27 citations · 34 across the 8 of their papers we have counts for
9 papers
Input margins can predict generalization too
Coenraad Mouton, Marthinus W. Theunissen, Marelie H. Davel
Understanding generalization in deep neural networks is an active area of research. A promising avenue of exploration has been that of margin measurements: the shortest distance to…
The Missing Margin: How Sample Corruption Affects Distance to the Boundary in ANNs
Marthinus W. Theunissen, Coenraad Mouton, Marelie H. Davel
Classification margins are commonly used to estimate the generalization ability of machine learning models. We present an empirical study of these margins in artificial neural netw…
Efficient acoustic feature transformation in mismatched environments using a Guided-GAN
Walter Heymans, Marelie H. Davel, Charl van Heerden
We propose a new framework to improve automatic speech recognition (ASR) systems in resource-scarce environments using a generative adversarial network (GAN) operating on acoustic…
Multi-style Training for South African Call Centre Audio
Walter Heymans, Marelie H. Davel, Charl van Heerden
Mismatched data is a challenging problem for automatic speech recognition (ASR) systems. One of the most common techniques used to address mismatched data is multi-style training (…
Exploring layerwise decision making in DNNs
Coenraad Mouton, Marelie H. Davel
While deep neural networks (DNNs) have become a standard architecture for many machine learning tasks, their internal decision-making process and general interpretability is still…
Tracking translation invariance in CNNs
Johannes C. Myburgh, Coenraad Mouton, Marelie H. Davel
Although Convolutional Neural Networks (CNNs) are widely used, their translation invariance (ability to deal with translated inputs) is still subject to some controversy. We explor…
Stride and Translation Invariance in CNNs
Coenraad Mouton, Johannes C. Myburgh, Marelie H. Davel
Convolutional Neural Networks have become the standard for image classification tasks, however, these architectures are not invariant to translations of the input image. This lack…
Pre-interpolation loss behaviour in neural networks
Arthur E. W. Venter, Marthinus W. Theunissen, Marelie H. Davel
When training neural networks as classifiers, it is common to observe an increase in average test loss while still maintaining or improving the overall classification accuracy on t…
DNNs as Layers of Cooperating Classifiers
Marelie H. Davel, Marthinus W. Theunissen, Arnold M. Pretorius +1
A robust theoretical framework that can describe and predict the generalization ability of deep neural networks (DNNs) in general circumstances remains elusive. Classical attempts…