We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Gilbert MacKenzie's contribution to the Discussion of 'the Discussion Meeting on Probabilistic and statistical aspects of machine learning'.
- Authors
MacKenzie, Gilbert
- Abstract
The article discusses Gilbert MacKenzie's contribution to a discussion on probabilistic and statistical aspects of machine learning. MacKenzie emphasizes the importance of collaboration between statisticians and machine learners in developing better models. He focuses on the issue of over-parametrization in feed-forward neural nets and suggests a non-linear regression approach. MacKenzie also discusses the use of Bayesian information criterion for model selection and the potential for framing neural net analysis in a regression-like format. The article acknowledges that this work is still in its early stages and highlights the unresolved problem of "explanation" in machine learning.
- Subjects
MACHINE learning; ARTIFICIAL neural networks; FEEDFORWARD neural networks
- Publication
Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2024, Vol 86, Issue 2, p307
- ISSN
1369-7412
- Publication type
Proceeding
- DOI
10.1093/jrsssb/qkad148