Experimental benchmarks show ESP-NOW protocol yields lowest latency in TinyML split learning on ESP32-S3, with a beam search optimizer achieving near-optimal split points in 0.1 seconds for small device groups.
Quantization and training of neural networks for efficient integer-arithmetic-only inference
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.NI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Optimizing Split Learning Latency in TinyML-Based IoT Systems
Experimental benchmarks show ESP-NOW protocol yields lowest latency in TinyML split learning on ESP32-S3, with a beam search optimizer achieving near-optimal split points in 0.1 seconds for small device groups.