Placeholder chapter page. Sections are produced by the book-skills pipeline.
Sections
- 59.1 Why sensory AI runs on-device
- 59.2 CPU, GPU, NPU, DSP, MCU tradeoffs
- 59.3 Memory, power, and thermal limits
- 59.4 Quantization (int8, int4, and hardware-native 2/4-bit)
- 59.5 Pruning and sparsity
- 59.6 Hardware-aware NAS
- 59.7 Edge-cloud partitioning and the deployment lifecycle
Lab 59
quantize and prune a sensor classifier for edge inference; measure latency/size/accuracy tradeoffs.