Uncertainty-aware Active Learning for Optimal Bayesian Classifier
Authors: Guang Zhao, Edward Dougherty, Byung-Jun Yoon, Francis Alexander, Xiaoning Qian
ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate its performance with both synthetic and real-world datasets. |
| Researcher Affiliation | Collaboration | 1Department of Electrical & Computer Engineering, 2Department of Computer Science & Engineering, Texas A&M University College Station, TX 77843, USA 3Computational Science Initiative, Brookhaven National Laborator Upton, NY 11973, USA |
| Pseudocode | Yes | Algorithm 1 Calculation for Weighted-MOCU based Acquisition Function |
| Open Source Code | Yes | The code for our experiments is made available at https://github.com/Qian Lab/WMOCU_AL. |
| Open Datasets | Yes | We also present the results on the UCI User Knowledge dataset (Kahraman et al., 2013). ... We also present the results on the UCI Letter Recognition dataset (Dua & Graff, 2017). |
| Dataset Splits | No | The paper describes train and test sets but does not explicitly mention or detail a validation set split for reproducibility. |
| Hardware Specification | No | The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory amounts, or cloud instance types) used for running its experiments. |
| Software Dependencies | No | The paper mentions that code is available but does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch x.x). |
| Experiment Setup | Yes | In the following experiments, we set c = 1 for the weighted MOCU function. ... We randomly sample 100 particles from the parameter prior with one of the particles as the true model parameter. The five active learning algorithms are compared for 500 iterations... We repeat the simulations for 500 runs. ... We test for 300 runs. ... We have randomly drawn 150 samples from each class as the candidate pool and perform the five different active learning algorithms. We repeat the whole procedure 150 times. ... We randomly take 100 data points first to construct the prior, and use the rest of the data as the pool to test the five active learning algorithms. For prior construction, we train a logistic regression model on the 100 data points and take the trained parameters as the mean of a normal distributed prior with the variance equal to 1. Then we sample 1000 particles from the prior as the uncertain parameter set. We repeat the whole procedure 100 times. |