Nonparametric Iterative Machine Teaching

Authors: Chen Zhang, Xiaofeng Cao, Weiyang Liu, Ivor Tsang, James Kwok

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we verify the correctness of our theoretical findings with extensive experiments in nonparametric scenarios. We test our RFT and GFT on both synthetic and real-world data, on which we find these two algorithms present satisfactory capability to tackle nonparametric teaching tasks.
Researcher Affiliation Academia 1School of Artificial Intelligence, Jilin University, China 2Max Planck Institute for Intelligent Systems, T ubingen, Germany 3University of Cambridge, United Kingdom 4Centre for Frontier AI Research and Institute of High Performance Computing, A*STAR, Singapore 5Hong Kong University of Science and Technology.
Pseudocode Yes Algorithm 1 Random / Greedy Functional Teaching
Open Source Code Yes Our source code is available at https://github.com/chen2hang/Nonparametric Teaching.
Open Datasets Yes Consider a digit (MNIST (Le Cun, 1998)) teaching instance, one can image a digit figure as a surface in 3D space... EMNIST from (Cohen et al., 2017)... we pick two facial figures form the ORL database (http://www.cam-orl.co.uk)
Dataset Splits No The paper does not explicitly provide training/test/validation dataset splits needed to reproduce the experiment. While it mentions using MNIST (both training and testing sets), it does not specify the split percentages or counts.
Hardware Specification Yes Our implementation is based on Intel(R) Core(TM) i7-8750H and NVIDIA GTX 1050 Ti with Max-Q Design.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., specific library versions or solver versions).
Experiment Setup Yes For this regression problem, we assume the loss function of the learner is square loss L = (y - f(x))^2... The learning rate ηt is fixed as 0.01. ... We set kernel as the popular and general RBF K(x, x') = exp(-||x - x'||^2 / (2 * sigma^2)).