Reproduce “Replica-Aware Split Inference for Edge IoT Systems” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zied Jenhani, Nasrin Razmi, Admela Jukan

The authors recover split inference after device failures by placing copies of failed segments on survivors, leaving original segments fixed. A mixed-integer linear program jointly chooses replica hosts and communication protocols under memory constraints to minimize modeled computation and transfer latency. In a simulated 15-device network, average latency was 3,361.92 ms, about 6.8% below Computation Greedy’s 3,607.94 ms; link-specific rate reductions were bounded at 75%.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.