Learning with Feature and Distribution Evolvable Streams

Authors: Zhen-Yu Zhang, Peng Zhao, Yuan Jiang, Zhi-Hua Zhou

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical studies on synthetic data verify the rationale of our proposed discrepancy measure, and extensive experiments on real-world tasks validate the effectiveness of our algorithm.
Researcher Affiliation Academia 1National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China. Correspondence to: Yuan Jiang <jiangy@lamda.nju.edu.cn>.
Pseudocode No The paper describes the algorithm steps and framework but does not include a formal pseudocode block or algorithm listing.
Open Source Code No The paper does not provide any statement or link regarding the availability of its source code.
Open Datasets Yes RFID Dataset (Hou et al., 2017) is real-time data streams collected by the RFID technique. Amazon Dataset (Mc Auley et al., 2015) contains the product s quality (label) from 2006 to 2008 according to the ratings of its users (feature). Reuters multilingual dataset (Amini et al., 2009) contains about 11K articles from 6 classes in 5 languages so that we can simulate the evolving stream by various languages.
Dataset Splits No The paper describes how evolving data is generated (e.g., '20% evolving data in each mini-batch') and how data is categorized or split for task generation, but it does not specify explicit train/validation/test dataset splits with percentages or sample counts.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions general software components like 'SGD', 'cross-entropy loss', 'MLP', and 'ReLU', but it does not provide specific version numbers for any libraries, frameworks, or programming languages used.
Experiment Setup Yes For implementations of the EDM algorithm, we set the main classifiers (min-player) and auxiliary classifiers (maxplayer) in the adversarial network as two 5-layer MLP with Re LU as activation functions. The model is trained by SGD with a learning rate of 0.004 and regularization weight decay 0.005.