https://tahabouhsine.com/blog/not-all-infinities-are-equal/https://tahabouhsine.com/blog/opposite-is-not-different/https://tahabouhsine.com/blog/organizing-randomness-jax/https://tahabouhsine.com/blog/poem-0-1/https://tahabouhsine.com/blog/qk-projections-jax-flax-nnx/https://tahabouhsine.com/blog/readout-as-convex-combination/https://tahabouhsine.com/blog/search/https://tahabouhsine.com/blog/tag/activation-functions/https://tahabouhsine.com/blog/tag/africa-tech/https://tahabouhsine.com/blog/tag/ai-education/https://tahabouhsine.com/blog/tag/ai-literacy/https://tahabouhsine.com/blog/tag/ai/https://tahabouhsine.com/blog/tag/attention/https://tahabouhsine.com/blog/tag/bilinear/https://tahabouhsine.com/blog/tag/clip/https://tahabouhsine.com/blog/tag/contrastive-learning/https://tahabouhsine.com/blog/tag/contrastive/https://tahabouhsine.com/blog/tag/cosine-similarity/https://tahabouhsine.com/blog/tag/cross-entropy/https://tahabouhsine.com/blog/tag/deep-learning/https://tahabouhsine.com/blog/tag/efficient-transformers/https://tahabouhsine.com/blog/tag/embeddings/https://tahabouhsine.com/blog/tag/flax/https://tahabouhsine.com/blog/tag/frame-theory/https://tahabouhsine.com/blog/tag/geometry/https://tahabouhsine.com/blog/tag/hallucination/https://tahabouhsine.com/blog/tag/hierarchy/https://tahabouhsine.com/blog/tag/implementation/https://tahabouhsine.com/blog/tag/infonce/https://tahabouhsine.com/blog/tag/information-theory/https://tahabouhsine.com/blog/tag/information/https://tahabouhsine.com/blog/tag/interpretability/https://tahabouhsine.com/blog/tag/jacobian/https://tahabouhsine.com/blog/tag/jax/https://tahabouhsine.com/blog/tag/kernel-methods/https://tahabouhsine.com/blog/tag/kernels/https://tahabouhsine.com/blog/tag/label-structure/https://tahabouhsine.com/blog/tag/latent-space/https://tahabouhsine.com/blog/tag/linear-attention/https://tahabouhsine.com/blog/tag/loss-functions/https://tahabouhsine.com/blog/tag/loss-landscape/https://tahabouhsine.com/blog/tag/mechanistic-interpretability/https://tahabouhsine.com/blog/tag/ml/https://tahabouhsine.com/blog/tag/mlp/https://tahabouhsine.com/blog/tag/modality-gap/https://tahabouhsine.com/blog/tag/morocco/https://tahabouhsine.com/blog/tag/nadaraya-watson/https://tahabouhsine.com/blog/tag/neural-collapse/https://tahabouhsine.com/blog/tag/neural-networks/https://tahabouhsine.com/blog/tag/nnx/https://tahabouhsine.com/blog/tag/optax/https://tahabouhsine.com/blog/tag/orthogonality/https://tahabouhsine.com/blog/tag/packing/https://tahabouhsine.com/blog/tag/performer/https://tahabouhsine.com/blog/tag/phase-transitions/https://tahabouhsine.com/blog/tag/poem/https://tahabouhsine.com/blog/tag/policy/https://tahabouhsine.com/blog/tag/prototypes/https://tahabouhsine.com/blog/tag/query-key/https://tahabouhsine.com/blog/tag/random-features/https://tahabouhsine.com/blog/tag/relu/https://tahabouhsine.com/blog/tag/representation-learning/https://tahabouhsine.com/blog/tag/residual-stream/https://tahabouhsine.com/blog/tag/rkhs/https://tahabouhsine.com/blog/tag/rope/https://tahabouhsine.com/blog/tag/self-attention/https://tahabouhsine.com/blog/tag/self-supervised-learning/https://tahabouhsine.com/blog/tag/siglip/https://tahabouhsine.com/blog/tag/simclr/https://tahabouhsine.com/blog/tag/simplex-etf/https://tahabouhsine.com/blog/tag/simplex-packing/https://tahabouhsine.com/blog/tag/simplex/https://tahabouhsine.com/blog/tag/spectral-embedding/https://tahabouhsine.com/blog/tag/supcon/https://tahabouhsine.com/blog/tag/taxonomy/https://tahabouhsine.com/blog/tag/tight-frames/https://tahabouhsine.com/blog/tag/training-dynamics/https://tahabouhsine.com/blog/tag/transformers/https://tahabouhsine.com/blog/tag/triplet-loss/https://tahabouhsine.com/blog/tag/welch-bound/https://tahabouhsine.com/blog/tag/yat-unit/https://tahabouhsine.com/blog/tags/https://tahabouhsine.com/blog/three-states-of-information/https://tahabouhsine.com/blog/untangling-the-moons/https://tahabouhsine.com/blog/welch-bound-good-latent-space/https://tahabouhsine.com/blog/welch-bound-jax-analysis/https://tahabouhsine.com/blog/what-an-mlp-knows/https://tahabouhsine.com/blog/why-attention-needs-qk-projections/