Jump to content
  • Advertisement

phi

Member
  • Content Count

    95
  • Joined

  • Last visited

Community Reputation

122 Neutral

About phi

  • Rank
    Member
  1. Ah, hopefully I haven't waste my time! lol. I'm looking at around 1000-2000 potential 'shops' and also would like to take into account past actions. The transition matrix would be huge if I had Village --> Shop, House --> Shop, Shop --> Forest etc.. (i.e. from every potential location to every other potential location). Would there be a way of being able to 'backtrack' through the past N states? My plan was to try and merge all information (including past locations) into one number so that it would be efficient to calculate any methods. From this number, I would be able to put it into a reverse function and derive one state previously and so one.
  2. Hi, I'm trying to monitor user's habitual patterns in game and so I thought the best way to go about it is to try and compress any data into a number for each state. I'm trying to create an adaptive user interface so that it's context sensitive as well as customised to the user themselves. Therefore, if the user often accesses the "buy" menu after he walks into a village, I can decide whether to preload the 'buy' menu or automatically display it on screen. I'm hoping this would a less intrusive UI. My original plan was to take all the factors (Location, Lives etc..) and try and convert them into a number. But maybe I'm going about this all wrong :/ ?
  3. Hi, Sorry for the late reply, I was unable to access the internet (but I survived! :D). For the time being, ignore any of my constraints mentioned in my other posts. What I'm trying to do is to have a large number of 'states' and compare their similarities to a given state. So for example, a state consists of 'Lives, Time Played and Level'(although there will be a few more factors besdies these three). Another state would consist of the same thing and so I'd like to compare their similarities. The slightly more confusing bit comes into a chain of states. I'd like the next state in the chain to incorporate the information from the previous state. To clarify: (read in columns) 1)State X1 = 10,30,50 1)State Y1 = 30,40,10 State Z1 = 1,2,3 2)State Z1 = 32,43,56 2)State Z1 = 36,16,75 As you can see, State Z1 would normally be 1,2,3 if it was the initial state. However, when preceded by something else, it changes. This way, I can then compare a chain of states rather than individual ones. In case a chain to extremely long, I'd rather not iterate through each one and compare. Hope that's a bit clearer (although it's quite late, so I may be confusing things further lol) Thanks
  4. lol I may not have explained it properly, although I think you've got it. I initially had the same idea as you did where I could just literally append everything together, but I felt this may not work given the nature of the task. Essentially what I'm trying to achieve is the ability to "encode" a certain state into a number. Similar states would have similar numbers. A state is something that is made up of various inputs (Time, Lives, Health etc..). To simplify this processing I'm going to convert all the inputs into numerical values. However, things get complicated when a previous state is also taken into account. E.G. A previous state acts as an input to the new state which means that a certain number will be produced even if all other inputs are the same. I'll take a look at hashing. I need to ensure that the hash function doesn't generate a random number using artibtary salts as this would not allow for similarity between states. Thanks again for the help.
  5. Hi, Thanks for the reply. 1) The ID's do not have to be unique (although it's preferable that fewer are to provide greater granularity) 2) For the time being, input sizes will be the same size for each input, but different from each input. e.g: Input 1 will always be 4 digits long Input 2 will always be 6 digits long Input 3 will always be 3 digits long etc..
  6. Wow, thanks for the quick reply! The Hamming Distance seems like a good method of finding the 'distance' between the outputs, but how would I generate the outputs to use?
  7. Hi everyone, I may be overcomplicating things, but how would I go about taking X number of inputs and combining them all to create an 'output ID'? The key thing about the ID is that it should take all the inputs into account (preferably with weighting such as having X1 slightly more important than X2) and produce a number where number can be used to compare similarity. For example: A) Inputs 1,2,3 ---> <perform function> ---> 0123456 B) Inputs 1,2,4 ---> <perform function> ---> 0123457 C) Inputs 5,3,8 ---> <perform function> ---> 1354565 Notice A and B are similar, so the outputs are similar to each other than A/B and C. Thank you for any help [Edited by - phi on July 24, 2009 5:30:16 PM]
  8. phi

    song signature

    Hi, I managed to find out about the MFCCs of a song. Apparently, they're quite good representations of a song's signature. http://en.wikipedia.org/wiki/Mel-frequency_cepstral_coefficient They won't necessarily identify the song, but it can give similarities between songs. Thanks for the help
  9. Hi, I don't know if this belongs here or the maths forums. I'm trying to create a recurring self organizing map. I've created a normal SOM, which works fine. However, in order to make it recurrent, I need to add a leaky integrator in the competition step. I understand the most of the theory of this (I'm not much of a mathematician), but I don't know how to convert that into programming. Does anyone have experience with RSOM's and psuedo code for how it should be set up. In the normal SOM, the competition step finds the distance (Euclidean) between the codebook node and the input vector, but how do I extend this to an RSOM. The leaky integrator formula is: Yi(n) = (1-alpha)Yi(n-1) + alpha(x(n) - Wi(n)) alpha is the learning rate and Yi is the difference vector. Any help would be appreciated. Thanks
  10. phi

    song signature

    Hi Kylotan, I agree that site was interesting but I feel there could be more aspects than just the frequency. I was playing a game called Audio Surf and it was really interesting how the levels were developed. I was wondering if there are other aspects, or more over, how the frequency data could be used in order to extract even more information.
  11. phi

    song signature

    Hello.What type of quantitative information makes up a song? I'm trying to analyse songs and create a list of numerical data (or signature) of the song. What information should I look? Thanks
  12. Hi Kylotan, I'm trying to achieve something similar to this: http://www.brains-n-brawn.com/default.aspx?vDir=aisommusic in terms of the frequency. Thank you.
  13. Hi everyone, I'm having some trouble with FMOD (C#) and am hoping someone can help me. I'm new to the world of audio processing and such, so I may be getting some terms wrong. I have a bunch of songs (mp3s) and would like to obtain the frequency data for each song without having to run the song in realtime. I'm using the Channel.getSpectrum() function which seems fine. However, my values are very small (0.00xxxxx) and are slightly different at times, even when the song is the same. I'm assuming that's down to the exact time the song is scanned. Is it possible to obtain this information without playing the sound? I've tried using system.setOutput(FMOD.OUTPUTTYPE.NOSOUND_NRT);, but it doesn't seem to have any effect as I still need the sound to be playing. Please let me know if you require any clarification. Thank you
  14. Continuing from the end of this thread: http://www.gamedev.net/community/forums/topic.asp?topic_id=499991&whichpage=1� I've been looking into Markov chains and understand some of the maths and probability side. However, I've noticed in many papers I've been reading that they mention Dynamic Bayesian Networks and sometimes use it synonymously with HMMs and other times as improvements of HMMs. Could someone clarify the relationship between DBNs and HMMs? I'd like to start getting some basic code done with HMMs but I'm not sure if I should be looking into DBNs instead, as they seem to be more efficient (from what I've read).
  15. It's finally clicked! :D Thanks for the help. I'm sure I'll be back lol, but I think I'm on some sort of path for now to do some own work. I'm at Nottingham Uni, which I just realised is in your hometown!
  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!