CapProNet: Deep Feature Learning via Orthogonal Projections onto Capsule Subspaces

Authors: Liheng Zhang, Marzieh Edraki, Guo-Jun Qi

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiment results on image datasets show the presented model can greatly improve the performance of the state-of-the-art Res Net backbones by 10 20% and that of the Densenet by 5 7% respectively at the same level of computing and memory expenses. The Cap Pro Net establishes the competitive state-of-the-art performance for the family of capsule nets by significantly reducing test errors on the benchmark datasets.
Researcher Affiliation Collaboration Laboratory for MAchine Perception and LEarning, University of Central Florida http://maple.cs.ucf.edu Huawei Cloud, Seattle, USA Corresponding author: G.-J. Qi, email: guojunq@gmail.com and guojun.qi@huawei.com.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes The source code is available at https://github.com/maple-research-lab.
Open Datasets Yes We use both CIFAR and SVHN datasets in experiments to evaluate the performance. CIFAR The CIFAR dataset contains 50,000 and 10,000 images of 32 32 pixels for the training and test sets respectively. SVHN The Street View House Number (SVHN) dataset has 73,257 and 26,032 images of colored digits in the training and test sets, with an additional 531,131 training images available. Image Net The Image Net data-set consists of 1.2 million training and 50k validation images.
Dataset Splits Yes CIFAR ... A separate validation set of 5,000 images are split from the training set to choose the model hyperparameters, and the final test errors are reported with the chosen hyperparameters by training the model on all 50,000 training images. SVHN ... a separate validation set of 6,000 images is split from the training set.
Hardware Specification No The paper mentions training on 'four GPUs' but does not specify the model or other hardware details.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup No The paper mentions using training strategies like 'learning rate schedule, parameter initialization, and the stochastic optimization solver' from backbone papers, but does not provide specific hyperparameter values or detailed configurations within the text.