No activity!

No recent post! It is pitty. I’ll try to solve it. I must do a few fundamental changes for making this blog active.

Ph.C.

Doing a Ph.D., according to this funny articles, can strongly be easen by eating chocolate. I have not tested that in a systematic way (yes! I eat chocolate too!), but I’ll try to do so. If my weight increases in the near future, you will know what the reason is.

Active Learning and Reinforcement Learning

The problem of active learning can be considered as a special case of reinforcement learning (Sanjoy Dasgupta noted it). We can consider it as learning a policy (which selects new data point) that maximizes the increase in some classification performance, e.g. empirical risk, our estimate about structural risk, or anything similar.

@NIPS 2005

Right now, I am in the middle of the NIPS 2005 conference waiting for another oral session to begin.
In this afternoon session, Sanjoy Dasgupta talked about his new result about active learning and the possible merit of it comparing with the traditional supervised case.
He defined something named searchability index which shows the difficulty of the problem in active learning. This index shows that to what extent the ideal binary search-like division of hypothesis space is applicable. If that index is high (or constant over all hypothesis space), the sample complexity of the problem is VC_dim*log(1/eps) (binary division). The other extreme case is VC_dim*(1/eps) (supervised case).
Anyway, you may like to read his paper (I’ll cite it later. I can’t remember the title of the talk.).

I like the idea of active learning. Moreover, I wonder if RL can benefits from it (or vice versa). It seems that active learning strategy is somehow like selecting the exploration strategy.