neural network - Pybrain Reinforcement Learning Example -

as question states looking explanation/example reinforcement learning in pybrain documentation on confuses me no end, can work don't understand how apply other things.

thanks tom

unfortunately, pybrain's documentation rl classes disappointing. have found this blog quite useful.

---

in summary, need identify following components (for implementation details follow tutorial on link):

1. an environment: env = environment(...)
2. a task --> task = task(env)
3. a controller, module (like table) keep action-value information --> controller = module(...)
4. a learner --> learner = sarsa() --> may add explorer learner. default epsilon-greedy epsilon = 0.3, decay = 0.9999.
5. an agent integrate controller , learner --> agent = agent(controller, learner)
6. an experiment integrate task , agent , actual iterations --> experiment = experiment(task, agent)

each of capitalized classes should replaced corresponding class pybrain.then run do-while cycle perform iterations , learn. note there several options set user, , in real-world problems need write sub-classes generalize basic classes of pybrain, steps same here.