Other Index 3: The Multistable System

1. █ Divider: 3
2. █ Section Title: The Multistable System
3. █ (Much of this section probably refers to the polystable - not distinguished till 5800)
4. █ In a multistable system with bounds changing at random, a line's endurance is inversley proportional to its length, 2011.
5. █ Marked effect on multistable system of a small change in level of activation, 2002.
6. █ The Traffic Principle, 1981.
7. █ The principle of "minimal bounds" 2064.
8. █ A factor which tends to lessen interaction between lines in a multistable system, 2485.
9. █ Replicated arcs cannot be corrected individually by the environment - 2541 - but if sample combinations of arcs are used, there is a tendency to correction, 2647.
10. █ Experimental method of splitting an environment into parts, 2851.
11. █ In the multistable system each arc does not test itself against its environment but the whole against the whole, 2889.
12. █ Example, in perception, of the multistable system adapting in parts, 2891.
13. █ Skinner's experiments showing how arcs interact, 3008.
14. █ Discussion of how the multistable system adapts, 3014.
15. █ Human beings like information to be highly redundant: does this link with replication in multistable systems? 2993.
16. █ Tendency of a multistable system to develop cycles, with a calculation 3063.
17. █ Adaptive behaviour, in the multistable system, is built out of scrap; it is the environment that enforces the perfection. 3015.
18. █ Length of line of behaviour and chance that a step function will change, 3294.
19. █ Law of the Invariance of Distribution 3418.
20. █ In a multistable system, counting the number of resting states is not practically possible, 3418.
21. █ Some of the step-functions can be used only once, so the system will tend gradually to lose versatility, 3436.
22. █ The multistable system should react sucessfully to the probabilities in its environment, 3478.
23. █ Multistable system as searcher for linear methods, 3511.
24. █ Multistable system responding to 'signal', 3520.
25. █ Some step-functions should be hardly accessible, 3531.
26. █ Number of variables necessary for multi-stability, 3637.
27. █ Polya's methods for solving problems correspond in detail with those of the multistable system, 3817.
28. █ Example of how a multiplicity of tests, each of low reliability, can give a result of high reliability, 3819.
29. █ The multistable system's problem is to get its forgetting right! 3860.
30. █ It may happen that part talks to part through the environment 3924.
31. █ Contained in the multistable system should be all mathematics, 3926.
32. █ Properties of each part depend more and more, as time goes on, what happens to it rather that on which initial state it started from. 3957.
33. █ Several examples from animal psychology of apparent multistability. { 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 }
34. █ The struggle between subsystems as 'war' 3454.
35. █ Multistable system will tend to avoid activating variables that ramify widely or magnify much, 3479.
36. █ The subsystems should gradually develop highly specialised means for each others' disruption 3451.
37. █ The multistable system must progress in adaptation by additive increments, like the genes. 3016.
38. █ The deliberate injection of noise into a multistable system may improve its performance in some specific learning. 3019.
39. █ Crystals as multistable system, 3546.
40. █ Self-reproducing arcs are of high importance, for good or evil 3932.
41. █ Eddies resemble the subsystems of a multistable system. 3438.
42. █ Behaviour of systems having a state from which the transition probability is very small 3291.
43. █ Probability of an absolute system whose variables are statistics of a much larger system, 3476.
44. █ Sub-systems of multistable system viewed topologically, 4007.
45. █ Multistable system tends to have terminal fields with little activation, 4011.
46. █ In the multistable system, essential variables must be mulitple if adaptation is to be progressive, 4015.
47. █ Lashley supports the statistical view, 4027.
48. █ Perhaps arcs in the cortex breed, 4054.
49. █ The intelligent adult is locked in this state 4068.
50. █ Random transformation used by card player - quotation, 4076.
51. █ Selective correction, 4077.
52. █ I do not argue that there must be Q's: I say that if there are responding Q's then they will be important. 4110.
53. █ Mongraph on Evolution, { 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 } .
54. █ By selecting from progressive classes, the intensity can be raised exponentially 4139.
55. █ For really intense selection (a) thermal noise or cosmic radiation must be used as source (b) select by progressive stages. 4140.
56. █ If a quality Q is to be self-reproducing the basic system must be fairly homogenous 4154.
57. █ Adaptation can use either step-function or basins, 4161.
58. █ Multistable systems does not need, but may advantageously have step-functions. 4162.
59. █ Essential variables must have a wide channel to the source of variety, 4167.
60. █ Separate clusters will tend to adapt in that way which gives minimal interaction, 4178.
61. █ Channel capacity in multistable systems for rapid adaptation. 4182.
62. █ Lashley shows that most of the cells of the cortex must be active most of the time. 4188.
63. █ Discriminative feedback of the multistable system. 4189, 4194.
64. █ Relation to Darwinian evolution reviewed 4198.
65. █ Communication between part and part can occur thru the environment, 4215.
66. █ Comparison with selection for egg-laying 4216.
67. █ Relation of multistable system as defined in Design, to the system with environmental variables separated. 4225.
68. █ Conditions for a system to be self-locking, 4229.
69. █ Multistable system will first stabilise two of the null-functions, 4233.
70. █ Why the multistable system takes invarients as fundamental, 4233.
71. █ A system with null-functions, if observed on a new set of dials, no longer shows null-functions (Ashby 3757).
72. █ The present-day, advanced, multistable system probably keeps melting down its adaptations, 4240.
73. █ Factors affecting speed of adaptation, 4246.
74. █ Mutistable system insufficiently general, as it has no levels, 4263.
75. █ Levels between essential variables. 4284.
76. █ An evolving system requires long term memories as well as short, 4374.
77. █ The rank of the differential matrix is unaltered by changing to new variables that are linear functions of the old. Ashby 3768.
78. █ Dispersion changes variety in value to variety in position. 4418.
79. █ Trials within trials are an important advanced method 4546.
80. █ The theory dependencies is that of distinguishabilities 4559.
81. █ In a system of part-functions, all infinitesimal displacements, in whatever direction, from a resting state activate the same set of variables, 3581, 3599.
82. █ Arcs activated by infinitesimal displacements from a resting state - all displacements activate the same arcs 3600.
83. █ The breaking of a whole into subsystems by part-functions is not invariant for change of coordinates. 3757.
84. █ Primary operation needs special definition when applied to a system with part-functions, 3761.
85. █ The theory of games as a chapter in that of the multistable system. 4592.
86. █ The multistable system resembles the Darwinian because both share large numbers of common theorums in the much larger theory of systems. 4620.
87. █ Parts of multistable system can adapt simultaneously provided channel width is adequate and a disperser is provided. 4641.
88. █ If the environment is not constrained, cortex can do nothing better than search at random 4649.
89. █ In the multistable system there must be parts not activated in each reaction. 4831.
90. █ Set theory may play a part in the workings of the multistable system. 4839.
91. █ Multistable system developing the structure of the world around it. 4899.
92. █ The number of arcs that a reaction activates; distribution and average, 3634.
93. █ Number of reactions that will use a given arc; distribution and average, 3634.
94. █ Dispersion proved necessary, 3851.
95. █ Dispersion depends on 'neutral' as well as on the active variables, 3459.
96. █ Dispersion must first occur in a memory-free region if one reaction is not to upset another. 2988.
97. █ The different 'arcs' of the multistable system should traverse the environment by different routes. 3140.
98. █ The most efficient way is to make the environment show the specific step-functions that are to be altered 2970.
99. █ Relation of the multistable system to its environments. 3082, 3026.
100. █ Discriminative feedback from essential variables. 2958.
101. █ Average run necessary to reach terminal set when, on average, m variables are activated and give chance of (½)^m, is e^m/2. So m can hardly exceed 6-10.
102. █ Systems of part-functions, to be stable, must have intrinsic stability. 2978.
103. █ A system of part-functions with k active is as difficult to stabilise as one of k full-functions 3022.
104. █ "To accumulate memories that are not affected by intervening events" is equivalent to "There must be small interaction between the events and the memories" 5304.
105. █ A factor making for fast adaptation 5314.
106. █ There is no general rule for how the feedback should be made discrimminating. Only when a particular machine is given can the question proceed to detail. 5336.
107. █ The multistable system has a bias towards becoming efficient and quickly adapting. 5342.
108. █ Multistability, i.e. separation of learning parts, proved necessary 5345. (Better proof than in Design for a Brain S.17/2)
109. █ A system that adapts by correcting multiple arcs moves to an equilibrial "percentage adapted" - for the more there are adapted, the greater is the chance that correction of one unadapted will upset the adapted, 5360.
110. █ How much should discriminative feedback spread? - As wide as the spread of the errors. 5369.
111. █ Destructive feedback may at first, in the history of the multistable system, be scattered widely but should steadily be scattered in narrower range. 5371.
112. █ Keeping reactions apart by giving them room to spread is costly in material and space. 5380.
113. █ If a change of step-mechanism (due to corrective feedback) makes an arc change place, more space is required. If the change makes an arc change nature but not place, more time is required. 5388.
114. █ (Some concepts and processes demand only that the system be "plastic".)
115. █ A system that accumulates adaptations must use discrimination in its distribution of corrective feedback. 5410.
116. █ Multistable system and gene-pattern both demand DIAGRAM 5421.
117. █ How should the multistable system be didvided? - So that each irreducible "Good" has its own ultrastable system. 5436.
118. █ Fundamental rule for the multistable system: if the major "Good" is to be obtained by the accumulation of minor "goods", then each minor good should have an ultrastable system to itself. 5436.
119. █ What in the environment are the special characteristics that the multistable system is specially adapted to? 5582.
120. █ Accumulation of adaptations 5592.
121. █ Newer theory of multistable system. 5601 et seq.
122. █ The multistable system is the natural specialisation to the disturbance that comes rarely but repetitively. 5618.
123. █ (Brief review of position) 5619.
124. █ I now (1957) regarded as a special case in a more general formulation. 5621.5.
125. █ One unit for adaptation (one "arc") may comprise many portions of both cortex and environment. 5634.3.
126. █ Taylor's experiments with reversing spectacles showing the disorderliness of the multistable system. 5697.
127. █ Multistable system, as defined in "Design for a Brain", is assumed to have its feedbacks discriminating. 5735.
128. █ Old definition of multistable system reviewed ten years later. 5735.
129. █ Multistable system re-specified to separate the aspect of :- (1) Having many almost independent essential variables. (2) Having SFs (step functions) separated into sets. 5775.
130. █ Multistable system distinguished from polystable. 5800.
131. █ Polystable system finds constraint automatically(?) 6372.