Epistemic Neural Networks
Ian Osband, Zheng Wen, Seyed Mohammad Asghari, Vikranth Dwaracherla, Morteza Ibrahimi, Xiuyuan Lu, and Benjamin Van Roy DeepMind, Efficient Agent Team, Mountain View Contact
Abstract
Intelligent agents need to know what they don’t know, and this capability can be evaluated through the quality of joint predictions. In principle, ensemble methods can produce effective joint predictions, but the compute costs are prohibitive for large models. We introduce the epinet: an architecture that can supplement any conventional neural network, including large pretrained models, and can be trained with modest incremental computation to estimate uncertainty. With an epinet, conventional neural networks outperform large ensembles of hundreds or more particles, and use orders of magnitude less computation. The epinet does not fit the traditional framework of Bayesian neural networks, so we introduce the epistemic neural network (ENN) as a general interface for models that generate joint predictions.
中文速览
让神经网络"知道自己不知道什么"是智能系统的核心难题，而衡量这一能力的关键在于模型能否给出高质量的联合预测（joint prediction）——即对多个输入同时预测时，能否区分"数据本身的随机性"和"训练数据不足导致的不确定性"。传统的集成方法（ensemble）虽然能做到这一点，但用几百个模型副本来逼近贝叶斯后验，计算代价对大模型而言完全不可接受。作者提出了一种叫做 epinet 的轻量附加网络结构，可以插接在任意已有神经网络（包括大型预训练模型）之上，通过引入一个"认知索引"（epistemic index）让模型输出随不确定性而变化，从而以极少的额外计算生成有意义的联合预测。实验表明，加了 epinet 的单个 ResNet 在联合预测质量上超过了拥有 100 个副本的大型集成模型，而总参数量还不到两个副本，这意味着在不牺牲分类精度的前提下，可以用几乎可忽略的算力代价换来大幅更优的不确定性估计，对强化学习、主动学习和安全决策等下游任务具有重要价值。
原文 arXiv:2107.08924;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2107.08924v8