Researchers introduce Neural networks with differentiable Belief-Function layers
The authors present a deep evidential pipeline that replaces the standard Dempster–Shafer normaliser with Yager’s 1987 rule of combination. The new rule is bilinear and avoids the $(1-\kappa)^{-2}$ term that causes gradient divergence when evidence sources disagree.
Key points
- Authors use Yager's rule to avoid gradient divergence in belief‑function layers
- Yager's vacuity is 9× larger than Dempster's on contested CIFAR‑10H images
- The network's ignorance output serves as an "I do not know" signal
In experiments on CIFAR‑10H, Yager’s vacuity is 9× larger on the most contested images, making the network’s ignorance output $m(\Theta)$ a readable "I do not know" signal. The model also performs competitively on out‑of‑distribution detection against strong single‑pass baselines, while a clamped Dempster variant becomes gradient‑unstable once the clamp is removed.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- MIT historian says early AI engineers added randomness to mimic human creativity · 1 src
- Researchers introduce MemPlan for text-based planning with PDDL · 1 src
- Researchers present CiteAnchor framework for detecting hallucinated citations · 1 src
- Researchers find self‑rewarded moral argumentation makes language models decisive but pliable · 1 src
- AQL-Lite improves Arabic LLMs with curated data and partial-parameter training · 1 src
Comments
via GitHub Discussions