zolax

cross-posted from: https://programming.dev/post/15553031

second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can "see" the whole iwndow rather than just the neighbouring entities.

comment from video:

the main issue with inputting game data relatively was how tricky it was to get the NN to recognise the bounds of the window which lead to it regularly trying to move out of the bounds of the game. an absolute view of the game has mostly fixed this issue.

the NN does generally perform better now; it is able to move its way through bullet patterns (01:38) and at one point in testing was able to stream - moving slowly while many honing bullets move in your direction.

cross-posted from: https://programming.dev/post/15553031

second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can "see" the whole iwndow rather than just the neighbouring entities.

comment from video:

the main issue with inputting game data relatively was how tricky it was to get the NN to recognise the bounds of the window which lead to it regularly trying to move out of the bounds of the game. an absolute view of the game has mostly fixed this issue.

the NN does generally perform better now; it is able to move its way through bullet patterns (01:38) and at one point in testing was able to stream - moving slowly while many honing bullets move in your direction.

cross-posted from: https://programming.dev/post/15553031

second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can "see" the whole iwndow rather than just the neighbouring entities.

comment from video:

the main issue with inputting game data relatively was how tricky it was to get the NN to recognise the bounds of the window which lead to it regularly trying to move out of the bounds of the game. an absolute view of the game has mostly fixed this issue.

the NN does generally perform better now; it is able to move its way through bullet patterns (01:38) and at one point in testing was able to stream - moving slowly while many honing bullets move in your direction.

second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can "see" the whole iwndow rather than just the neighbouring entities.

comment from video:

the main issue with inputting game data relatively was how tricky it was to get the NN to recognise the bounds of the window which lead to it regularly trying to move out of the bounds of the game. an absolute view of the game has mostly fixed this issue.

the NN does generally perform better now; it is able to move its way through bullet patterns (01:38) and at one point in testing was able to stream - moving slowly while many honing bullets move in your direction.

a gapless, ringtone version of the song "Triple Baka" from the first few seconds.

reply or DM if you want a loseless download because this upload is compressed.

cross-posted from: https://programming.dev/post/14979173

• neural network is trained with deep Q-learning in its own training environment
• controls the game with twinject

demonstration video of the neural network playing Touhou (Imperishable Night):

it actually makes progress up to the stage boss which is fairly impressive. it performs okay in its training environment but performs poorly in an existing bullet hell game and makes a lot of mistakes.

let me know your thoughts and any questions you have!

cross-posted from: https://programming.dev/post/14979173

• neural network is trained with deep Q-learning in its own training environment
• controls the game with twinject

demonstration video of the neural network playing Touhou (Imperishable Night):

it actually makes progress up to the stage boss which is fairly impressive. it performs okay in its training environment but performs poorly in an existing bullet hell game and makes a lot of mistakes.

let me know your thoughts and any questions you have!

cross-posted from: https://programming.dev/post/14979173

• neural network is trained with deep Q-learning in its own training environment
• controls the game with twinject

demonstration video of the neural network playing Touhou (Imperishable Night):

it actually makes progress up to the stage boss which is fairly impressive. it performs okay in its training environment but performs poorly in an existing bullet hell game and makes a lot of mistakes.

let me know your thoughts and any questions you have!