FIRAL: An Active Learning Algorithm for Multinomial Logistic Regression
Authors: Youguang Chen, George Biros
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments on synthetic datasets. Furthermore, we compare FIRAL with five other methods and found that our scheme outperforms them: it consistently produces the smallest classification error in the multiclass logistic regression setting, as demonstrated through experiments on MNIST, CIFAR-10, and 50-class Image Net. |
| Researcher Affiliation | Academia | Youguang Chen George Biros Oden Institute for Computational Engineering and Sciences The University of Texas at Austin |
| Pseudocode | Yes | Algorithm 1 FIRAL(b, U, S0, 0) |
| Open Source Code | No | The paper does not provide any statement about releasing source code or a link to a code repository for the methodology described. |
| Open Datasets | Yes | We demonstrate the effectiveness of our active learning algorithm using three real-world datasets: MNIST [18], CIFAR-10 [19], and Image Net [20]. |
| Dataset Splits | No | The paper mentions initializing labeled data and evaluating classification accuracy on the unlabeled pool, but it does not specify explicit training, validation, and test splits (e.g., percentages, sample counts, or explicit standard splits for reproducibility) for the datasets used. |
| Hardware Specification | No | The paper mentions 'Computing time on the Texas Advanced Computing Centers Stampede system was provided by an allocation from TACC and the NSF,' which names a system but does not provide specific hardware details such as GPU/CPU models, memory, or processor types. |
| Software Dependencies | No | The paper discusses various algorithms and methods but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, or specific libraries with their versions). |
| Experiment Setup | Yes | For MNIST, we calculate the normalized Laplacian of the training data and use the spectral subspace of the 20 smallest eigenvalues. For CIFAR-10 and Image Net-50, we use a contrastive learning Sim CLR model [21]; then we compute the normalized nearest-neighbor Laplacian and select the subspace of the 20 smallest eigenvalues; For Image Net-50 we select the subspace of the 40 smallest eigenvalues. For each dataset, we initialize the labeled data S0 by randomly selecting one sample from each class. Further details about tuning hyperparameter and data pre-processing are given in Appendix G.2. |