Examining Redundancy in the Context of Safe Machine Learning
This work addresses the problem of ensuring safety and dependability in machine learning systems, particularly for critical applications, but it is incremental as it focuses on preliminary experiments with basic methods.
The paper investigated naive implementations of redundant neural network architectures on the MNIST digit database to explore safe and dependable machine learning, finding that these approaches highlight expected difficulties in using such classifiers for safety-critical systems.
This paper describes a set of experiments with neural network classifiers on the MNIST database of digits. The purpose is to investigate naïve implementations of redundant architectures as a first step towards safe and dependable machine learning. We report on a set of measurements using the MNIST database which ultimately serve to underline the expected difficulties in using NN classifiers in safe and dependable systems.