Category Archives: Paper

Paper: Binary Action Search for Learning Continuous-Action Control Policies

NOTES -“Binary Action Search” BAS search the whole action range by increment and decrement the action-values for an “internal binary policy” -> what is that policy BAS eliminates restrictive modification steps of “Adaptive Action Modification” basement: – Least-Squares Policy Iteration … Continue reading

Posted in Paper | Leave a comment

Paper: Speeding up RRL

RRL and TG/RRL-TG algorithm – RRL:combination of “inductive logic programming” and “reinforcement learning” – improvement: combined RRL with TG – improvement(small): using query-packs mechanism to lower the needed storage orignal RRL algorithm standard RL-algorithm -> add to knowledge base prediction … Continue reading

Posted in Paper | Leave a comment

Paper: Mining High-Speed Data Streams

VFDT and Hoeffing bound VFDT is for nearly infinite data streams and shows best performace after 100k samples(in this paper) VFDT is noise resistant and performs nearly equivilant on noise-free data. Hoeffding tree algorithm garantees near/equal trees as batch learner’s … Continue reading

Posted in Paper | Leave a comment