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.
Differentiable Distributionally Robust Optimization Layers
Authors: Xutao Ma, Chao Ning, Wenli Du
ICML 2024 | Venue PDF | LLM Run Details | Input Tokens: 26,310 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 3,566 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 | To validate the effectiveness of the proposed differentiable DRO layers, we conduct experiments on a toy example and the portfolio management problem. |
| Researcher Affiliation | Academia | 1Department of Automation, Shanghai Jiao Tong University, Shanghai 200240, China 2The Key Laboratory of Smart Manufacturing in Energy Chemical Process, Ministry of Education, East China University of Science and Technology, Shanghai 200237, China. |
| Pseudocode | No | The paper describes algorithmic steps in narrative text, but it does not include any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Source code of all the experiments is available at https://github.com/DOCU-Lab/Differentiable DRO Layers. |
| Open Datasets | No | In generating data, we first sample covariate z, and then sample y conditioned on z. ... In the multi-item newsvendor problem (19) with n = 4, we set ac = (0.25, 0.5, 0.75, 1), ad = 0.95 ac, v = (10.0, 13.0, 16.0, 19.0), b = (2.0, 4.0, 6.0, 8.0), d = (0.5, 1.0, 1.5, 2.0). The covariate z follows the uniform distribution on [0, 1], and conditioned on z, we set the distribution of demand y by... |
| Dataset Splits | Yes | We use 2,500 data for training, 500 data for validation, and 1,000 data for testing. ... In each case, the size of the validation data set is set to N/5. |
| Hardware Specification | Yes | The experiments are conducted on a laptop with an i7 CPU and 32G RAM. |
| Software Dependencies | No | Processes 2 and 4 are built on Cvxpylayers (Agrawal et al., 2019a), and Process 3 is built on commercial solver Gurobi. The paper mentions these software tools but does not provide specific version numbers for them. |
| Experiment Setup | Yes | We use 2,500 data for training, 500 data for validation, and 1,000 data for testing. ... In the gradient estimation, we solve the DRO 3 times to construct (15) and use 4 samples to estimate the gradient term by importance sampling. For the energy parameter λ in (10), we initially set it to 10, and subsequently reduce it by one-third every 30 epochs. ... We use the following two-layer full-connected neural network to learn the ambiguity set parameter in both our method and the method proposed in Costa & Iyengar (2023). FC(5, 22) FC(22, 27) FC(27, m) |