maze-rlnn - Reinforcement learning with a neuronal network as Q-function in a maze (toy example)

Age	Commit message (Collapse)	Author
2016-02-05	irgendwas mit plots	Florian Jung

2016-02-03	nn3.log	Florian Jung

2016-02-03	proper frameskip for NN	Florian Jung

2016-01-08	dirty hack: do not ever quit early	Florian Jung

2016-01-08	graphs	Florian Jung

2016-01-08	more doc	Florian Jung

2016-01-08	fixed neuronal network approach:	Florian Jung
	- make input values fit into 0..1 range (not 0..5/0..7) - reward of 0.5 instead of 10 - randomize initial weights much more to avoid degeneration
2016-01-08	display range fix	Florian Jung

2016-01-08	--sleep flag	Florian Jung

2016-01-07	Update LICENSE.md	Florian Jung

2016-01-07	QNN stuff, does not work. lowered reward to 0.7	Florian Jung

2016-01-07	fix bug	Florian Jung

2016-01-07	doc	Florian Jung

2016-01-07	episodic batch learning for QArray. probably broke NN.	Florian Jung

2016-01-07	configurable number of episodes	Florian Jung

2016-01-06	episode no-op	Florian Jung

2016-01-06	switchable NN/array	Florian Jung

2016-01-05	reduced neuron number to 40	Florian Jung

2016-01-05	wider output	Florian Jung

2016-01-05	neuronal net. converges a bit. maybe. sometimes. slow.	Florian Jung

2016-01-05	abstract Q array through object	Florian Jung

2016-01-05	cleanup	Florian Jung

2016-01-05	remove policy evaluation	Florian Jung

2016-01-05	reinforcement learning homework	Florian Jung