Active Learning Through a Covering Lens

Authors: Ofer Yehuda, Avihu Dekel, Guy Hacohen, Daphna Weinshall

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conclude with extensive experiments, evaluating Prob Cover in the low-budget regime. We show that our principled active learning strategy improves the state-of-the-art in the low-budget regime in several image recognition benchmarks. ... In Section 4 we empirically evaluate the performance of Prob Cover on several computer vision datasets, including CIFAR-10, CIFAR-100, Tiny-Image Net, Image Net and its subsets.
Researcher Affiliation Academia Ofer Yehuda , Avihu Dekel , Guy Hacohen , Daphna Weinshall School of Computer Science & Engineering Edmond and Lily Safra Center for Brain Sciences The Hebrew University of Jerusalem Jerusalem 91904, Israel {ofer.yehuda,avihu.dekel,guy.hacohen,daphna}@mail.huji.ac.il
Pseudocode Yes Algorithm 1 Prob Cover
Open Source Code Yes Code is available at https://github.com/avihu111/Typi Clust.
Open Datasets Yes We empirically evaluate the performance of Prob Cover on several computer vision datasets, including CIFAR-10, CIFAR-100, Tiny-Image Net, Image Net and its subsets. ... When considering CIFAR-10/100 and Tiny Image Net, we use as input the embedding of Sim CLR [9] across all methods. When considering Image Net we use as input the embedding of DINO [5] throughout.
Dataset Splits No The paper mentions running experiments for a fixed number of active learning rounds and reporting mean test accuracy, but it does not specify explicit training, validation, and testing dataset splits (e.g., 80/10/10) for the entire dataset used in the experiments. It focuses on how samples are queried and added to the labeled set over rounds.
Hardware Specification No The paper does not explicitly specify the hardware used for the experiments, such as GPU models, CPU types, or memory. In the checklist at the end, it states: '(d) Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [No]'
Software Dependencies No The paper mentions using specific models and frameworks like 'Res Net-18', 'Sim CLR', 'DINO', and 'Flex Match', and an 'evaluation kit created by Munjal et al. [33]'. However, it does not provide specific version numbers for any of these software dependencies or underlying libraries (e.g., Python, PyTorch versions), which is required for reproducibility.
Experiment Setup Yes Details concerning specific networks and hyper-parameters can be found in App. C, and in the attached code in the supplementary material.