Blackboard temperature

Blackboard's temperature has two roles:

1. it measures the perceptual organization of the current solution in the Blackboard. It is a rough indication of the value of the answer given by BAsCET (the lower the temperature value, the better the answer).
2. it controls the randomness used in each decision (the higher the temperature value, the less deterministic the decision).

Randomness is used for the choice of the agents in order to give to BAsCET a "creativity" power: a same treatment, ran twice on the same problem won't always yield twice the same solution, imitating the human behaviour. A human being does not react equally when he is calm or angry, hasty or not... According to his state of mind, he won't solve problems exactly the same, all the time.

Temperature can be expressed as the average value of the objects' Eminence, weighted by their Importance (cf. BAsCET Blackboard to know how Eminence is computed). A high Temperature level points out the lack of trustworthy information on which to found a decision; a lower Temperature level shows that information is more sure, and that one can trust the indications given by the Concept Network: agents urgency values are more relevant when the Temperature value is high.

That's why the choice of the agents is turned nondeterministic when the Temperature value is high, and more deterministic when it is low. It allows to proceed in a broad-first manner in the beginning, when available information is scarce and untrustworthy. All the agents have then same likelihood to be run, thus all possible solutions are explored with the same "haste". Conversely, when the Temperature slumps, agents with a higher urgency value (whose father nodes are the more activated) are preferred.

urgency values  +-----+----------+------+--+------+
of the agents   | 56  |   100    |  84  |30|  91  |
in the coderack +-----+----------+------+--+------+

adjustment      +----+-----------+-------++-------+
according to    | 50 |    111    |  89   ||  99   |
Temperature     +----+-----------+-------++-------+
(T=30)                              ^
                                    |
Random value=189--------------------+

The new urgency value nuv is computed according to the starting urgency value uv, the Temperature value T, and the urgency values of the n agents uv_i:

nuv = uv + ( T - 50 ) / 50 x ( (Σⁿ_i=1 uv_i) / n - uv)

This formula shows that with an average Temperature value (50), urgency values are not modified. It is easy to see that when the Temperature value is maximal (100), the new urgency value nuv is uv + (Σⁿ_i=1 uv_i) / n - uv, thus equal to the average of the urgency values, for all uv. Thus, at maximal Temperature, each agents running has the same probability value.

As the writer is French, his English is not perfect. All remarks are accepted (Message Inbox).