Towards Better Selective Classification
Authors: Leo Feng, Mohamed Osama Ahmed, Hossein Hajimirsadeghi, Amir H. Abdi
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The results suggest that the superior performance of state-of-the-art methods is owed to training a more generalizable classifier rather than their proposed selection mechanisms. Our proposed selection mechanism with the proposed entropy-based regularizer achieves new state-of-the-art results. 5 EXPERIMENTS For the following experiments, we evaluate the following state-of-the-art methods (1) Selective Net (SN), (2) Self-Adaptive Training (SAT), and (3) Deep Gamblers. |
| Researcher Affiliation | Collaboration | Leo Feng Mila Université de Montréal & Borealis AI leo.feng@mila.quebec Mohamed Osama Ahmed Borealis AI mohamed.o.ahmed@borealisai.com Hossein Hajimirsadeghi Borealis AI hossein.hajimirsadeghi@borealisai.com Amir Abdi Borealis AI amir.abdi@borealisai.com |
| Pseudocode | No | The paper does not contain any sections explicitly labeled 'Pseudocode' or 'Algorithm', nor are there any structured code-like blocks. |
| Open Source Code | Yes | Our code is available at https://github.com/Borealis AI/towards-better-sel-cls. |
| Open Datasets | Yes | We introduce new datasets: Stanford Cars, Food101, Imagenet, Imagenet100 and Imagenet Subset, for the selective classification problem setting and benchmark the existing state-of-the-art methods. Imagenet (Deng et al., 2009) Food101. The Food dataset (Bossard et al., 2014) Stanford Cars. The Cars dataset (Krause et al., 2013) CIFAR-10. The CIFAR-10 dataset (Krizhevsky, 2009) |
| Dataset Splits | Yes | For hyperparameter tuning, we split Imagenet100 s training data into 80% training data and 20% validation data evenly across the different classes. |
| Hardware Specification | Yes | The experiments were primarily run on a GTX 1080 Ti. |
| Software Dependencies | No | The paper mentions 'Pytorch implementation' but does not specify a version number for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | For hyperparameter tuning, we split Imagenet100 s training data into 80% training data and 20% validation data evenly across the different classes. We tested the following values for the entropy minimization coefficient β {0.1, 0.01, 0.001, 0.0001}. ... Self-Adaptive Training models are trained using SGD with an initial learning rate of 0.1 and a momentum of 0.9. Food101/Imagenet100/Imagenet Subset. The models were trained for 500 epochs with a mini-batch size of 128. The learning rate was reduced by 0.5 every 25 epochs. The entropy-minimization term was β = 0.01. |