Sampling-Free Learning of Bayesian Quantized Neural Networks
Authors: Jiahao Su, Milan Cvitkovic, Furong Huang
ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate BQNs on MNIST, Fashion-MNIST, KMNIST and CIFAR10 image classification datasets. compared against bootstrap ensemble of QNNs (E-QNN). We demonstrate BQNs achieve both lower predictive errors and better-calibrated uncertainties than E-QNN (with less than 20% of the negative log-likelihood). |
| Researcher Affiliation | Collaboration | Jiahao Su Department of Electrical and Computer Engineering University of Maryland College Park, MD 20740 jiahaosu@umd.edu Milan Cvitkovic Amazon Web Services Seattle, WA, USA cvitkom@amazon.com Furong Huang Department of Computer Science University of Maryland College Park, MD 20740 furongh@cs.umd.edu |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about open-source code availability or links to a code repository for the described methodology. |
| Open Datasets | Yes | We evaluate BQNs on MNIST, Fashion-MNIST, KMNIST and CIFAR10 image classification datasets. |
| Dataset Splits | No | The paper mentions using training and test sets and provides sample counts for the Boston housing dataset, but it does not provide specific details on validation dataset splits or the methodology for creating these splits for other datasets like MNIST or CIFAR-10. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., GPU models, CPU types) used for running the experiments. |
| Software Dependencies | No | The paper mentions using the ADAM optimizer but does not specify its version or any other software dependencies with version numbers (e.g., programming languages, libraries, or frameworks). |
| Experiment Setup | Yes | In training, each image is augmented by a random shift within 2 pixels (with an additional random flipping for CIFAR10), and no augmentation is used in test. In the experiments, we consider a class of quantized neural networks, with both binary weights and activations (i.e. Q = { 1, 1}) with sign activations σ( ) = sign( ). For BQNs, the distribution parameters φ are initialized by Xavier s uniform initializer, and all models are trained by ADAM optimizer (Kingma & Ba, 2014) for 100 epochs (and 300 epochs for CIFAR10) with batch size 100 and initial learning rate 10 2, which decays by 0.98 per epoch. |