Nonparametric Bayesian Deep Networks with Local Competition

Authors: Konstantinos Panousis, Sotirios Chatzis, Sergios Theodoridis

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental As we experimentally show using benchmark datasets, our approach yields networks with less computational footprint than the state-of-the-art, and with no compromises in predictive accuracy.
Researcher Affiliation Academia 1Dept. of Informatics & Telecommunications, National and Kapodistrian University of Athens, Greece 2Dept. of Electrical Eng., Computer Eng., and Informatics, Cyprus University of Technology, Limassol, Cyprus 3The Chinese University of Hong Kong, Shenzen, China.
Pseudocode No The paper describes training and inference algorithms but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not include any explicit statements about releasing source code or provide a link to a code repository for the described methodology.
Open Datasets Yes We use MNIST in these experiments. ... We train the network from scratch on the MNIST dataset... Finally, we perform experimental evaluations on a more challenging benchmark dataset, namely CIFAR-10 (Krizhevsky & Hinton, 2009).
Dataset Splits No The paper mentions training on MNIST and CIFAR-10 datasets and evaluating on a 'test set', but it does not provide specific details on the train/validation/test dataset splits (e.g., percentages, sample counts, or explicit reference to predefined splits).
Hardware Specification Yes We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.
Software Dependencies No The paper mentions using ADAM for ELBO maximization but does not specify any software names with version numbers for libraries, frameworks, or programming languages.
Experiment Setup Yes In our experiments, the stick variables are drawn from a Beta(1, 1) prior. The hyperparameters of the approximate Kumaraswamy posteriors of the sticks are initialized as follows: the ak s are set equal to the number of LWTA blocks of their corresponding layer; the bk s are always set equal to 1. All other initializations are random within the corresponding support sets. The employed cut-off threshold, τ, is set to 10 2.