Saffari et al. propose an online random forest algorithm (see  for the original paper introducing random forests) based on online bagging . Note the difference between on-line and incremental algorithms: While incremental algorithms have memory, that is they may store each incoming sample, an online algorithm sees each sample exactly once. On common datasets, as for example the USPS Dataset, Saffari et al. show that online random forests converge to the offline algorithm.
The implementation is currently under development, however, may be made publicly available within the next few weeks.
What is your opinion on the summarized work? Or do you know related work that is of interest? Let me know your thoughts in the comments below: