Collapsed Inference for Bayesian Deep Learning

Authors: Zhe Zeng, Guy Van den Broeck

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On various regression and classification tasks, our collapsed Bayesian deep learning approach demonstrates significant improvements over existing methods and sets a new state of the art in terms of uncertainty estimation as well as predictive performance.
Researcher Affiliation Academia Zhe Zeng Computer Science Department University of California, Los Angeles zhezeng@cs.ucla.edu Guy Van den Broeck Computer Science Department University of California, Los Angeles guyvdb@cs.ucla.edu
Pseudocode Yes We summarize our proposed algorithm CIBER in Algorithm 1 in the Appendix.
Open Source Code Yes Code and experiments are available at https://github.com/UCLA-Star AI/CIBER.
Open Datasets Yes We experiment on 5 small UCI datasets: boston, concrete, yacht, naval and energy. We further experiment on 6 large UCI datasets: elevators, keggdirected, keggundirected, pol, protein and skillcraft. CIFAR-10 and CIFAR-100 (Krizhevsky et al., 2009)
Dataset Splits Yes The hyperparameters including learning rates and weight decay are tuned by performing a grid search to maximize the Gaussian log likelihood using a validation split.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory, or cloud instance types) used for running the experiments.
Software Dependencies No The paper mentions 'pywmi library' but does not specify version numbers for any software dependencies required to replicate the experiments.
Experiment Setup Yes All the network models are trained for 300 epochs using SGD. We start the weight collection after epoch 160 with step size 5. We follow exactly the same hyperparameters as Maddox et al. (2019) including learning rates and weight decay parameters.