ml.lib nime 2015 slides
TRANSCRIPT
ml.lib: Robust, Cross-platform, Open-source Machine Learning for Max and Pure Data
Jamie Bullock Associate Professor of Music Technology
Birmingham City University
Ali Momeni Associate Professor of Art Carnegie Mellon University
Cont, A. (2008). Antescofo: Anticipatory Synchronization and Control of Interactive Parameters in Computer Music (pp. 33–40). Presented at the International Computer Music Conference, Belfast, Ireland: Ann Arbor, MI: Scholarly Publishing Office, University of Michigan Library.
Deyle, T., Palinko, S., & Poole, E. S. (2007). Hambone: A bio-acoustic gesture interface. … Computers. http://doi.org/10.1109/ISWC.2007.4373768
Sato, M., Poupyrev, I., & Harrison, C. (2012). Touché: enhancing touch interaction on humans, screens, liquids, and everyday objects. Chi, 483–492. http://doi.org/10.1145/2207676.2207743
Ono, M., Shizuki, B., & Tanaka, J. (2013). Touch & activate (pp. 31–40). Presented at the the 26th annual ACM symposium, New York, New York, USA: ACM Press. http://doi.org/10.1145/2501988.2501989
Poupyrev, Ivan, et al. "Botanicus Interacticus: interactive plants technology." ACM SIGGRAPH 2012 Emerging Technologies. ACM, 2012.
Nielsen,Usability Engineering (1993)
Syst
em a
ccep
tabi
lity
Social acceptability
Cost Compa-
tibility
Relia-bility
Etc.
Utility
Easy to learn
Efficient to use
Easy to remember
Few errors
Subjectively pleasing
Practical accepta-bility
Usefulness
Usability
. Raskin, The Humane Interface (2000)
• modeless: user actions should have the same effect regardless of the application’s state
addtrainmap
<class> <values …>3 .1 .4 .1 .5 .9 .2 .6
<values …>.1 .4 .1 .5 .9 .2 .6 3
recordManually segment time-varying input vectors
Dlibmlpack
Shark
• Efficient • Wide range of algorithms • Well supported • Good documentation
GRT
libsvm + others
GRT + FLEXT
=ml.lib
+ UCD
ClassificationAdaptive Boosting
Adaptive Naive BayesBootstrap Aggregator
Decision Trees Dynamic Time Warping
Finite State Machine Gaussian Mixture ModelHidden Markov Modelk-Nearest Neighbour
Linear Discriminant AnalysisMinimum DistanceParticle Classifier Random Forests
Support Vector Machines
Regression
Artificial Neural NetworkLinear Regression
Logistic RegressionMultidimensional Regression
Regression Tree
Feature Extraction
Peak DetectionMin / Max
Zero Crossings
Architecture
live
offline
algorithmtrainingvector
in-memorymodel
“add”
storedmodel
“train”
storeddata
“map”
“read / write”
in-memorydata
inputvector
outputvalue
Common Attributesprobs <0/1>
scaling <0/1>
Object-specific Attributesrandomize_training_order <0/1>
mode <0/1>
num_outputs <1..N>
Live Demo
Swept Frequency Sensing
Future Work
• Documentation!!! • Sort out HMMs • Implement GRT clustering algorithms • Possible threaded “train” • Maybe more feature extraction, LibXtract?
Thank You!