Open Access
Review
Table 5
Cross-layer affinity matrix: Mapping compression to edge hardware dispatching
| Compression paradigm | Hardware affinity | System-level optimization | Primary metric impact |
|---|---|---|---|
| Quantization | NPU/GPU | Operational intensity teconfiguration | ↑ Throughput (batching) |
| Structured pruning | NPU (SIMD) | Execution pattern alignment | ↓ Latency (vectorization) |
| Token pruning | NPU/DSP | SRAM footprint reduction | ↓ Memory usage (KV-cache) |
| Mixed-precision MoE | SoC | Dynamic resource offloading | ↓ Energy per inference |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.