Structure-Aware Convolutional Neural Networks

Authors: Jianlong Chang, Jie Gu, Lingfeng Wang, GAOFENG MENG, SHIMING XIANG, Chunhong Pan

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on eleven datasets strongly evidence that SACNNs outperform current models on various machine learning tasks, including image classification and clustering, text categorization, skeleton-based action recognition, molecular activity detection, and taxi flow prediction.
Researcher Affiliation Academia 1NLPR, Institute of Automation, Chinese Academy of Sciences 2School of Artificial Intelligence, University of Chinese Academy of Sciences
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes Specifically, Our core code will be released at https://github.com/vector-1127/SACNNs.
Open Datasets Yes We perform experiments on six Euclidean and five non-Euclidean structured datasets to verify the capability of SACNNs. Six Euclidean structured datasets include the Mnist [26], Cifar-10 [23], Cifar-100 [23], STL-10 [8], Image10 [6], and Image Dog [6] image datasets. Five non-Euclidean structured datasets contain the text categorization datasets 20NEWS and Reuters [25], the action recognition dataset NTU [36], the molecular activity dataset DPP4 [20], and the taxi flow dataset TF-198 [42] that consists of the taxis flow data at 198 traffic intersections in a city.
Dataset Splits Yes Figure 3 (a) illustrates the validation errors of SACNNs with different t. (b) Influence of channels... validation error (Cifar-10). The paper utilizes standard benchmark datasets (e.g., MNIST, CIFAR-10) that commonly have predefined train/validation/test splits.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific version numbers for ancillary software or libraries used in the experiments.
Experiment Setup Yes The hyper-parameters in SACNNs are set as follows. In our experiments, the max pooling and the Graclus method [11] are employed as the pooling operations to coarsen the feature maps in SACNNs when managing Euclidean and non-Euclidean structured data respectively, the Re LU function [13] is used as the activation function, batch normalization [17] is employed to normalize the inputs of all layers, parameters are randomly initialized with a uniform distribution U( 0.1, 0.1), the order of polynomials t is set to the maximum number of neighbors among the whole spatial domains (e.g., t = 9 if we attempt to learn 3 3 filters in images). During the training stage, the Adam optimizer [21] with the initial learning rate 0.001 is utilized to train SACNNs, the mini-batch size is set to 32, the categorical cross entropy loss is used in the classification tasks, and the mean squared error loss is used in the regression tasks.