Competing against Adaptive Strategies in Online Learning via Hints
Competing against Adaptive Strategies in Online Learning via Hints
复制标题
DOI：
--
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Aditya Bhaskara;Kamesh Munagala
中科院分区：
文献类型：
--
作者：
Aditya Bhaskara;Kamesh Munagala
对于许多经典的在线学习环境，众所周知，在预测之前，对损失功能有“提示”，从而在这项工作中可以保证Sig-nifore sig-ni，这是否使我们超越了提示。遗憾的标准概念（与最佳固定策略竞争），并与自适应或动态策略竞争，如果提示是完美的，我们可以与完全动态的策略竞争。转换后悔的下限，即，算法所产生的损失与最佳的最佳策略在最多l时间切换状态，其中l是一些参数。经典的专家问题。
For many of the classic online learning settings, it is known that having a “hint” about the loss function before making a prediction yields sig-nificantly better regret guarantees. In this work we study the question, do hints allow us to go beyond the standard notion of regret (which competes against the best fixed strategy) and compete against adaptive or dynamic strategies? After all, if hints were perfect, we can clearly compete against a fully dynamic strategy. For some common online learning settings, we provide upper and lower bounds for the switching regret, i.e., the difference between the loss incurred by the algorithm and the optimal strategy in hind-sight that switches state at most L times, where L is some parameter. We show positive results for online linear optimization and the classic experts problem. Interestingly, such results turn out to be impossible for the classic bandit setting.