Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 
 
CONTEMPORARY APPROACHES TO LIVE COMPUTER MUSIC:  
THE EVOLUTION OF THE PERFORMER COMPOSER  
 
 
BY 
 
 
OWEN SKIPPER VALLIS 
 
 
 
 
 
 
 
 
A thesis  
submitted to the Victoria University of 
Wellington  
in fulfillment of the requirements for the degree 
of  
Doctor of Philosophy 
 
 
 
 
 
Victoria University of Wellington 
 
2013

  iii
Supervisory Committee 
 
Dr. Ajay Kapur (New Zealand School of Music) 
Supervisor 
 
Dr. Dugal McKinnon (New Zealand School of Music) 
Co-Supervisor 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
? OWEN VALLIS, 2013 
NEW ZEALAND SCHOOL OF MUSIC 
  iv 
ABSTRACT 
 
This thesis examines contemporary approaches to live computer music, and the 
impact they have on the evolution of the composer performer. How do online 
resources and communities impact the design and creation of new musical 
interfaces used for live computer music? Can we use machine learning to 
augment and extend the expressive potential of a single live musician? How can 
these tools be integrated into ensembles of computer musicians? Given these 
tools, can we understand the computer musician within the traditional context of 
acoustic instrumentalists, or do we require new concepts and taxonomies? Lastly, 
how do audiences perceive and understand these new technologies, and what 
does this mean for the connection between musician and audience? 
The focus of the research presented in this dissertation examines the application 
of current computing technology towards furthering the field of live computer 
music. This field is diverse and rich, with individual live computer musicians 
developing custom instruments and unique modes of performance. This 
diversity leads to the development of new models of performance, and the 
evolution of established approaches to live instrumental music.  
This research was conducted in several parts. The first section examines how 
online communities are iteratively developing interfaces for computer music. 
Several case studies are presented as examples of how online communities are 
helping to drive new developments in musical interface design. 
This thesis also presents research into designing real-time interactive systems 
capable of creating a virtual model of an existing performer, that then allows the 
model?s output to be contextualized by a second performer?s live input. These 
systems allow for a solo live musician?s single action to be multiplied into many 
different, but contextually dependent, actions. 
  v 
Additionally, this thesis looks at contemporary approaches to local networked 
ensembles, the concept of shared social instruments, and the ways in which the 
previously described research can be used in these ensembles.  
The primary contributions of these efforts include (1) the development of 
several new open-source interfaces for live computer music, and the examination 
of the effect that online communities have on the evolution of musical 
interfaces; (2) the development of a novel approach to search based interactive 
musical agents; (3) examining how networked music ensembles can provided 
new forms of shared social instruments. 
  

  vii
ACKNOWLEDGEMENTS 
 
The author wishes to express sincere appreciation to all those who have helped 
with the realization of this thesis. The work that follows is the culmination of the 
last ten years of my life, and was made possible by the many artists, colleagues, 
and friends who have inspired and helped me along the way. 
Ajay Kapur, thank you first and foremost for your exceptional tutelage in not 
only my research, but also my life. I would not be here today if it was not for 
your insight, advice, mentorship, and motivation. You have the amazing gift of 
seeing the potential in those around you, and the genius to help us realize that 
potential.  
Jordan Hochenbaum, thank you for all the projects, music, and art that we have 
worked on over the last few years, for all the incredible projects we will work on 
in the coming years, and most of all for being an incredible friend. Our work as 
Flipmu has been a source of many of the ideas in this thesis, and I look forward 
to all the coding, installation, music, and general madness that are to come. You 
are an incredible musician, amazing artist, and constant source of inspiration. 
Many thanks to Dugal McKinnon, Nick Collins, Michael Norris, Martijn 
Zwartjes, Jim Murphy, Brad Hill, Tim Exley, Jason Edwards, Michael Darling, 
and Johnny McClymont for their help, ideas, and advice along the way. I would 
not have been able to complete this thesis without you sharing your expertise 
and thoughts. 
Much of this thesis is the direct result of my experiences performing with other 
musicians in live computer music ensembles. Thank you to The KarmetiK 
Machine Orchestra; whom?s broad musical vision allowed me to perform both 
with incredible musicians, and a shared robotic instrument. Thank you to 
Trimpin, Curtis Bahn, and Tomie Hahn for sharing your incredible work. Your 
  viii 
ideas have helped to inspire my own efforts. Thank you to Jeffery Lufkin for 
being a part of the genesis for many of the ideas found in this thesis. 
Lastly, thank you to my family for their support throughout this journey. To my 
parents for their encouragement and support; to my incredible and loving wife 
Liv, for being my source of strength and my muse; and to Rory, for being an 
inspiration for my writing, and helping me through many challenging 
discussions, thoughts, and edits.  
For any of my friends who I have most regrettably left out of these 
acknowledgments, please know that you have my deepest gratitude for your help. 
 
 TABLE OF CONTENTS 
CHAPTER 1 INTRODUCTION ................................................................................................. 1

1.1
 INTERACTION CONTEXTS ...................................................................................... 2

1.2
 PERSONAL MOTIVATIONS ...................................................................................... 5

1.3
 THESIS OVERVIEW ................................................................................................... 7

CHAPTER 2 COMMUNITY BASED DESIGN: ITERATIVE MUSICAL INTERFACE 
DEVELOPMENT ...................................................................................................................... 11

2.1
 GOALS AND MOTIVATION .................................................................................... 12

2.2
 INTERFACE DEVELOPMENT PRIOR TO ONLINE COMMUNITIES .................... 13

2.3
 COMMUNITY BASED DESIGN ............................................................................... 15

2.4
 ITERATIVE DEVELOPMENT .................................................................................. 16

2.5
 BACKGROUND CASE STUDIES ON THE MONOME ........................................... 16

2.5.1
 Monome .......................................................................................................... 17

2.5.2
 The computer musican as digital luthier ..................................................... 22

2.6
 NEW WORK: ARDUINOME AND CHRONOME ................................................... 22

2.7
 MONOME, TENORI-ON COMPARISON ............................................................... 28

2.8
 DISCUSSION ............................................................................................................ 29

CHAPTER 3 ARMY OF ME: AUTONOMOUS AGENTS AND THE SOLO PERFORMER .... 33

3.1
 GOALS AND MOTIVATION .................................................................................... 34

3.2
 BACKGROUND ........................................................................................................ 35

3.2.1
 Twentieth century composers ...................................................................... 36

3.2.2
 Computer aided algorithimic composition ................................................ 38

3.2.3
 Interactive musical agents ............................................................................. 40

3.2.4
 Contemporary systems .................................................................................. 41

3.3
 DEFINE THE CHALLENGE .................................................................................... 42

3.4
 IMPLEMENTATION: LIVE PERFORMANCE SYSTEMS ........................................ 43

3.4.1
 Search based systems ..................................................................................... 44

3.4.2
 Considerations for use with improvisation ................................................ 45

3.4.3
 S2MP: a similarity matching algorithm ....................................................... 46

3.4.4
 Training the system, and linking controller data ....................................... 47

3.4.5
 Implementation of the system for use in performance ............................ 49

3.4.6
 Plugin design ................................................................................................... 53

Table of Contents 
 
x 
3.4.7
 Analysis ............................................................................................................ 55

3.4.8
 Challenges with using search based systems .............................................. 64

3.5
 DISCUSSION ............................................................................................................. 65

3.5.1
 Architecture of an interactive musical agent .............................................. 65

3.5.2
 The army of me .............................................................................................. 67

CHAPTER 4 THE ART OF COMMUNICATION: SHARED INSTRUMENTS AND 
NETWORKED MUSICAL ENSEMBLES ................................................................................... 69

4.1
 BACKGROUND ........................................................................................................ 72

4.2
 PHYSICALITY IN COMPUTER MUSIC PEFORMANCE, AND EXTENDING 
SHARED CONTROL TO MUSICAL ROBOTICS ................................................................. 74

4.3
 MUSICAL ROBOTICS AND THE KARMETIK MACHINE ORCHESTRA ............. 76

4.4
 COMPOSITIONS AND PERFORMANCES ............................................................... 78

4.4.1
 January 27, 2010 REDCAT - The Machine Orchestra ............................ 78

4.4.2
 August 14, 2010 ? Karmetik Collective ...................................................... 82

4.4.3
 April 12, 2012 REDCAT ? Samsara The Machine Orchestra ................ 84

4.5
 DISCUSSION ............................................................................................................. 86

CHAPTER 5 CONCLUSION .................................................................................................... 89

5.1.2
 Improvisation in live computer music ........................................................ 93

5.2
 CONTRIBUTIONS .................................................................................................... 95

5.2.1
 Online Community based iterative design and the Chronome ............... 96

5.2.2
 S2MP and an interactive system for continuous control ......................... 97

5.2.3
 Shared social musical robotics ...................................................................... 97

5.3
 FUTURE WORK AND PHILOSOPHY ...................................................................... 98

5.3.1
 Bridging the gap between performer and audience .................................. 99

5.3.2
 Final thoughts ............................................................................................... 102

APPENDIX A RELATED PUBLICATIONS .......................................................................... 105

APPENDIX B CHRONOME TECHNICAL FILES ................................................................. 107

APPENDIX C COMPARATIVE SURVEY OF LOCAL NETWORK ENSEMBLES AND SOLO 
LIVE COMPUTER MUSIC ...................................................................................................... 121

APPENDIX D PROBABILITIES AND MARKOV MODELS ................................................ 131

APPENDIX E SEARCH-BASED ALGORITHMS ................................................................... 145

APPENDIX F REGRESSION SYSTEMS ................................................................................ 151

APPENDIX G COMPARISONS AND REQUIREMENTS ..................................................... 157

Table of Contents 
 
xi 
BIBLIOGRAPHY .................................................................................................................... 167


 LIST OF FIGURES 
  
Figure 1: Tools used in the evolving roles of the performer composer ......................... 7

Figure 2: Overview of the thesis layout ............................................................................... 8

Figure 3: Iterations of the Monome ................................................................................... 17

Figure 4: Commercially developed Native Instruments Maschine, Akai APC 40, 
Novation Launchpad, and Ableton PUSH ............................................................. 21

Figure 5: The first Arduinome. Built using Monome 40h buttons ................................ 24

Figure 6: Chronome prototype ............................................................................................ 26

Figure 7: BrickTable version I, II, and III ......................................................................... 27

Figure 8: Monome left and Tenori-On right ..................................................................... 28

Figure 9: Interactive musical agent for modeling continuous control data from 
musical interfaces ......................................................................................................... 35

 Figure 10: Controller data from both performers stored as 2-D and flattened 1-D 
collections ..................................................................................................................... 47

Figure 11 Search algorithm training algorithm. The two performers are sampled by 
the plugin, and performer B?s current item set is linked with Performer A?s 
previous item set. ......................................................................................................... 48

Figure 12: Stored performances for S2MP algorithm. Sequential item sets for a 
performance are stored as Seq2(n). Each jth item set represents both Performer A 
and B, and has a MIDI buffer associated with it. ................................................... 49

Figure 13: Constraining the S2MP search to current position within a bar. This 
represents a good compromise between efficiency and search as it cuts the 
number of compares and is likely to return good matches. .................................. 50

Figure 14: Mapping score matrix. The intersection is taken between every ith item set 
in Seq1 and the jth item set in the current sub-sequence from Seq2(n). ................. 51

Figure 15: Performance diagram of search algorithm. Sampling Performer A along 
with the current agent output generates the input sequence. This is then 
compared against sub-sequences within the database, providing similarity 
scores. The MIDI buffer is taken from the sub-sequence with the highest 
similarity score, and then appended to the plugin?s MIDI output. ..................... 53

Figure 16: Continuous Control S2MP plugin ................................................................... 54

Figure 17: Parsing MIDI buffers into sample windows .................................................. 54

List of Figures 
 
xiv 
Figure 18: Routing setup for S2MP training in Ableton Live, and MIDI CC training 
data ................................................................................................................................. 56

Figure 19: Target CC sequence (top) vs. Output sequence (bottom). The output 
sequence didn't match the target sequence in the first bar, but otherwise was a 
perfect match. ............................................................................................................... 56

Figure 20: Initial S2MP plugin test - The training sequence length was increased from 
one to four bars. ........................................................................................................... 57

Figure 21: Similarity percentage between output sequence and target sequence based 
on Mapping / Order weighting ................................................................................. 59

Figure 22: Mapping / Order - Average distance of output sequence from the target 
sequence ........................................................................................................................ 60

Figure 23: Factor increase of discontinuities between output and target sequence .... 61

Figure 24: Similarity percentage between output sequence and target sequence based 
on number of item sets in the input sequence, and number of trained transitions
 ........................................................................................................................................ 62

Figure 25: Increasing numbers of item sets - average distance of output sequence 
from the target sequence for input sequences ........................................................ 63

Figure 26: Factor increase of number of discontinuous matches by sequence size .... 64

Figure 27: Different design approaches for interactive musical agents. Clockwise 
from top: (A) All inputs affecting the model?s output; (B) The model is only 
affected by itself, and live input is applied as a fitness function; (C) Inputs are 
split into simpler individual models, all acting independent of each other ......... 66

Figure 28: Network topology of The Machine Orchestra ensemble ............................. 71

Figure 29: The League of Automatic Composers 1980 ................................................... 72

Figure 30: View of the marimbas from the musical robot Tammy ............................... 76

 Figure 31: The Machine Orchestra at REDCAT 2010 ................................................... 79

Figure 32: The Machine Orchestra performing Samsara 2012 ....................................... 85

Figure 33: The Monome can be both highly programmable or immediately usable .. 90

Figure 34 Preference performing solo computer music vs. networked ensembles .. 122

Figure 35: A Markov Model representing the transition probabilities for the set of 
notes C through A. The notes on the left represent the source states, and the 
notes along the top are destination states. The values in the matrix are the 
transition probabilities for moving from a source state to a destination state. 133

Figure 36: Algorithm for determining the next destination state in a Markov Model 
given a row of transition probabilities .................................................................... 134

List of Figures 
 
xv 
Figure 37: 2nd order Markov Model ................................................................................ 135

Figure 38: 2nd order Markov Model shown as 1st order Markov Model .................. 135

Figure 39: The description of a transition probability, containing the Source State, 
Destination State, and the number of times the transition has been observed 
during training ............................................................................................................ 137

Figure 40: Markov source states stored as tree structure. This allows for searching 
variable length source state sequences. .................................................................. 137

Figure 41: Longest matching source state sequence for input sequence F G D C ... 138

Figure 42: Longest matching source state sequence for input sequence G A D C .. 138

Figure 43: Training diagram of Markov Model .............................................................. 141

Figure 44: Performance diagram of Markov Model ...................................................... 142

Figure 45: Input to Markov Model that does not match any previously seen source 
state .............................................................................................................................. 143

Figure 46: Matching sequences return the next stored state form the database ........ 146

Figure 47: Matching sequences return values from a second linked sequence .......... 147

Figure 48: Matrix with ideal rating scores comparing input sequence along the 
columns, and scored sequence down the rows ..................................................... 148

Figure 49: Insertion test: this tests for notes played by the performer that are not in 
the original score ........................................................................................................ 149

Figure 50: Deletion test: this tests for notes from the score that are skipped over by 
the performer ............................................................................................................. 149

Figure 51: Substitution test: this tests for notes substituted by the performer .......... 149

Figure 52: Repetition test: this looks for notes from the score that are repeated by the 
performer .................................................................................................................... 149

Figure 53: Constrained search to speed up sequence comparison .............................. 150

Figure 54: Basic linear regression ...................................................................................... 152

Figure 55 Similarity between linear regression and basic sequential ANN ................ 152

Figure 56 Diagram of multi-variate linear regression model ........................................ 154


 LIST OF TABLES 
Table 1 Reasons for solo or group performance preference ........................................ 124

Table 2 Descriptions of each musician's roll within the ensemble .............................. 125

Table 3 Description of new performance modes afforded by computer ensembles 127

Table 4 Descriptions of the challenges of performing in computer ensembles ........ 129

Table 5 Overview of algorithms for designing interactive musical agents ................. 164