Federated Contextual Cascading Bandits with Asynchronous Communication and Heterogeneous Users
Authors: Hantao Yang, Xutong Liu, Zhiyong Wang, Hong Xie, John C. S. Lui, Defu Lian, Enhong Chen
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical Evaluation. Finally, we conduct experiments on both synthetic and real data and show the effectiveness of the user clustering procedure and the pt-auxiliary communication, where our Fed C3UCB-H achieves superior performance regarding regrets and communication cost. |
| Researcher Affiliation | Academia | Hantao Yang1, Xutong Liu2*, Zhiyong Wang2, Hong Xie1, John C. S. Lui2, Defu Lian1, Enhong Chen1 1University of Science and Technology of China 2The Chinese University of Hong Kong |
| Pseudocode | Yes | Algorithm 1: Fed C3UCB-H; Algorithm 2: Local Agent(t, ut, pt, αc); Algorithm 3: Server(t, ut, αd) |
| Open Source Code | No | The paper does not include any explicit statement or link indicating that the source code for the methodology is openly available. |
| Open Datasets | Yes | We conduct experiments on the Yelp dataset, which contains 4.7 million ratings of 1.57 105 restaurants from 1.18 million users.3 http://www.yelp.com/dataset challenge |
| Dataset Splits | No | The paper describes the datasets used (synthetic and Yelp) and how they were processed, but it does not provide specific details on how the data was split into training, validation, and test sets. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU/CPU models, processor types, or memory used for running the experiments. |
| Software Dependencies | No | The paper mentions general techniques like k-means clustering and SVD, but it does not specify any particular software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, specific libraries with versions). |
| Experiment Setup | Yes | Synthetic Experiments. We conduct experiments in a synthetic environment with |U| = 40 users, J = 5 or J = 10 clusters with orthogonal θ, and T = 1000, 000 rounds. The preference and feature vectors are of dimension d = 20, with each entry drawn from a standard Gaussian distribution, and are normalized to vectors with . 2 = 1 (Li et al. 2019). At round t, a user ut randomly chosen from U and and It = 200 items are generated, where each item is associated with a random xt,i Rd generated as above. The agent needs to recommend K = 4 items as cascading arms at to the user, leading to a random reward f(at, wt) as defined in Eq. (3). The agents then observed feedback according to Eq. (1). Experiments on Real Dataset. ...Then we randomly sample |U| = 40 users and use k-means clustering (Ahmed, Seraj, and Islam 2020) to generate J = 10 user clusters. Each cluster s center vector represents the true preference vector for users within that cluster. The experiment s remaining parameters match those of the synthetic experiments. |