/deep_evidential_regression_loss_pytorch

Implementation of Deep evidential regression paper

Primary LanguagePythonApache License 2.0Apache-2.0


Deep Evidential Regression Loss Function

Paper Conference

Description

The paper "Deep Evidential Uncertainty/Regression" was submitted to ICLR where it was rejected[1]. The idea is inline with light of Sensoy et al.[2] and Malinin & Gales[3]. It was rejected because of lack of experiments and similar ideas with Malinin thesis. The goal is to implement the loss function and validate the results.

Installation

Typical Install

pip install git+https://github.com/deebuls/deep_evidential_regression_loss_pytorch

Development

git clone https://github.com/deebuls/deep_evidential_regression_loss_pytorch
cd deep_evidential_regression_loss_pytorch
pip install -e .[dev]

Tests can then be run from the root of the project using:

nosetests

Usage

To use this code EvidentialLossSumOfSquares and create loss function. loss.py implements the evidential loss function.

Check examples for detailed usage example

ToDo

  1. Different variation of the loss (NLL, with log(alpha, beta, lambda))
  2. When output is image as case of VAE
  3. Examples
  4. Test cases

Abstract

Deterministic neural networks (NNs) are increasingly being deployed in safety critical domains, where calibrated, robust and efficient measures of uncertainty are crucial. While it is possible to train regression networks to output the parameters of a probability distribution by maximizing a Gaussian likelihood function, the resulting model remains oblivious to the underlying confidence of its predictions. In this paper, we propose a novel method for training deterministic NNs to not only estimate the desired target but also the associated evidence in support of that target. We accomplish this by placing evidential priors over our original Gaussian likelihood function and training our NN to infer the hyperparameters of our evidential distribution. We impose priors during training such that the model is penalized when its predicted evidence is not aligned with the correct output. Thus the model estimates not only the probabilistic mean and variance of our target but also the underlying uncertainty associated with each of those parameters. We observe that our evidential regression method learns well-calibrated measures of uncertainty on various benchmarks, scales to complex computer vision tasks, and is robust to adversarial input perturbations.

References