This is taking a long time

2. CES302 July 14, 2011 10:15am

6. Spent the last ten years in teaching and research, particularly on the development of the Starcat framework, here at SDSU

7. Continuing this while also experiencing “research migration”

9. Population-based techniques (some more than others, e.g. genetic algorithms) map nicely into such an approach

10. My longstanding commitment to certain cognitive models for emergent and adaptive computation are inherently both distributed and parallel

12. Mike also responded with interest to my suggestion that my own recent move into the world of research with GPGPU hardware could result in a joint effort to produce a similar kind of framework to SPEEDES for that rapidly growing technique

13. *(caveat): I was unable to follow up with Mike very quickly due to administrative obligations of our new certificate program; John stepped in to facilitate, so that here we are today (thanks!)

15. ...which generated unexpected excitement for me as I have been ruminating on both distributed and parallel possibilities for that architecture—indeed have been developing a more close-to-the-silicon version of Starcat in support of mult-agent simulations leveraging the hundreds of parallel cores in the Nvidia Tesla board (GPU)

16. ...but which may have as much or more chance of success in a CPU based mature framework such as SPEEDES.

17. The ultimate goal would be to produce self-organizing behavior in a computational system that adapts to a changing environment and learns new behavior in a manner more cognitively plausible than much of the body of work in traditional machine learning (Hawkins for exp.)

19. Developing an analog of SPEEDES for leveraging GPGPU hardware is a completely independent line of research we could begin as soon as feasible; likely there are sources of support for something of this nature—or at least a clear need (and thus client base) for that framework (I resisted the urge to invent a cute cousin acronym for it)

20. A subsequent slide will present some very preliminary thoughts regarding the feasibility of such a “port” of the SPEEDES product to this different hardware configuration

22. It is FULL of possibilities of interest to everyone from the DOD to, for example, the Neurosciences Institute here in SD

23. I've been wanting to bring the two ideas together for a couple of years now

24. I'd love to get involved in such a collaboration! Publications guaranteed. Follow-on funding and applications too...

25. Again, following a bit later are some slides to motivate some technical discussion about adaptive computation and the Starcat framework

29. String on finger, book by door, pheromones .

30. Internal stigmergy occurs in a complex system when changes in internal dynamics persist and have influence on future action.

31. When that persistence is closely coupled with particular interactions with the environment, the changes can be called emergent representations of the phenomona from the environment.

33. Some have gone so far as to eschew it completely (Brooks, 1991).

34. But a representation that emerges as a side-effect of the interaction of very many agents with the environment and each other, guided by long-trained system dynamics, is another matter.

35. Partial sufficient representation—no more need to maintain fidelity of a model.

36. Such a representation might also be leveraged externally, as a usable “product” of the system’s processes

38. So the system’s own “trail” of behavior plays a role in guiding the system’s future actions.

39. Coupled with the fact that global behavior arises from the interaction of very many locally acting agents, we have the possibility of adaptation of behavior.

40. It only calls for us to define the microbehaviors that are possible and the concepts which motivate them.

42. facilitating the development of open-ended systems

46. Bottom-up pressures from perceptual input and context

47. Fluid concepts that shift under these pressures

48. Flexible representations, situated through behavior, that emerge from the interactions of very many simple agents

51. Arbitrary connectivity among and choice of components

52. Data structure / Metabolism model for components

53. Success-based and failure-based slipnet activation

54. Genericity of workspace structures

55. Broader spectrum of codelet release

56. Original biological metaphor for codelets amplified

57. Elaborated modulation of workspace access patterns

58. XML configuration of slipnet and component parameters

60. Workspace as spatial array captures physical constraints

61. Ants, pheromones, food are basic objects

62. Codelets (MoveAnt, EvaporatePheromone,…) imbue workspace structures with dynamics

63. Stigmergic behavior provides optimization “solution” to familiar asymmetric bridge experiment and in general finding paths from nest to food

64. Explores notion of stigmergy as learning

66. Copycat was one of the first systems to realize the idea of emergent representation and was a predecessor to Starcat

67. Want to show the new system has similar capabilities

68. Problem domain is analogies between letter-string pairs

69. Initial results here; a complete redevelopment using a refined Starcat architecture and extending the Copycat capabilities is underway

71. We want to explore the interactions between multiple instantiations of the framework

72. We have an ALife application where each individual agent has a slipnet and a coderack for regulating its behavior and they all interact in a shared workspace

73. Patterns of interaction change depending on the internal states of the many slipnets

74. Fear, hunger and contentedness nodes rise and fall in activation in response to environment, driving individual agent’s behaviors (exploring, fleeing, attacking, eating)

75. Hunter and prey agents (both lazy and active varieties) have slipnets with different biases and generated codelets

77. Lazy prey agents had the peak lifespan in a single experiment

78. Lazy hunters moved dramatically less than any other agent (scavenging)

79. Active prey agents were attacked more than lazy ones (increased interaction)

80. Lazy prey moved more than any agent type on average, across all experiments (fear, hunger and longer lifespans)

81. Scavenging in general (more eating than movement/attacking)

82. Changing the slipnets affects these tendencies

83. Tuning space huge and true emergent effects not clearly seen…an issue to consider

85. Madcat was the first system to embody the emergent representation and fluid concept network of Copycat in ongoing interaction with a problem domain

86. Madcat’s development motivated Starcat and its premises

87. Madcat highlighted the possibility of autopoiesis in a computational system and the use of partial, sufficient representation

88. Want to show the new system has similar capabilities and explore additional features

92. Performs analysis of musical concepts at multiple levels of abstraction (notes, phrases, etc.)

93. Provides ability to ‘seed’ the system with different types of musical styles

94. Workspace is time-driven (similar to Madcat) with a ‘window’ of analysis

95. It can be influenced by past structures but only build new structures from present and beyond

96. Workspace has subdivisions for melody track and chord progressions

97. Has the potential for real-time interaction with performing musician

98. Rather dense slipnet with core of basic musical concepts and smaller set of style-specific concepts

99. Phrasing and melodic shaping are emergent properties of the system

100. One of the more successful Starcat implementations

102. Integrates a genetic algorithm into the framework.

103. The slipnet links and link lengths are determined by this GA.

104. Populations of 50-100 Botcat agents, initially randomly configured, are evolved over 1500 generations.

105. This ability to train the network with a GA has been a goal of the project since the beginning.

106. We also want to provide an evolutionary process over the development of codelets.

107. Together these might bring about something like learning, provided the GA is always running, similar to the classifier system.

108. Botcat experimented with different configurations of the framework and with new types of codelets (fuzzy and composite).

109. Demonstrated two interesting truly “emergent” phenomena: ramming and raking, to cope with the underlying Robocode frameworks evaluator patterns

111. We want this network to self-organize its configuration (who is talking to who, who is monitoring the signal, how to collaborate for increased confidence, etc.)

112. Subject to various constraints (orientations, available power, etc.)

113. Use Starcat as a fine-grained implementation of an artificial immune system model

117. Growing/evolving rather than programming

118. Slipnet is core (and hardest and exploratory)

120. … the principles of complex systems at play

121. New issues arise with each effort

123. Autonomy arises because the system simply produces ongoing behavior in response to a changing environment without external intervention