Sorry, I am lazy and won't read back how the conversation started. I just want to vent an idea (?) Have you tested how bayesian inference would work suggesting playlists ? Basically: a. db based - This playlist is all 60's pop, next pleasing song could be pop of the same era (score: 5/10). b. behavior based - Last 2 times we added the song I was ready to suggest, it got skipped from this player (1/10). Another one maybe (5/10) - Most times we played the last song in the playlist on this player, near this time of day, it was chained with another one, and this is not a pop song (score: 10/10)
=> a+b : add the not pop song 1st, and 60's pop song as a safer fallback in 2nd position. (then record what happens: skip or play ) Well you get the idea. Just like mail filters, but continuously learning. Since we have a server, in the cloud further, recording bits of information 24/7 seems not a problem. That's a lot of maths, so I never went beyond wondering if this would work in real life (which clues would be actionable.) ------------------------------------------------------------------------ epoch1970's Profile: http://forums.slimdevices.com/member.php?userid=16711 View this thread: http://forums.slimdevices.com/showthread.php?t=98467
_______________________________________________ discuss mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/discuss
