Neural networks in 90 seconds: the actual math of forward propagation
HarperSCarroll · x · 2026-11-08
Educator Harper Carroll walks through the core math of neural networks with a concrete example: layers of neurons connected by weighted edges, inputs first transformed into embeddings (thousands of dimensions in real models, 2 here). With inputs 5 and 2, a neuron with edge weights 10 and 2 computes 5×10+2×2=54; a neighbor with weights 20 and 4 yields 108. Multiply, add, pass forward, repeat layer by layer — nearly every generative AI model runs on this calculation. She notes the demo omits nonlinear activations and teases follow-ups on output layers and backpropagation.
More from Research
- LLMs Are Now Both Authors and Reviewers, Quip Highlights Peer Review Loop — miniapeur · 2026-11-08
- Why lowering sampling temperature fixes generative models: new study on limited-data bias — bravo_abad · 2026-11-08
- Academic protest: reviewers caught using AI to auto-generate peer reviews — TuhinChakr · 2026-11-08
- The AI Papers Iceberg: From Attention to Obscure RL Papers in 7 Tiers — burny_tech · 2026-11-08
- LEV-2.6B jumps 63.2% to 76.2% on JevBench as depth goes 1→4; on-demand depth saves little — RobertTLange · 2026-11-08
- Oxford researchers argue LLMs can't invent: 'Theory Is All You Need' paper on data vs. human cognition — alex_verem · 2026-11-08