Method of a network that learns

• US 6,029,099 A
• Filed: 03/10/1997
• Issued: 02/22/2000
• Est. Priority Date: 11/22/1993
• Status: Expired due to Fees

• Current Assignee: Robert Alan Brown
• Original Assignee: Robert Alan Brown
• Inventors: Brown, Robert Alan
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): PENDER JR, MICHAEL J
• Application Number:
  US08/814,282
• Time in Patent Office: 1,079 Days
• Field of Search: 395/20, 395/11, 395/23, 700/245, 700/250
• US Class Current: 700/245
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

A self-learning network with: (a) one or more sensor(s) in one or more sensor unit(s) with a plurality of memory cells that forward-select values of one or more intermediate variable(s) I₁, J₁, . . . according to their stored values and the conditions identified by its sensor(s) in a given sensing and action period, (b) with or without one or more intermediate units with a plurality of bidirectional memory cells that forward-select a value of one or more intermediate output variable(s) I₂, J₂, . . . according to their stored values and the value(s) of the intermediate input variable(s) I₁, J₁, . . . , and (c) one or more actuator(s) in one or more actuator unit(s) connected to one or more intermediate variable(s) I₁, J₁, . . . or I₂, J₂, . . . with a plurality of bidirectional memory cells that forward-select values of one or more actuators according to their stored values and the value(s) of its intermediate variable(s) I₁, J₁, . . . or I₂, J₂, . . . . The bidirectional memory cells then back-select a value of each intermediate co-variable I₁ '"'"', J₁ '"'"', . . . or I₂ '"'"', J₂ '"'"', . . . according to their stored values and the actions that actually occur within each specific sensing and action periods. The stored values are reduced in the forward-selecting memory cells, and the stored values are increased in the back-selecting bidirectional memory cells, and are increased in the memory cells at the intersection of the back-selected value(s) of the intermediate co-variable I₁ '"'"', J₁ '"'"', . . . and the conditions identified by its sensor(s) in that sensing and action period.

Claims

1. A method for operating a network using a bidirectional memory cell means that stores an analog memory record of a sensitivity, said bidirectional memory cell means being connected to one of a value of an input variable means and a one of a value of an input co-variable means, and one of a value of an output variable means and one of a value of an output co-variable means, uniquely, said method of said bidirectional memory cell means comprising:

a. allowing said bidirectional memory cell means to make a forward-selection of a forward-selected value of said output variable means, and

b. making said forward selection by a forward-selecting bidirectional memory cell means, said forward-selecting bidirectional memory cell means being connected to an extant value of said input variable means and said forward-selected value of said output variable means, said forward-selection being made when a forward-selecting voltage ramp means reaches a required voltage level determined by an extant value of said sensitivity, and

c. making a reduction is said sensitivity of said forward-selecting bidirectional memory cell means, said reduction being equal to said extant value of said sensitivity multiplied by a predetermined constant, said predetermined constant being in a range from one to a value greater than zero, and

d. prohibiting said forward-selection of another value of said output variable means, and

e. allowing said bidirectional memory cell means to make a back-selection of a back-selected value of said input co-variable means, and

f. making said back-selection by a back-selecting bidirectional memory cell means, said back-selecting bidirectional memory cell means being connected to an extant value of said output co-variable means and said back-selected value of said input co-variable means, said back-selection being made when a back-selecting voltage ramp means reaches said required voltage level determined by said extant value of said sensitivity, and

g. making an increase is said sensitivity of said back-selecting bidirectional memory cell means, said increase being equal to a difference between a maximum sensitivity and said extant value of said sensitivity, said difference being multiplied by a predetermined constant, said predetermined constant being in a range from one to a value greater than zero, and

h. prohibiting said back-selection of another value of said input co-variable means, wherein

i. said sensitivity of said bidirectional memory cell means is determined by a number of times said increase is made in relation to a number of times said decrease is made, whereby

j. said bidirectional memory cell means is useful in a network of a self-learning unit means.

2. The method of claim 1 further comprising a bidirectional memory submatrix means, where said bidirectional memory submatrix means is made up of a plurality of said bidirectional memory cell means, wherein said bidirectional memory submatrix means is connected to one said input variable means and said input co-variable means, and one said output variable means and said output co-variable means, said bidirectional memory submatrix means operating in a sequence of a sensing and action period, a method of said bidirectional memory submatrix means comprising:

a. identifying an extant value of said input variable means in said sensing and action period, and

b. prohibiting said bidirectional memory submatrix means from identifying another said extant value of said input variable means in said sensing and action period, and allowing said bidirectional memory submatrix means to make said forward selection, and

c. making said forward-selection of said forward-selected value of said output variable means within said sensing and action period, said forward-selection being made by one said forward-selecting bidirectional memory cell means in said bidirectional memory submatrix means, said forward-selecting bidirectional memory cell means being connected to said forward-selected value of said output variable means, said forward-selecting bidirectional memory cell means also being in a forward-selecting energized set, said forward-selecting energized set being made up of said bidirectional memory cell means connected to said extant value of said input variable means, said forward-selection being made by comparing a sensitivity value stored in each said bidirectional memory cell means in said forward-selecting energized set of said bidirectional memory cell means, said forward-selecting bidirectional memory cell means having a highest of said sensitivity value in said forward-selecting energized set, and

d. making said reduction of said sensitivity of said forward-selecting bidirectional memory cell means, and

e. prohibiting said bidirectional memory submatrix means from making another said forward-selection in said sensing and action period, and allowing said bidirectional memory submatrix means to identify an extant value of said output co-variable means, and

f. identifying said extant value of said output co-variable means after a predetermined delay from said making said forward-selection, and

g. prohibiting said bidirectional memory submatrix means from identifying another said extant value of said output co-variable means in said sensing and action period, and allowing said bidirectional memory submatrix means to make said back-selection, and

h. making said back-selection of said back-selected value of said input co-variable means within said sensing and action period, said back-selection being made by one said back-selecting bidirectional memory cell means in said bidirectional memory submatrix means, said back-selecting bidirectional memory cell means being connected to said back-selected value of said input co-variable means, said back-selecting bidirectional memory cell means also being in a back-selecting energized set, said back-selecting energized set being made up of said bidirectional memory cell means connected to said extant value of said output co-variable means, said back-selection being made by comparing said sensitivity value stored in each said bidirectional memory cell means in said back-selecting energized set of bidirectional memory cell means, said back-selecting bidirectional memory cell means having a highest of said sensitivity value in said back-selecting energized set, and

i. making said increase in said sensitivity of said back-selecting bidirectional memory cell means, and

j. prohibiting said bidirectional memory submatrix means from making another said back-selection in said sensing and action period, wherein

k. each said forward-selected value has occurred with a highest historical probability with each said extant value of said input variable means in said sensing and action period, and wherein

l. each said back-selected value has occurred with said highest historical probability with each said extant value of said output co-variable means in said sensing and action period, whereby

m. each said forward-selected value and each said back-selected value of said bidirectional memory submatrix means can influence other said self-learning unit means in said network of said self-learning unit means.

3. The method of claim 2, further comprising a plurality of said input variable means and said input co-variable means, a plurality of said output variable means and said output co-variable means, and a plurality of said bidirectional memory submatrix means, wherein each said bidirectional memory submatrix means is connected to one said input variable means and said input co-variable means, and one said output variable means and said output co-variable means uniquely, forming a bidirectional monolithic matrix means, a method of said bidirectional monolithic matrix means further comprising:

a. identifying said extant value of each said input variable means in said sensing and action period, and

b. prohibiting another said extant value of each said input variable means from being identified in said sensing and action period, and allowing said bidirectional monolithic matrix means to make a forward-selection of a forward-selected value of each said output variable means, and

c. making said forward-selection of said forward-selected value in each said output variable means, where one said forward-selecting bidirectional memory cell means in each said bidirectional memory submatrix means contributes to said forward-selection, where said forward-selecting bidirectional memory cell means belong to one of a forward-selecting set, where any one of said forward-selecting set is connected to only one said forward-selected value of only one said output variable means, and where each said forward-selecting set also belongs to at least one of a forward-selecting energized set of said bidirectional memory cell means, said forward-selecting energized set being connected to an extant value of said input variable means, and wherein

d. each said forward-selecting set contains a forward-selecting bidirectional memory cell means with a lowest sensitivity, wherein said forward-selecting bidirectional memory cell means with a lowest sensitivity has a higher sensitivity than a non-selecting bidirectional memory cell means with a lowest sensitivity in any of a non-selecting set, each said non-selecting set being connected to an unselected value of said output variable means, said non-selecting bidirectional memory cell means also being in said forward-selecting energized set, and

e. making said reduction in said sensitivity in each said forward-selecting bidirectional memory cell means, and

f. prohibiting another said forward-selection of said unselected value of said output variable means in said sensing and action period, and allowing an identification of said extant value of each said output co-variable means, and

g. identifying said extant value of each said output co-variable means after said predetermined delay from said making said forward-selection, and

h. prohibiting another said extant value of each said output co-variable means from being identified in said sensing and action period, and allowing said bidirectional monolithic matrix means to make said back-selection, and

i. making said back-selection of said back-selected value of each said input co-variable means, where one of a back-selecting bidirectional memory cell means in each said bidirectional memory submatrix means contributes to said back-selection, where said back-selecting bidirectional memory cell means belong to one of a back-selecting set, where any one of said back-selecting set is connected to only one said back-selected value of only one said input co-variable means, and where each said back-selecting set also belongs to at least one of a back-selecting energized set of said bidirectional memory cell means, said back-selecting energized set being connected to said extant value of said output co-variable means, and wherein

j. each said back-selecting set contains a back-selecting bidirectional memory cell means with a lowest sensitivity, wherein said back-selecting bidirectional memory cell means with a lowest sensitivity has a higher sensitivity than a non-back-selecting bidirectional memory cell means with a lowest sensitivity in any of a non-back-selecting set, each said non-back-selecting set being connected to an unselected value of said input co-variable means, said non-back-selecting bidirectional memory cell means also being in said back-selecting energized set, and

k. making said increase in said sensitivity in each said back-selecting bidirectional memory cell means, and

l. prohibiting another said back-selection of said unselected value of said input co-variable means in said sensing and action period, wherein

m. said forward-selected value of said output variable means has a highest historical probability of occurring with said extant value of said input variable means in said sensing and action period, said highest historical probability being due to a multiple of said reduction in said sensitivity and said increase in said sensitivity, and wherein

n. said back-selected value of said input co-variable means has a highest historical probability of occurring with said extant value of said output co-variable means in said sensing and action period, said highest historical probability being due to a multiple of said reduction in said sensitivity and said increase in said sensitivity, whereby

o. said method of said bidirectional monolithic matrix means is useful in a self-learning intermediate unit means, and in a self-learning actuator unit means in said network of said self-learning unit means.

4. The method of claim 3 further including at least one of an actuator variable means, one said actuator variable means being connected to one said output variable means and said output co-variable means exclusively, said bidirectional monolithic matrix means and said actuator variable means forming said self-learning actuator unit means, a method of said self-learning actuator unit means further comprising:

a. identifying said extant value of each said input variable means in said sensing and action period, and

b. prohibiting another said extant value of each said input variable means from being identified in said sensing and action period, and allowing said bidirectional monolithic matrix means to make said forward-selection of said forward-selected value of each said output variable means, and

C. making said forward-selection of said forward-selected value of each said output variable means, and

d. making said reduction in said sensitivity of said forward-selecting bidirectional memory cell means, and

e. prohibiting another said forward-selection of said unselected value of said output variable means in said sensing and action period, and

f. making an attempt to produce said forward-selected value, said attempt being made by said actuator variable means, said attempt being made after a predetermined delay from said forward-selection, said attempt being made for a predetermined action period, wherein

g. each said actuator variable means arrives at an extant actuator value during said predetermined action period, said extant actuator value being determined by said forward-selected value and forces applied to said actuator variable means, and

h. creating a holding force upon said actuator variable means at an end of said predetermined action period, said holding force being made by an actuator brake means, said holding force continuing except when said actuator variable means makes said attempt to produce said forward-selected value, and

i. making a measurement of said extant actuator value of each said actuator variable means during a predetermined feedback period within said sensing and action period, said measurement being made by said output co-variable means, said predetermined feedback period occurring after said predetermined action period, and

j. prohibiting said measurement of another said extant actuator value in said sensing and action period, and allowing said bidirectional monolithic matrix means to make said back-selection of said back-selected value of each input co-variable means, and

k. making said back-selection of said back-selected value of said input co-variable means, said back-selecting being made by said back-selecting bidirectional memory cell means, and

l. making said increase in said sensitivity of said back-selecting bidirectional memory cell means, and

m. prohibiting another said back-selection of said unselected value of said input co-variable means in said sensing and action period, wherein

n. said forward-selected value of said output variable means represents a value of said actuator variable means that is most likely to be carried out with said extant value of said input variable means in said sensing and action period, said forward-selected value being most likely by virtue of a multiple of said reduction in sensitivity and said increase in said sensitivity, and

o. said back-selected value of said input co-variable means represents a value of said input variable means that is most likely to forward-select said extant actuator value in said sensing and action period, said back-selected value being most likely by virtue of a multiple of said reduction in sensitivity and multiple said increase in said sensitivity, whereby

p. said method of said self-learning actuator unit means is useful in said network of said self-learning unit means.

5. The method of claim 4 further including a self-learning sensor unit means, said self-learning sensor unit means having at least one of a submatrix means, said submatrix means being made up of a plurality of a memory cell means, wherein each said submatrix means is connected to one of a sensor variable means and one of an output variable means of said self-learning sensor unit means and output co-variable means of said self-learning sensor unit means uniquely, and wherein each said memory cell means is connected to one of a value of said sensor variable means and one of a value of said output variable means of said self-learning sensor unit means and said output co-variable means of said self-learning sensor unit means uniquely, and where each said output variable means of said self-learning sensor unit means and said output co-variable means of said self-learning sensor unit means are connected to said input variable means and input co-variable means of said self-learning actuator unit means exclusively, said self-learning sensor unit means and said self-learning actuator unit means forming a duplex network of said self-learning unit means, a method of said duplex network of said self-learning unit means further comprising:

a. identifying said extant value of each said sensor variable means at a beginning of said sensing and action period, and

b. prohibiting another said extant value of said sensor variable means from being identified in said sensing and action period, and allowing said self-learning sensor unit means to make a forward-selection of a forward-selected value of each said output variable means of said self-learning sensor unit means, where

c. a forward-selecting memory cell means in each said submatrix means makes said forward-selection of said forward-selected value of said output variable means of said self-learning sensor unit means, where said forward-selecting memory cell means belong to a forward-selecting set of said memory cells means, said forward-selecting set being connected to said forward-selected value of said output variable means, and where members of said forward-selecting set also belong to a forward-selecting energized set of said memory cell means, said forward-selecting energized set being connected to said extant value of said sensor variable means at said beginning of said sensing and action period, and wherein

d. a lowest sensitivity of said forward-selecting memory cell means in each said forward-selecting set is higher than a lowest sensitivity of a non-selecting memory cell means in any of a non-selecting set of said non-selecting memory cell means, each said non-selecting set being connected to an unselected value of said output variable means of said self-learning sensor unit means, said non-selecting memory cell means also being in said forward-selecting energized set, and

e. making a reduction in a sensitivity of each said forward-selecting memory cell means, said reduction being equal to a sensitivity of each said forward-selecting memory cell means multiplied by a predetermined constant, said predetermined constant being in a range greater than zero and less than one, and

f. prohibiting another said forward-selection of said unselected value of said output variable means of said self-learning sensor unit means in said sensing and action period, and allowing said self-learning actuator unit means to make said forward-selection of said forward-selected value of each said output variable means of said self-learning actuator unit means, and

g. making said forward-selection of said forward-selected value of each of said output variable means of said self-learning actuator unit means, said forward-selection being made by one said forward-selecting bidirectional memory cell means in each said bidirectional memory submatrix means in said bidirectional monolithic matrix means of said self-learning actuator unit means, said forward-selection being made according to said sensitivity of said bidirectional memory cell means connected to each said forward-selected value of each of said output variable means of said self-learning sensor unit means, said forward selection being made in the manner of claim 4, and

h. making said reduction in said sensitivity of said forward-selecting bidirectional memory cells means in each said bidirectional memory submatrix means, and

i. prohibiting another said forward-selection of said unselected value of said output variable means of said self-learning actuator unit means in said sensing and action period, and allowing said actuator variable means to attempt to produce said forward-selected value, and

j. making said attempt to produce said forward-selected value of said actuator variable means, said attempt being made in the manner of claim 4, wherein

k. said actuator variable means arrives at said extant actuator value at said end of said predetermined action period, and

l. creating said holding force upon said actuator means at said end of said predetermined action period, and

m. making said measurement of said extant actuator value of each said actuator means at said end of said predetermined action period, and

n. prohibiting said measurement of another said extant actuator value in said sensing and action period, and allowing said self-learning actuator unit means to make said back-selection of said back-selected value of each said input co-variable means of said self-learning actuator unit means, and

o. making said back-selection of said back-selected value of said input co-variable means of said self-learning actuator unit means, said back-selection being made by a back-selecting bidirectional memory cell means in each said bidirectional memory submatrix means in said bidirectional monolithic matrix means of said self-learning actuator unit means, said back-selection being made according to said sensitivity of said bidirectional memory means connected to said extant actuator value, said back-selection being made in the manner of claim 4, and

p. making said increase in said sensitivity of said back-selecting bidirectional memory cell means, and

q. prohibiting another said back-selection of said unselected value of said input co-variable means of said self-learning actuator unit means in said sensing and action period, and allowing said self-learning sensor unit means to make said increase in sensitivity of said feedback-selected memory cell means, and

r. making said increase in said sensitivity of each said feedback-selected memory cell means in each said submatrix means in said self-learning sensor unit means, said feedback-selected memory cell means being at an intersection of said extant value of said sensor variable means and said back-selected value of said input co-variable means of said self-learning actuator unit means, said increase being equal to a difference between a maximum possible sensitivity of said memory cell means and said sensitivity of said memory cell means, said difference being multiplied by a predetermined constant, wherein

s. said duplex network of said self-learning unit means selects said forward-selected value of each said actuator variable means, said forward-selected value of each said actuator variable means having a highest historical probability of being carried out with said extant value of each said sensor variable means in said sensing and action period, said highest historical probability being due to a multiple said decrease in said sensitivity and a multiple said increase in said sensitivity, whereby

t. said method of said duplex network of said self-learning unit means requires fewer said memory cell means and said bidirectional memory cell means when fewer of a sensor/actuator relations are required by a control task, since only as many of a value of an intermediate variable means are required as said sensor/actuator relations, said duplex network of said self-learning unit means forming a self-learning network.

6. The method of claim 5, further including a digitizing means, said digitizing means having digitized connections between each of said self-learning sensor unit means and said sensor variable means forming said digitized self-learning sensor unit means, and digitized connections between each of said self-learning actuator unit means and said actuator variable means forming said digitized self-learning actuator unit means, said digitized self-learning sensor unit means and said digitized self-learning actuator unit means forming a digitized duplex network of said self-learning unit means, a method of said digitized duplex network of said self-learning unit means further comprising:

a. decomposing said extant value of a sensor variable means by an encoder means into a unique combination of values of a plurality of an aggregate input variable means of said self-learning sensor unit means, and

b. identifying said extant value of each said aggregate input variable means at said beginning of said sensing and action period, and

c. prohibiting another said extant value of each said aggregate input variable means from being identified in said sensing and action period, and allowing said self-learning sensor unit means to make said forward-selection of said forward-selected value of each said output variable means of said self-learning sensor unit means, and

d. making said forward-selection by said self-learning sensor unit means of said value of each said output variable means of said self-learning sensor unit means, said forward selection being made according to said sensitivity of said memory cell means connected to said unique combination of values of said plurality of said aggregate input variable means, said forward-selection being made in the manner of claim 5, and

e. making said reduction in a sensitivity of each said forward-selecting memory cell means in said self-learning sensor unit means, and

f. prohibiting another said forward-selection by said self-learning sensor unit means in said sensing and action period, and allowing said self-learning actuator unit means to make said forward-selection, and

g. making said forward-selection by said self-learning actuator unit means of a unique combination of values of a plurality of an aggregate output variable means, said forward-selection being made in the manner of claim 5, and

h. making said reduction in a sensitivity of each said forward-selecting bidirectional memory cell means in said self-learning actuator unit means, and

i. prohibiting another said forward-selection by said self-learning actuator unit means in said sensing and action period, and

j. synthesizing said forward-selected value of said actuator variable means by a decoder means from said unique combination of values of said plurality of said aggregate output variable means of said self-learning actuator unit means, and

k. making said attempt by said actuator variable means to produce said forward-selected value of said actuator variable means, wherein

l. said actuator variable means arrives at said extant actuator value at said end of said predetermined action period, and

m. creating said holding force upon said actuator variable means, and

n. making said measurement of said extant value of said actuator variable means at said end of said predetermined period, and

o. decomposing said extant value of said actuator variable means by an encoder means into a unique combination of values of a plurality of an aggregate output co-variable means of said self-learning actuator unit means, and

p. prohibiting said measurement of another said extant actuator value in said sensing and action period, and allowing said self-learning actuator unit means to make said back-selection of said back-selected value of each said input co-variable means of said self-learning actuator unit means, and

q. making said back-selection of said back-selected value of said input co-variable means of said self-learning actuator unit means, said back-selection being made according to said sensitivity of said bidirectional memory cell means connect to said unique combination of values of said plurality of said aggregate output co-variable means, said back-selection being made in the manner of claim 5, and

r. making said increase in said sensitivity of each said back-selecting bidirectional memory cell means in said self-learning actuator unit means, and

s. prohibiting another said back-selection by said self-learning actuator unit means in said sensing and action period, and allowing said self-learning sensor unit means to make said increase in sensitivity of said feedback-selected memory cell means, and

t. making said increase in a sensitivity of each said feedback-selected memory cell means in said self-learning sensor unit means, said feedback-selected memory cell means being at said intersection of said back-selected value of said input co-variable means of said self-learning actuator unit means and said unique combination of values of said plurality of said aggregate input variable means, wherein

u. said forward-selected value of said actuator variable means has a highest historical probability of being carried out with said extant value of said sensor variable means in each said sensing and action period, said highest historical probability being due to a multiple of said decrease in said sensitivity and said increase in said sensitivity, whereby

v. said method of said digitized duplex network of said self-learning unit means allows said digitized duplex network of said self-learning unit means to have fewer said memory cell means and said bidirectional memory cell means than an undigitized duplex network of said self-learning unit means, said digitized duplex network of said self-learning unit means forming a self-learn network.

7. The method claim 5 further including at least one of a self-learning intermediate unit means, said duplex network of said self-learning unit means and said self-learning intermediate unit means forming a nodal network of said self-learning unit means, where said self-learning intermediate unit means contains at least one said bidirectional memory submatrix means, wherein each said bidirectional memory submatrix means is connected to one of an input variable means of said self-learning intermediate unit means and an input co-variable means of said self-learning intermediate unit means, and one of an output variable means of said self-learning intermediate unit means and an output co-variable means of said self-learning intermediate unit means uniquely, and wherein said input variable means and said input co-variable means of said self-learning intermediate unit means is connected to said output variable means and said output co-variable means of at least one of said self-learning sensor unit means, and said output variable means and said output co-variable means of said self-learning intermediate unit means is connected to an said input variable means and said input co-variable means of at least one of said self-learning actuator unit means, said self-learning intermediate unit means operating in said sensing and action period, a method of said nodal network of said self-learning unit means further comprising:

a. making an identification of said extant value of said sensor variable means of said self-learning sensor unit means, said identification being made at said beginning of said sensing and action period, and

b. prohibiting another said extant value of said sensor variable means from being identified in said sensing and action period, and allowing said self-learning sensor unit means to make said forward-selection of said forward-selected value of each said output variable means of said self-learning sensor unit means, and

c. making said forward-selection of said forward-selected value of said output variable means of said self-learning sensor unit means in said sensing and action period, and

d. making said reduction in said sensitivity of each said forward-selecting memory cell means in said self-learning sensor unit means, and

e. prohibiting another said forward-selection by said self-learning sensor unit means in said sensing and action period, and allowing said self-learning intermediate unit means to make said forward-selection, and

f. making said forward-selection of a forward-selected value of said output variable means of said self-learning intermediate unit means in said sensing and action period, said forward-selection being made according to said sensitivity of said bidirectional memory cell means connected to said forward-selected value of said output variable means of said self-learning sensor unit means, said forward-selection being made in the manner of said self-learning actuator unit means of claim 5, and

g. making said reduction in said sensitivity of each of said forward-selecting bidirectional memory cell means in said self-learning intermediate unit means, and

h. prohibiting another said forward-selection by said self-learning intermediate unit means in said sensing and action period, and allowing said self-learning actuator unit means to make said forward-selection, and

i. making said forward-selection of said forward-selected value of each of said output variable means of said self-learning actuator unit means, said forward-selection being made according to said sensitivity of said bidirectional memory cell means connected to said forward-selected value of said output variable means of said self-learning intermediate unit means, said forward-selection being made in the manner of claim 5, and

j. making said reduction in said sensitivity of each said forward-selecting bidirectional memory cell means in each said self-learning actuator unit means, and

k. prohibiting another said forward-selection by said self-learning actuator unit means in said sensing and action period, and

l. making said attempt to produce said forward-selected value of each said actuator unit means, and

m. creating said holding force on each said actuator variable means by said actuator brake means, and

n. making said measurement of said extant value of each said actuator variable means after said delay to allow said actuator variable means to reach said forward-selected value, and

o. prohibiting said measurement of another said extant actuator value in said sensing and action period, and allowing said self-learning actuator unit means to make said back-selection, and

p. making said back-selection in said self-learning actuator unit means of each said back-selected value of said input co-variable means of said self-learning actuator unit means, said back-selection in said self-learning actuator unit means being made in the manner of claim 5, and

q. making said increase in said sensitivity of each said back-selecting bidirectional memory cell means in said self-learning actuator unit means, and

r. prohibiting another said back-selection by said self-learning actuator unit means in said sensing and action period, and allowing said self-learning intermediate unit means to make said back-selection, and

s. making a back-selection by said self-learning intermediate unit means of a back-selected value of each said input co-variable means of said self-learning intermediate unit means, said back-selection in said self-learning intermediate unit means being made according to said sensitivity of said bidirectional memory cell means connected to said back-selected value of each said input co-variable means of said self-learning actuator unit means, said back-selection in said self-learning intermediate unit means being made in the manner of claim 5, and

t. making said increase in said sensitivity of each said back-selecting memory cell means in said self-learning intermediate unit means, and

u. prohibiting another said back-selection by said self-learning intermediate unit means in said sensing and action period, and allowing said self-learning sensor unit means to make said increase in sensitivity of said feedback-selected memory cell means, and

v. making said increase in said sensitivity of each said feedback-selected memory cell means in said self-learning sensor unit means, said feedback-selected memory cell means being at said intersection of said back-selected value of each said input co-variable means of said self-learning intermediate unit means and said extent value of said sensor variable means, wherein

w. said forward-selected value of each said actuator variable means has a highest historical probability of having been carried out with said extant value of each said sensor variable means in each said sensing and action period, said highest historical value being due to a multiple of said decrease in said sensitivity and a multiple of said increase in said sensitivity, whereby

x. said method of said nodal network of said self-learning unit means allows multiple said self-learning sensor unit means to be connected to multiple said self-learning actuator unit means with fewer interconnections and a smaller memory matrix means than multiple of said duplex network of said self-learning unit means, said nodal network of said self-learning unit means being a self-learning network.

8. The method of claim 7, further including a digitizing means, said digitizing means having digitized connections between each of said self-learning sensor unit means and said sensor variable means forming said digitized self-learning sensor unit means, and digitized connections between each of said self-learning actuator unit means and said actuator variable means forming said digitized self-learning actuator unit means, said digitized self-learning sensor unit means and said digitized self-learning actuator unit means forming a digitized nodal network of said self-learning unit means, a method of said digitized nodal network of said self-learning unit means further comprising:

a. decomposing said extant value of said sensor variable means by an encoder means into a unique combination of values of a plurality of an aggregate input variable means, and

b. identifying said extant value of each said aggregate input variable means at said beginning of said sensing and action period, and

c. prohibiting another said extant value of each said aggregate input variable means from being identified in said sensing and action period, and allowing said self-learning sensor unit means to make said forward-selection of said forward-selected value of each said output variable means of said self-learning sensor unit means, and

d. making said forward-selection of said forward-selected value of each of said output variable means of said self-learning sensor unit means, said forward-selection being made according to said unique combination of values of said plurality of said aggregate input variable means, said forward-selection being made in the manner of claim 7, and

e. making said reduction in said sensitivity of each said forward-selecting memory cell means in said self-learning sensor unit means, and

f. prohibiting said self-learning sensor unit means from making another said forward-selection in said sensing and action period, and allowing said self-learning intermediate unit means to make said forward-selection, and

g. making said forward-selection of said forward-selected value of each of said output variable means of said self-learning intermediate unit means, said forward-selection being made in the manner of claim 7, and

h. making said reduction in said sensitivity of each said forward-selecting bidirectional memory cell means in said self-learning intermediate unit means, and

i. prohibiting said self-learning intermediate unit means from making another said forward-selection in said sensing and action period, and allowing said self-learning actuator unit means to make said forward-selection, and

j. making said forward-selection of a unique combination of values of a plurality of an aggregate output variable means, said forward-selection being made by said self-learning actuator unit means, said forward-selection being made according to said sensitivity of said bidirectional memory cell means connected to said forward-selected value of each said output variable means of said self-learning intermediate unit means, said forward-selection being made in the manner of claim 7, and

k. making said reduction in said sensitivity of each of said forward-selecting bidirectional memory cell means in said self-learning actuator unit means, and

l. prohibiting another said forward-selection by said self-learning actuator unit means in said sensing and action period, and

m. synthesizing said forward-selected value of each of said actuator variable means by a decoder means from said unique combination of values of said plurality of said aggregate output variable means of said self-learning actuator unit means, and

n. making said attempt to produce said forward-selected value of said actuator variable means during said predetermined action period, and

o. creating said holding force upon said actuator variable means after said predetermined action period, and

p. making said measurement of said extant actuator value after a sufficient delay to allow said actuator variable means to arrive at said forward-selected value, and

q. decomposing said extant actuator value of said actuator variable means by an encoder means into a unique combination of values of a plurality of aggregate output co-variable means of each said self-learning actuator unit means, and

r. prohibiting said measurement of another said extant actuator value in said sensing and action period, and allowing said self-learning actuator unit means to make said back-selection, and

s. making said back-selection of said back-selected value of said input co-variable means of said self-learning actuator unit means, said back-selection being made according to said unique combination of values of said plurality of said aggregate output co-variable means, said back-selection being made in the manner of claim 7, and

t. making said increase in said sensitivity of each said back-selecting bidirectional memory cell means in each said self-learning actuator unit means, and

u. prohibiting another said back-selection by said self-learning actuator unit means in said sensing and action period, and allowing said self-learning intermediate unit means to make said back-selection, and

v. making said back-selection of said back-selected value of said input co-variable means of said self-learning intermediate unit means, and

w. making said increase in said sensitivity of each said back-selecting bidirectional memory cell means in said self-learning intermediate unit means, and

x. prohibiting another said back-selection by said self-learning intermediate unit means in said sensing and action period, and allowing said self-learning sensor unit means to make said increase in sensitivity of said feedback-selected memory cell means, and

y. making said increase in said sensitivity of each said feedback-selected memory cell means in said self-learning sensor unit means, said increase being made at said intersection of said back-selected value of said input co-variable interface means of said self-learning intermediate unit means and said extant value of said aggregate input variable means, wherein

z. said digitized nodal network of said self-learning unit means forward-selects said forward-selected value of said actuator variable means, said forward-selected value of said actuator variable means having been carried out with a highest historical probability with said extant value of said sensor variable means in each said sensing and action period, said highest historical probability being due to a multiple of said decrease and said increase in said sensitivity, whereby

aa. said method of said digitized nodal network of said self-learning unit means operates with fewer said memory cell means and fewer said bidirectional memory cell means than an undigitized nodal network of said self-learning unit means, said digitized nodal network of said self-learning unit means being a self-learning network.