Research map / Label-efficient and robust learning / Recognition and long-tail
Inference Offloading for Cost-Sensitive Binary Classification at the Edge
Vishnu Narayanan Moothedath, Umang Agarwal, Umeshraja N, James Richard Gross, Jaya Prakash Champati, Sharayu Moharir
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use two thresholds (offload ambiguity region plus cost-sensitive local prediction) instead of a single threshold | hierarchical inference policy · BreakHis, Chest, Synthetic, BreaCh | average cost: up to 55% reduction at beta=0.6 (Synthetic) | Improved |
| use H2T2 instead of single-threshold HI on Phishing | hierarchical inference policy · Phishing | average cost | No clear effect |
| online H2T2 versus offline-optimal single-threshold policy | H2T2 · two of the datasets | average cost | Mixed |
| offload decisions learned online under distribution shift | H2T2 · BreaCh (OOD) | false negative rate (%): 38% -> below 1% | Improved |
| H2T2 vs static no-offload policy at high offload cost | H2T2 · Phishing (beta>0.3) | average cost | Hurt |
| increase threshold resolution beyond 0.0625 (ablation) | H2T2 · studied datasets | average cost | No clear effect |
Closest papers by meaning
- Efficient Edge Inference by Selective Query (ICLR 2023 · similarity 0.86)
- REDRESS: Generating Compressed Models for Edge Inference Using Tsetlin Machines (TPAMI 2023 · similarity 0.86)
- OCCAM: Towards Cost-Efficient and Accuracy-Aware Classification Inference (ICLR 2025 · similarity 0.84)
- Soft ascent-descent as a stable and flexible alternative to flooding (NeurIPS 2024 · similarity 0.83)
- Entropy-Gated Latent Recursion (NeurIPS 2026 · similarity 0.82)
- A Universal Growth Rate for Learning with Smooth Surrogate Losses (NeurIPS 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
