Semi-Supervised Streaming Learning with Emerging New Labels

Authors: Yong-Nan Zhu, Yu-Feng Li7015-7022

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical studies on several datasets validate that the algorithm can accurately classify points on a dynamic stream with a small number of labeled examples and emerging new classes.
Researcher Affiliation Academia National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China
Pseudocode Yes Algorithm 1 summarizes the approach. Algorithm 2 summarizes the construction of SEENTree. SEENLP is presented in Algorithm 3.
Open Source Code No The paper does not provide any explicit statement or link for open-source code availability.
Open Datasets Yes To evaluate the predictive performance of the proposed SEEN approach, we use four multi-class benchmark datasets ( segment , satimage , usps , pendigits , detailed information is shown in Table 2)1. 1https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass.html
Dataset Splits No The paper describes how the data stream is simulated, with new classes emerging over time, but it does not specify traditional train/validation/test dataset splits with exact percentages or sample counts for static partitioning.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper mentions algorithms used (e.g., SVM, K-means) and their settings, but it does not specify software dependencies with version numbers (e.g., 'Python 3.x', 'PyTorch 1.x').
Experiment Setup Yes For SEENForest, hm = 7, s = 50, φ = 128, and k is half the number of features. For SEENLP, τ = 50 and we use the standard RBF similarity to inialize the weight matrix, Wxi,xj = exp( xi xj 2/σ2), where σ can be obtained by crossvalidation from the first period. The number of labeled instances in each class in the four datasets as ordered in Table 2 is 20, 50, 80, and 50.