Contrastive Classification and Representation Learning with Probabilistic Interpretation
Authors: Rahaf Aljundi, Yash Patel, Milan Sulc, Nikolay Chumerin, Daniel Olmeda Reino
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically show that our proposed objective functions show a significant improvement over the standard cross entropy loss with more training stability and robustness in various challenging settings. |
| Researcher Affiliation | Collaboration | Rahaf Aljundi1, Yash Patel2, Milan Sulc 2, Nikolay Chumerin1, Daniel Olmeda Reino1 1 Toyota Motor Europe 2 Visual Recognition Group, Czech Technical University in Prague rahaf.al.jundi@toyota-europe.com |
| Pseudocode | No | The paper describes the methods through mathematical equations and textual explanations but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states 'Supplementary materials can be found in Ar Xiv.' in the abstract, but does not provide a direct link to a code repository or explicitly state that source code for their methodology is released. |
| Open Datasets | Yes | Datasets We consider Cifar-100, Cifar-10 (Krizhevsky and Hinton 2009), Tiny Image Net (Le and Yang 2015) (a subset of 200 classes from Image Net (Deng et ol. 2009), rescaled to the 32 32) datasets and Caltech256 (Griffin, Holub, and Perona 2007). |
| Dataset Splits | No | The paper mentions using training and test sets from standard datasets (e.g., 'Cifar-10', 'Cifar-100') and describes manipulations to the training data (low-sample, imbalanced, noisy) with different N values or imbalance/noise rates. However, it does not explicitly specify the training/validation/test split percentages or sample counts for all datasets needed to reproduce the exact data partitioning. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory amounts used for running the experiments. |
| Software Dependencies | No | The paper mentions using ResNet50 as the main network and details architectural components like projection heads, but it does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | No | The paper states that 'hyper-parameters were estimated on Cifar-10 dataset and fixed for the rest' and describes data scenarios (low-sample, imbalanced, noisy) with rates (e.g., N, IR, NR). However, it does not provide specific hyperparameter values like learning rate, batch size, or optimizer settings in the main text. |