Page  311 ï~~NEURAL NETWORK DIFFERENT THAT LI MUSIC Christiane Linster TU Graz Austria Abstract Based on the idea that listening to music in' rhythm structures, we motivate and present preferences of different musical styles. We di; structure and show how the encoding problem network's elements has been solved in this partic The system can be presented with an arbitrarya of the styles the system has learned during the t

Page  312 ï~~J -I "I o JJ'.Nh 40 k / 74 flI n7 0=-f J no' isar I oil% e1 J Jn /1 k, 5"J 0 4a JS 1 S Fig Two different grouping structures for A neural network will find the solution which is closest the training phase. The idea to design a system that learns to play experience. Encoding What information should the elements of the r signal, which gives us information about discrete mo describes: the presence of a musical event, its- amplitu

Page  313 ï~~A more compact architecture requires additio network structure is trained with all the training patter into the context units. (Fig 4) OUTPUT PATE,C op U N ITFS rs Â~ I V a a a a a I m El \ Fig 4: Network arc Conclusion We have designed a neural network structure rhythm analysis and Derformance of simple melodies