Time to threshold

tags
Transfer learning, Reinforcement learning

This is a simple metric first mentioned in (Taylor et al. 2007; Taylor, Stone 2007). In the paper by Taylor Stone and Liu, it is defined as:

Time-to-Threshold: Measure the time needed to reach a performance threshold in the target task.

In other words, this metric measures the time spent to reach a target performance for a given learning system.

To write down this metric, we use a

In these works, the metric is defined for only one threshold value.

Bibliography

  1. . . "Transfer Learning via Inter-task Mappings for Temporal Difference Learning". J. Mach. Learn. Res. 8:2125–67. http://dl.acm.org/citation.cfm?id=1314569.
  2. . . "Cross-domain Transfer for Reinforcement Learning". In Machine Learning, Proceedings of the Twenty-fourth International Conference (ICML 2007), Corvallis, Oregon, USA, June 20-24, 2007, edited by Zoubin Ghahramani, 227:879–86. ACM International Conference Proceeding Series. ACM. DOI.

Comments


← Back to Notes