2018-01-30 20:24:32 +01:00
|
|
|
this program is largely incomplete and not in a presentable state.
|
2018-01-30 19:52:40 +01:00
|
|
|
please be mindful when sharing it.
|
2018-01-30 20:24:32 +01:00
|
|
|
however, feel free to copy any snippets of code you find useful.
|
|
|
|
|
|
|
|
|
|
TODOs: (that i can remember right now)
|
2018-03-26 16:32:00 +02:00
|
|
|
- finish implementing ARS
|
|
|
|
|
'-> running mean/std normalization of all inputs (aka ARS V2)
|
|
|
|
|
'-> normalize and/or embed sprite inputs
|
2018-01-30 20:24:32 +01:00
|
|
|
- settle on a network architecture
|
|
|
|
|
- fix lag-frames skipped-inputs bug
|
2018-03-23 09:33:02 +01:00
|
|
|
- detect frames when Mario is in a controllable state
|
2018-01-30 20:24:32 +01:00
|
|
|
- fix offscreen sprites sometimes being visible to network
|
|
|
|
|
- add some detection for enemies later in the game
|
2018-03-23 09:33:02 +01:00
|
|
|
- compute how many input neurons the network needs instead of hardcoding
|
2018-01-30 20:24:32 +01:00
|
|
|
|
|
|
|
|
naive:
|
|
|
|
|
- learn any combination of buttons, starting from title screen
|
|
|
|
|
- learn to run network without frameskip
|
|
|
|
|
- learn other games
|
