Feature and Parameter Selection in Stochastic Linear Bandits

Authors: Ahmadreza Moradipari, Berkay Turan, Yasin Abbasi-Yadkori, Mahnoosh Alizadeh, Mohammad Ghavamzadeh

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the performances of our FS-SCB and PS-OFUL algorithms using a synthetic LB problem and image classification problems: MNIST (Le Cun et al., 1998) and CIFAR10 (Krizhevsky et al., 2009). We report the details of our experimental setup and additional results in Appendix F.
Researcher Affiliation Collaboration 1Department of Electrical and Computer Engineering, University of California, Santa Barbara, USA 2Deep Mind, London, UK 3Google Research, Mountain View, USA.
Pseudocode Yes Algorithm 1 Feature Selection Square-CB (FS-SCB) Algorithm 2 Parameter Selection OFUL (PS-OFUL) Algorithm 3 Sequential Prediction with Expert Advice
Open Source Code No The paper does not include any explicit statements about releasing source code or provide links to a code repository for the described methodology.
Open Datasets Yes We evaluate the performances of our FS-SCB and PS-OFUL algorithms using a synthetic LB problem and image classification problems: MNIST (Le Cun et al., 1998) and CIFAR10 (Krizhevsky et al., 2009).
Dataset Splits No The paper specifies training and test set sizes for MNIST and CIFAR-100 (e.g., "MNIST dataset consists of 60000 training and 10000 test images..."), but it does not explicitly mention or detail a separate validation split or its methodology.
Hardware Specification No The paper does not provide any specific details about the hardware used for running experiments, such as GPU models, CPU types, or memory configurations. It only mentions general activities like "We train a convolutional neural network (CNN)."
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks like PyTorch, TensorFlow, scikit-learn, etc.) used for implementation or experimentation.
Experiment Setup Yes We train a convolutional neural network (CNN) with M different number of epochs on MNIST data... At each round t [T], the learner is given an action set consist of 10 numbers from A = {1, 2, . . . , 10, 000}. The reward of each action a is φ1(a), θ + ηt, where ηt U[ 0.5, 0.5]. ... The action set in each round t [T] consists of 10 vectors {φt(aj)}10 j=1 N(0, 0.01Id), and the reward of the selected action at is defined as φt(at), θ + ηt, ηt U[ 0.5, 0.5].