Prediction with Corrupted Expert Advice
Authors: Idan Amir, Idan Attias, Tomer Koren, Yishay Mansour, Roi Livni
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conducted a basic numerical experiment to illustrate our regret bounds and the gap between OMD and FTRL discussed above. The experiment setup consists of two experts with different gaps {0.05, 0.15, 0.25, 0.4}. The losses were taken as Bernoullis and the corruption strategy injected contamination in the first rounds up to a total budget of C, inflicting maximal loss on the best expert while zeroing the losses of the other expert. The results, shown in Fig. 1, demonstrate that for the stochastic case without corruption (C= 0) OMD achieves better pseudo regret, but is substantially outperformed by FTRL when C> 0. In Fig. 2 we further show the inverse dependence of the pseudo-regret on the minimal gap , which precisely supports our theoretical finding discussed in Section 3.3. |
| Researcher Affiliation | Collaboration | Idan Amir Tel Aviv University idanamir@mail.tau.ac.il; Idan Attias Ben-Gurion University idanatti@post.bgu.ac.il; Tomer Koren Tel Aviv University and Google tkoren@tauex.tau.ac.il; Roi Livni Tel Aviv University rlivni@tauex.tau.ac.il; Yishay Mansour Tel Aviv University and Google mansour.yishay@gmail.com |
| Pseudocode | No | The paper provides algorithmic update rules in equations (3), (4), and (5), but these are not presented in a formal 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | No | The paper does not explicitly state that source code for the described methodology is available, nor does it provide a link to a code repository. |
| Open Datasets | No | The paper describes a synthetic experiment setup where losses were taken as Bernoullis and a corruption strategy was injected. It does not refer to a publicly available dataset with concrete access information. |
| Dataset Splits | No | The paper describes the experiment setup and reports results, but it does not specify training, validation, or test dataset splits or cross-validation methods. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the numerical simulations. |
| Software Dependencies | No | The paper does not specify any software dependencies (e.g., programming languages, libraries, or frameworks) with version numbers used for the experiments. |
| Experiment Setup | Yes | The experiment setup consists of two experts with different gaps {0.05, 0.15, 0.25, 0.4}. The losses were taken as Bernoullis and the corruption strategy injected contamination in the first rounds up to a total budget of C, inflicting maximal loss on the best expert while zeroing the losses of the other expert. |