Synthetic Causal Contextual Bandit Dataset

该数据集是为了评估因果上下文老虎机算法的性能而生成的。它包含了基于一个已知的因果图的多重中间上下文和干预措施。在特定的干预措施下，奖励按照伯努利分布（0.5 + ε）来分配，而在其他情况下则按照伯努利分布（0.5）来分配，其中ε的值为0.3。该数据集的规模为：25个中间上下文和25个变量。所面临的任务是因果上下文老虎机探索。

This dataset was generated to evaluate the performance of causal contextual bandit algorithms. It contains multiple intermediate contexts and interventions grounded in a known causal graph. Under specific interventions, rewards follow a Bernoulli distribution with parameter (0.5 + ε), while under other conditions, they follow a Bernoulli distribution with parameter 0.5, where ε is set to 0.3. The dataset consists of 25 intermediate contexts and 25 variables. The targeted task is causal contextual bandit exploration.