devlog (15/06): neural network plays 東方永夜抄 ～ Imperishable Night
second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can \
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.
devlog (15/06): neural network plays 東方永夜抄 ～ Imperishable Night
second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can \
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.
devlog (15/06): neural network plays 東方永夜抄 ～ Imperishable Night
second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can \
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.
devlog (15/06): neural network plays 東方永夜抄 ～ Imperishable Night
second devlog of a neural network playing Touhou, though now playing the second stage of Imperishable Night with 8 players (lives). the NN can \
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.
currently, yes, but this is more an investigation into how well a neural network could play a bullet hell game
very few bullet hell AI programs rely on machine learning and virtually all of the popular ones use algorithms.
but it is interesting to see how it mimics human behaviour, skills and strategies and how different methods of machine learning perform and why
(plus I understand machine learning more than the theory behind those bullet hell bots.)
the training environment is pretty basic right now so all bullets shoot from the top of the screen with no enemy to destroy.
additionally, the program I'm using to get player and bullet data (twinject) doesn't support enemy detection so the neural network wouldn't be able to see enemies in an existing bullet hell game. the character used has a wide bullet spread and honing bullets so the neural network inadvertently destroys the enemies on screen.
the time spent in each training session is constant rather than dependent on survival time because the scoring system is based on the total bullet distance only.
definitely. usually algorithms are used to calculate the difficulty of a game (eg. in osu!, a rhythm game) so there's definitely a practical application there
one problem ive seen with these game ai projects is that you have to constantly tweak it and reset training because it eventually ends up in a loop of bad habits and doesnt progress
you're correct that this is a recurring problem with a lot of machine learning projects, but this is more a problem with some evolutionary algorithms (simulating evolution to create better-performing neural networks) where the randomness of evolution usually leads to unintended behaviour and an eventual lack of progression, while this project instead uses deep Q-learning.
the neural network is scored based on its total distance between every bullet. so while the neural network doesn't perform well in-game, it does actually score very good (better than me in most attempts).
so is it even possible to complete such a project with this kind of approach as it seems to take too much time to get anywhere without insane server farms?
the vast majority of these kind of projects - including mine - aren't created to solve a problem. they just investigate the potential of such an algorithm as a learning experience and for others to learn off of.
the only practical applications for this project would be to replace the "CPU" in 2 player bullet hell games and maybe to automatically gauge a game's difficulty and programs already exist to play bullet hell games automatically so the application is quite limited.
the body of the post has the ringtone attached. I might need to edit it to make it viewable through Photon but you can also view it on a browser
Triple Baka ringtone
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.
I always find it interesting to see how optimization algorithms play games and to see how their habits can change how we would approach the game.
me too! there aren't many attempts at machine learning in this type of game so I wasn't really sure what to expect.
Humans would usually try to find the safest area on the screen and leave generous amounts of space in their dodges, whereas the AI here seems happy to make minimal motions and cut dodges as closely as possible.
yeah, the NN did this as well in the training environment. most likely it just doesn't understand these tactics as well as it could so it's less aware of (and therefore more comfortable) to make smaller, more riskier dodges.
I also wonder if the AI has any concept of time or ability to predict the future.
this was one of its main weaknesses. the timespan of the input and output data are both 0.1 seconds - meaning it sees 0.1 seconds into the past to perform moves for 0.1 seconds into the future - and that amount of time is only really suitable for quick, last-minute dodges, not complex sequences of moves to dodge several bullets at a time.
If not, I imagine it could get cornered easily if it dodges into an area where all of its escape routes are about to get closed off.
the method used to input data meant it couldn't see the bounds of the game window so it does frequently corner itself. I am working on a different method that prevents this issue, luckily.
I did create a music NN and started coding an UNO NN, but apart from that, no
yeah, the training environment was a basic bullet hell "game" (really just bullets being fired at the player and at random directions) to teach the neural network basic bullet dodging skills
- the white dot with 2 surrounding squares is the player and the red dots are bullets
- the data input from the environment is at the top-left and the confidence levels for each key (green = pressed) are at the bottom-left
- the scoring system is basically the total of all bullet distances
- this was one of the training sessions
- the fitness does improve but stops improving pretty quickly
- the increase in validation error (while training error decreased) is indicated overfitting
- it's kinda hard to explain here but basically the neural network performs well with the training data it is trained with but doesn't perform well with training data it isn't (which it should also be good at)
training a neural network to play a bullet hell game
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! >
training a neural network to play a bullet hell game
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! >
training a neural network to play a bullet hell game
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! >
training a neural network to play a bullet hell game
- 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!
lol I understand the feeling
thanks for this!! there's so much info on this comment
i'm currently using Logseq w/ Syncthing but i'll be looking at Org Mode and DokuWiki
they're all really good! senya is a really good singer
let's go
favourite opening: OP 4 favourite ending: either the one from EP 14 or Opposite World
this is the ad-free version, which is available with the exact same (if I'm correct) features on F-Droid for free, along with the source code on GitHub.
the versions on the Play Store (paid version and free version with ads) likely just help pay the developer for their work
and as others have said already, free software is free as in freedom, not free beer.
browse (Old) Reddit with Lemmy style
don't know if this is the correct community to post this on, but is there a way to style Reddit (most likely old Reddit but it doesn't matter) with a similar style to Lemmy? there are a few small subreddits I still browse (because there is no Lemmy equivalent) and it would be cool if I could style it like the regular Lemmy style (if you need an example: lemmy.ml)
I know there are frontends for viewing Lemmy instances like Reddit, but I haven't heard of any way to do the reverse.
it could be through a stylesheet, an alternative frontend, even just a pointer on how you could style a website into a different style. \ thanks!
[HELP] represent PC speaker beep with audio file
for a while, I have been using hardware through Linux that uses the beep from the PC speaker. I'm actually really used to it, so when I switched to using hardware with an unusably loud (volume can't be changed) volume (and also different frequency), I started looking into "exporting" the original beep to an audio file that could be played at different volumes and for other purposes.
looking through the internet, however, I haven't found any attempts to represent any actual PC speaker beep in an audio file, so I'm asking you guys if you know how to do so. presumably, the beep is just a short, simple waveform at a certain frequency, but I am not sure what that waveform is, or what the easiest way to do so is.
[IDEA] self-hosted advertisements for personal sites
so I was looking at someone's personal website from Mastodon, and noticed that they had banners to advertise other people's servers. while server lists like fediring exist, I was thinking of a more automatic method of advertisement within someone's website.
the concept is this: people could store advertisements (small banners, gifs) on their websites with a server and people willing to embed them could use an API to retrieve a random ad onto their website.
people would self-host their ads and "federate" with other websites to embed other ads on their website. not sure if this would scale up as well, though.
what do you think? just curious on lemmy's POV
edit: going by the comments, this idea is quite flawed and webrings (in small sizes) are a better approach.
thanks for the help