사용자의 입력을 받아올 수 있는 있는 도구
Kinect
LeapMotion
Tilt Sensor
Input : sensor: game controller WebCam GPS
Chuck
Processing
Processing/ Decision Making: The computer decides what action to take
Output: Sound, synthesis, Game, Animation, etc
Wekinator
Nearest Neighbor
K-Nearest Neighbor
Decision stamp
Decision Tree
learning capacity and task
computers numerical value inputs
output Changes smoothly as the inputs char
Nearest - neighbor classifier
Euclidian distance
K Nearest Neighbor
- No training time Capable of arbitrarily complex boundaries
Decision stump
occam's razor
Decision tree entropy
Naive Bayes = 조건부 확률
Traning running can get information about how
Adaboost - Less likely to overfit
Decision stump - Very unlikely to overfit
SVM - Support Vector Machine - make higher dimensional and cut, if it is linear, make a bend to 3d spaces and cut the top. Very powerful several kernel types computation on new example is fast. 1. kernel projects data into higher Dimensional space Linear (least complex)
Audio
MFCCs
FFT
ConstantQ
Maximilian Marsyas
Dynamic Time DTW
Warping
fit the start and end Time and divide value into minimal size
Smoothing
Filtering with average sensors value
naive Bayes decision stump Nearest neighbor
HMM Hidden Markov Model
GVF
Weka toolkit
Data Mining
Practical Machine
Learning Tools and Tech
Pattern Recognition and Machine
- Classification
- Regression
- Temporal Modeling
Conference or Group or Resources
New InterFaces for Musical Expression
International Society of Music Information
International Symposium on Electronic Art
ACM multimedia CHI
ACM Cretivity & Cognition
Association for Computation Creativity
Unsupervised - make a cluster
Computational Creativity
Experiments in musical Intelligence <David cope>