linucb

Yahoo! news article recommendation system by linUCB

Multi Armed Bandits implementation using the Yahoo! Front Page Today Module User Click Log Dataset

Contextual bandit algorithm called LinUCB / Linear Upper Confidence Bounds as proposed by Li, Langford and Schapire

Predict and recommend the news articles, user is most likely to click in real time.

Bandit algorithms

A comprehensive Python library implementing a variety of contextual and non-contextual multi-armed bandit algorithms—including LinUCB, Epsilon-Greedy, Upper Confidence Bound (UCB), Thompson Sampling, KernelUCB, NeuralLinearBandit, and DecisionTreeBandit—designed for reinforcement learning applications

Some visualizations of bandit algorithm outputs.

Reinforcement learning models for warfarin dose online estimation

LinUCB with disjoint linear models

python implementation of e-Greedy, UCB, LinUCB, LinThompson, and offline evaluator

A Contextual-bandit approach on MIND Datasets for News Recommendation Systems.

Implementation of the Multi-Armed Bandit where each arm returns continuous numerical rewards. Covers Epsilon-Greedy, UCB1, and Thompson Sampling with detailed explanations.

RL-based DBMS Buffer Pool Auto-Tuning for Optimal and Economical Memory Utilization. Thus, controls memory footprint of a relational database

Recommendation using LinUCB algorithm

Research implementation of a focused web crawler that combines adaptive Q-learning, LinUCB contextual bandits, and graph neural network embeddings for topic-directed crawl decisions.

Reproducibility artifact for distributed LinUCB-based adaptive DB-LBT in Wi-Fi/NR-U coexistence

Data Mining course at ETH Zürich.

A collection of implementations of the bandit problem.

Learning project: fraud scoring on the public ULB card-fraud data (time split, leakage-safe velocity/target encoding, GBDT), approval-vs-fraud-loss thresholds, LinUCB acquirer routing on SIMULATED acquirer outcomes, FastAPI scorer with measured p95. No real IDs or acquirer data.

Contextual bandit router that learns online which QA strategy (retrieval, lookup, or calculator) to use per query, from reward alone. LinUCB implemented from scratch, benchmarked against random, static, and oracle baselines.