Robust full Bayesian methods for neural networks
In this paper, we propose a full Bayesian model for neural networks. This model treats the model dimension (number of neurons), model parameters, regularisation parameters and noise parameters as random variables that need to be estimated. We then propose a reversible jump Markov chain Monte Carlo (...
| Váldodahkkit: | , , |
|---|---|
| Materiálatiipa: | Conference item |
| Almmustuhtton: |
Neural information processing systems foundation
2000
|
| _version_ | 1826281649894064128 |
|---|---|
| author | Andrieu, C de Freitas, J Doucet, A |
| author_facet | Andrieu, C de Freitas, J Doucet, A |
| author_sort | Andrieu, C |
| collection | OXFORD |
| description | In this paper, we propose a full Bayesian model for neural networks. This model treats the model dimension (number of neurons), model parameters, regularisation parameters and noise parameters as random variables that need to be estimated. We then propose a reversible jump Markov chain Monte Carlo (MCMC) method to perform the necessary computations. We find that the results are not only better than the previously reported ones, but also appear to be robust with respect to the prior specification. Moreover, we present a geometric convergence theorem for the algorithm. |
| first_indexed | 2024-03-07T00:31:59Z |
| format | Conference item |
| id | oxford-uuid:80200a99-02e5-4fb2-884f-b08fc91666ae |
| institution | University of Oxford |
| last_indexed | 2024-03-07T00:31:59Z |
| publishDate | 2000 |
| publisher | Neural information processing systems foundation |
| record_format | dspace |
| spelling | oxford-uuid:80200a99-02e5-4fb2-884f-b08fc91666ae2022-03-26T21:21:13ZRobust full Bayesian methods for neural networksConference itemhttp://purl.org/coar/resource_type/c_5794uuid:80200a99-02e5-4fb2-884f-b08fc91666aeSymplectic Elements at OxfordNeural information processing systems foundation2000Andrieu, Cde Freitas, JDoucet, AIn this paper, we propose a full Bayesian model for neural networks. This model treats the model dimension (number of neurons), model parameters, regularisation parameters and noise parameters as random variables that need to be estimated. We then propose a reversible jump Markov chain Monte Carlo (MCMC) method to perform the necessary computations. We find that the results are not only better than the previously reported ones, but also appear to be robust with respect to the prior specification. Moreover, we present a geometric convergence theorem for the algorithm. |
| spellingShingle | Andrieu, C de Freitas, J Doucet, A Robust full Bayesian methods for neural networks |
| title | Robust full Bayesian methods for neural networks |
| title_full | Robust full Bayesian methods for neural networks |
| title_fullStr | Robust full Bayesian methods for neural networks |
| title_full_unstemmed | Robust full Bayesian methods for neural networks |
| title_short | Robust full Bayesian methods for neural networks |
| title_sort | robust full bayesian methods for neural networks |
| work_keys_str_mv | AT andrieuc robustfullbayesianmethodsforneuralnetworks AT defreitasj robustfullbayesianmethodsforneuralnetworks AT douceta robustfullbayesianmethodsforneuralnetworks |