Decision-Focused Learning without Decision-Making: Learning Locally Optimized Decision Losses
Authors: Sanket Shah, Kai Wang, Bryan Wilder, Andrew Perrault, Milind Tambe
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our approach on three resource allocation problems from the literature and find that our approach outperforms learning without taking into account task-structure in all three domains, and even hand-crafted surrogates from the literature. and 5 Experiments To validate the efficacy of our approach, we run experiments on three resource allocation tasks from the literature. |
| Researcher Affiliation | Academia | Sanket Shah Harvard University sanketshah@g.harvard.edu Kai Wang Harvard University kaiwang@g.harvard.edu Bryan Wilder Carnegie Mellon University bwilder@andrew.cmu.edu Andrew Perrault The Ohio State University perrault.17@osu.edu Milind Tambe Harvard University milind tambe@harvard.edu |
| Pseudocode | No | No pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | Yes | Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] As supplemental material. |
| Open Datasets | Yes | true CTRs ym from the Yahoo! Webscope Dataset [28] and historical data from 2004 to 2017 for a set of N = 50 stocks from the Quandl WIKI dataset [22]. |
| Dataset Splits | No | The paper mentions 'More experimental setup details are provided in Appendix A.' and in the checklist, 'Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] In the appendix'. However, the appendix content is not provided in the given text, so specific split percentages or sample counts for validation are not explicitly stated within the main paper text. |
| Hardware Specification | No | The paper mentions that hardware details are in Appendix A: 'Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [Yes] In the appendix'. However, Appendix A is not provided in the given text, and the main body does not contain specific GPU/CPU models, processor types, or detailed computer specifications used for experiments. |
| Software Dependencies | No | The paper indicates that training details are in Appendix A: 'Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] In the appendix'. However, the appendix content is not provided in the given text, and the main paper text does not list specific software dependencies with version numbers (e.g., library names like PyTorch, TensorFlow, or Python versions). |
| Experiment Setup | Yes | We train either a linear model (for the Linear Model domain) or a 2-layer fully-connected neural network with 500 hidden units (for the other domains) using LODLs and compare it to:... and We train each LODL for 100 gradient descent steps using 5000 samples and train the predictive model for 500 steps (the same as the setup as Table 1). and λ = 0.1 is the risk aversion constant. |