Monotonic Alpha-divergence Minimisation for Variational Inference
Authors: Kamélia Daudel, Randal Douc, François Roueff
JMLR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Lastly, we provide empirical evidence that our methodology yields improved results on several multimodal target distributions and on a real data example. [...] Section 8. Numerical Experiments |
| Researcher Affiliation | Academia | 1: LTCI, Telecom Paris Institut Polytechnique de Paris, France 2: Department of Statistics, University of Oxford Oxford OX1 3TG, United Kingdom [...] SAMOVAR, Telecom Sud Paris Institut Polytechnique de Paris, France |
| Pseudocode | Yes | Algorithm 1: Maximisation approach algorithm for mixture models |
| Open Source Code | No | No explicit statement about open-source code release or a link to a repository is provided in the paper. |
| Open Datasets | Yes | We select the Covertype data set (581, 012 data points and 54 features, available at https: //www.csie.ntu.edu.tw/ cjlin/libsvmtools/datasets/binary.html) |
| Dataset Splits | No | The paper uses the Covertype data set for Bayesian Logistic Regression but does not specify how the dataset was split into training, validation, or test sets for experimental evaluation. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, cloud instances) used for running the experiments. |
| Software Dependencies | No | The paper does not explicitly state specific software dependencies or their version numbers, such as programming languages, libraries, or frameworks used for implementation. |
| Experiment Setup | Yes | The covariance matrices of the mixture components are fixed and equal to σ2Id with σ2 = 1, α = 0.2, J {10, 50}, d = 16, M = 200, the total number of iterations N is equal to 100, Θ1 is generated by sampling from a centered normal distribution with covariance matrix 10Id, λ1 = [1/J, . . . , 1/J] and for all time n = 1 . . . N, κn = 0, ηn = 0. and γn = γ with γ {0.1, 0.5, 1.}. The experiments are replicated independently 30 times and the convergence of the RGD and of the MG approaches is monitored for the three multimodal examples (i), (ii) and (iii) by computing a Monte Carlo estimator of the VR Bound... |