DeepAI AI Chat
Log In Sign Up

Evaluating Neural Machine Comprehension Model Robustness to Noisy Inputs and Adversarial Attacks

05/01/2020
by   Winston Wu, et al.
Johns Hopkins University
PNNL
0

We evaluate machine comprehension models' robustness to noise and adversarial attacks by performing novel perturbations at the character, word, and sentence level. We experiment with different amounts of perturbations to examine model confidence and misclassification rate, and contrast model performance in adversarial training with different embedding types on two benchmark datasets. We demonstrate improving model performance with ensembling. Finally, we analyze factors that effect model behavior under adversarial training and develop a model to predict model errors during adversarial attacks.

READ FULL TEXT
06/08/2020

Adversarial Feature Desensitization

Deep neural networks can now perform many tasks that were once thought t...
09/06/2019

Learning to Discriminate Perturbations for Blocking Adversarial Attacks in Text Classification

Adversarial attacks against machine learning models have threatened vari...
04/23/2021

Evaluating Deception Detection Model Robustness To Linguistic Variation

With the increasing use of machine-learning driven algorithmic judgement...
04/29/2022

Logically Consistent Adversarial Attacks for Soft Theorem Provers

Recent efforts within the AI community have yielded impressive results t...
03/23/2023

Decentralized Adversarial Training over Graphs

The vulnerability of machine learning models to adversarial attacks has ...
06/17/2021

Evaluating the Robustness of Bayesian Neural Networks Against Different Types of Attacks

To evaluate the robustness gain of Bayesian neural networks on image cla...