Trusted Multi-view Learning with Label Noise

Authors: Cai Xu, Yilin Zhang, Ziyu Guan, Wei Zhao

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically compare TMNR with state-of-the-art trusted multi-view learning and label noise learning baselines on 5 publicly available datasets. Experiment results show that TMNR outperforms baseline methods on accuracy, reliability and robustness.
Researcher Affiliation Academia School of Computer Science and Technology, Xidian University
Pseudocode Yes Algorithm 1 TMNR algorithm
Open Source Code Yes The code and appendix are released at https://github.com/Yilin Zhang107/TMNR.
Open Datasets Yes UCI1 contains features for handwritten numerals (0-9). The average of pixels in 240 windows, 47 Zernike moments, and 6 morphological features are used as 3 views. PIE2 consists of 680 face images from 68 experimenters. We extracted 3 views from it: intensity, LBP and Gabor. BBC3 includes 685 documents from BBC News that can be categorised into 5 categories and are depicted by 4 views. Caltech1014 contains 8677 images from 101 categories, extracting features as different views with 6 different methods: Gabor, Wavelet Moments, CENTRIST, HOG, GIST, and LBP. we chose the first 20 categories. Leaves1005 consists of 1600 leaf samples from 100 plant species. We extracted shape descriptors, fine-scale edges, and texture histograms as 3 views.
Dataset Splits No No specific details regarding a validation dataset split were explicitly mentioned. The paper only states 'In all datasets, 20% of the instances are split as the test set.'
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for running experiments were mentioned in the paper.
Software Dependencies Yes We implement all methods on Py Torch 1.13 framework.
Experiment Setup Yes We utilize the Adam optimizer with a learning rate of 1e-3 and l2-norm regularization set to 1e-5.