Causal Bandits with Propagating Inference
Authors: Akihiro Yabe, Daisuke Hatano, Hanna Sumita, Shinji Ito, Naonori Kakimura, Takuro Fukunaga, Ken-ichi Kawarabayashi
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5. Experiments We now demonstrate the performance of the proposed algorithm through experimental evaluations and compare it with a baseline algorithm (Audibert & Bubeck, 2010) which was proposed for the general bandit problem and thus cannot take advantage of known causal graph structure. ... Experimental results Figure 4.3(i) shows the average regrets over the synthetic instances against the number of rounds T {C, 2C, . . . , 9C}. Figures 4.3 (ii) and (iii) respectively illustrate the average regrets for the real-world instances constructed from the Alarm and the Water data sets. |
| Researcher Affiliation | Collaboration | 1NEC Corporation, Japan 2RIKEN AIP, Japan 3Tokyo Metropolitan University, Japan 4Keio University, Japan 5National Institute of Informatics, Japan. |
| Pseudocode | Yes | Algorithm 1 Estimation of β ... Algorithm 2 Estimation of α ... Algorithm 3 Causal Bandit |
| Open Source Code | No | The paper refers to the full version and an arXiv preprint for proofs, but does not include an explicit statement about releasing the source code for the described methodology or a direct link to a code repository. |
| Open Datasets | Yes | In the realworld instances, the DAG G is constructed from the Alarm and the Water data sets in a Bayesian Network Repository1. ... 1http://www.cs.huji.ac.il/ galel/ Repository/ |
| Dataset Splits | No | The paper discusses the total number of experimental rounds (T) and how these rounds are distributed across the algorithm's phases (e.g., T/3 for the first phase, 2T/3 for the second phase), but it does not specify any dataset splits for training, validation, or testing in terms of percentages, sample counts, or methodology. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU models, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies, such as programming languages, libraries, or solvers with their respective version numbers, that would be needed to replicate the experiments. |
| Experiment Setup | No | The paper describes parameters internal to the algorithm's phases (e.g., λ, C) and implementation modifications (e.g., setting ηA), but it does not provide specific experimental setup details such as hyperparameter values (learning rate, batch size, epochs), optimizer settings, or other system-level training configurations typically found in machine learning experiments. |