Notice: This website is continuously updated as new conference proceedings and journal volumes are published. Consequently, the figures presented here may differ from those reported in our publications.
A brief overview of our methodology is provided on the Methods page, along with references to publications documenting our approach and links to the code and data used in each study.
The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates.
DICE: Diversity in Deep Ensembles via Conditional Redundancy Adversarial Estimation
Authors: Alexandre Rame, Matthieu Cord
ICLR 2021 | Venue PDF | LLM Run Details | Input Tokens: 38,146 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 4,910 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We obtain state-of-the-art accuracy results on CIFAR-10/100: for example, an ensemble of 5 networks trained with DICE matches an ensemble of 7 networks trained independently. We further analyze the consequences on calibration, uncertainty estimation, out-of-distribution detection and online co-distillation. In this section, we present our experimental results on the CIFAR-10 and CIFAR-100 (Krizhevsky et al., 2009) datasets. |
| Researcher Affiliation | Collaboration | Alexandre Rame Sorbonne Universit e Paris, France EMAIL Matthieu Cord Sorbonne Universit e & valeo.ai Paris, France EMAIL |
| Pseudocode | Yes | Algorithm 1: Full DICE Procedure for M = 2 members |
| Open Source Code | No | The paper states: 'We borrowed the evaluation code from https://github.com/ uoguelph-mlrg/confidence_estimation (De Vries & Taylor, 2018).' It does not provide an explicit statement or link for the open-source code of their own proposed method (DICE). |
| Open Datasets | Yes | In this section, we present our experimental results on the CIFAR-10 and CIFAR-100 (Krizhevsky et al., 2009) datasets. |
| Dataset Splits | Yes | Hyperparameters for adversarial training and information bottleneck were ο¬ne-tuned on a validation dataset made of 5% of the training dataset, see Appendix D.1. For hyperparameter selection and ablation studies, we train on 95% of the training dataset, and analyze performances on the validation dataset made of the remaining 5%. |
| Hardware Specification | No | This work was granted access to the HPC resources of IDRIS under the allocation 20XXAD011011953 made by GENCI. |
| Software Dependencies | No | The paper mentions general software components like 'Res Net' and 'Wide Res Net' architectures and optimization algorithms like 'SGD', but it does not specify version numbers for programming languages or libraries (e.g., Python, PyTorch, TensorFlow). |
| Experiment Setup | Yes | Following (Chen et al., 2020b), we used SGD with Nesterov with momentum of 0.9, mini-batch size of 128, weight decay of 5e-4, 300 epochs, a standard learning rate scheduler that sets values {0.1, 0.001, 0.0001} at steps {0, 150, 225} for CIFAR-10/100. log(Ξ²ceb) reaches values {100, 10, 2, 1.5, 1} at steps {0, 8, 175, 250, 300}. |