摘要：
阅读全文
摘要：
skip over 阅读全文
摘要：
... 阅读全文
摘要：
阅读全文
摘要：
阅读全文
摘要：
jump over this lecture 阅读全文
摘要：
... 阅读全文
摘要：
... 阅读全文
摘要：
after the break, we'll extend our IRL into continuous spaces 阅读全文
摘要：
make compromise between learnt policy and minimal cost！ π hat is using states π theta is using observations 阅读全文
摘要：
MPC means replan every step Every N step, rebuild the dynamic model 阅读全文
摘要：
transition possibility is unknown and we even don't need to estimate the possibility 阅读全文
摘要：
understand that correlated samples cause problem. and how paralled solve the problem another solution is replay buffers, fully ultilizing the advantag 阅读全文
摘要：
in most AC algorithms, we actually just fit value function. less common to fit Q function as well. batch：off line， monte carlo。online： bootstrap，TD in 阅读全文
摘要：
green bar is the reward function, blue curve is the possibility of differenct trajectories if green bars are equally increased to yellow bars, the res 阅读全文
摘要：
first order markov chain on policy algorithm is easier to be paralleled off policy algorithm has to fit transition net, and policy net. much more comp 阅读全文
摘要：
前面弄错了，应该看2017的秋季课，结果看了春季课了。 neural network control a virtual robot, by imitating human motion Domain shift cause the failure of supervised learning in 阅读全文
摘要：
initialization dramatically influences the trajectory. the current state depends on all the past decision. ones reflect the dimensions being counted. 阅读全文