Information processing system using threshold passive modification

• US 4,254,474 A
• Filed: 08/02/1979
• Issued: 03/03/1981
• Est. Priority Date: 08/02/1979
• Status: Expired due to Term

• Current Assignee: Nestor, Inc.
• Original Assignee: Nestor Associates
• Inventors: Cooper, Leon N., Elbaum, Charles
• Primary Examiner(s): Springborn, Harvey E.
• Application Number:
  US06/063,071
• Time in Patent Office: 579 Days
• Field of Search: 364/200 MS File, 364/900 MS File, 364/715
• US Class Current: 706/41
• CPC Class Codes: G06F 17/10 Complex mathematical operations function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544
• CPC Class Codes: G06F 18/24143 Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

An adaptive information processing system includes input terminals adapted to receive N input signals s₁, s₂, . . . s_N representing an "input"; output terminals adapted to present n output responses r₁, R₂, . . . r_n representing a "response"; and information processing apparatus, connecting the input with the output terminals, and operative to carry out the processing steps required to "map" different inputs into responses in accordance with a particular passive modification algorithm. This algorithm permits the recognition of unknown inputs and separates inputs which are in different classes but which are originally close to one another.

Claims

1. An information processing system comprising, in combination:

a. a plurality (n) of input terminals 1, 2 . . . ,j . . . , N adapted to receive N input signals s₁, s₂ . . . ,s_j . . . ,s_N, respectively;

b. a plurality (n) of output terminals 1, 2 . . . , i . . . , n adapted to present n output responses r₁,r₂ . . . ,r_i . . . , r_n, respectively;

c. a plurality of junction elements, called mnemonders, each mnemonder coupling one of said input terminals (input j) with one of said output terminal (output i) and providing a transfer of information from input j to output i in dependence upon the signal s_j appearing at the input j and upon the mnemonder transfer function A_ij ; and

d. means for modifying the transfer function A_ij of at least one of said mnemonders, when in a learning mode, in dependence upon the product of at least one of said input signals and one of said output responses;

wherein modifications to the transfer functions of the mnemonders, when in a learning mode, take the form;

space="preserve" listing-type="equation">A.sub.ij (t)=γ

A.sub.ij (t-1)+δ

A.sub.ij (t)

where δ

A_ij (t) has the form;

space="preserve" listing-type="equation">δ

A.sub.ij (t)=η

(r.sub.i)r.sub.i s.sub.j

where A_ij (t) is the new (modified) transfer function, A_ij (t-1) is the previous transfer function, γ

is a decay constant in the range 0≦

γ

≦

1 and η

is a variable learning constant that is dependent upon the output r_i.

2. The information processing system defined in claim 1, further comprising logic element means, connected between each output terminal and the mnemonders coupled thereto, for receiving mnemonder outputs s_j '"'"' and producing an output response r_i in dependence upon said mnemonder outputs.

3. The information processing system defined in claim 2, wherein each of said logic element means produces said output response r_i in proportion to the sum of the mnemonder outputs s_j '"'"' applied thereto:

whereby said output response r_i takes the form;

##EQU8##

4. The information processing system defined in claim 1, wherein said modifying means includes means for changing all (n×

N) of said transfer functions A₁₁, A₁₂, . . . A₂₁, A₂₂, . . . A_nN in accordance with said formulae for A_ij (t).

5. The information processing system defined in claim 1, wherein said modifying means includes means for varying the modification of said transfer functions A_ij.

6. The information processing system defined in claim 5, wherein said modification varying means includes means for reducing the modification of said transfer functions A_ij to zero,

whereby said information processing system operates as a distributed memory.

7. The information processing system defined in any one of claims 1-6 wherein η

is larger than zero for r_i ≦

θ

_M, where θ

_M is the modification threshold, and smaller than zero for r_i <

θ

_M.

8. The information processing system defined in claim 7 wherein δ

A_ij (t) has the form;

space="preserve" listing-type="equation">δ

A.sub.ij (t)=±

η

.sup.±

r.sub.i s.sub.j (t)

where η

is greater or equal to zero having one value η

^- if r_i is less than said threshold value θ

_M and another value η

⁺ if r_i is greater than said threshold value θ

_M.

9. The information processing system defined in claim 8, wherein η

^- and η

⁺ are equal.

10. The information processing system defined in claim 8, wherein η

^- and η

⁺ are unequal.

11. The information processing system defined in claim 7, wherein θ

_M is a variable quantity.

12. The information processing system defined in claim 11 wherein θ

_M is a function of the output, r_i.

13. A method of processing information comprising the steps of:

a. receiving a plurality of input signals s₁,s₂ . . . , s_j . . . ,s_N,

b. producing a plurality of intermediate signals s'"'"'_i1, s'"'"'_i2, . . . s'"'"'_ij . . . ,s'"'"'_iN, each s'"'"'_ij of which is dependent upon a respective one of said input signals s_j and an associated transfer function A_ij ;

c. producing an output response r_i which is dependent upon at least one of said intermediate signals s'"'"'_ij ; and

d. modifying at least one of said transfer functions A_ij, when in a learning mode, according to the formula;

space="preserve" listing-type="equation">A.sub.ij (t)=γ

A.sub.ij (t-1)+δ

A.sub.ij (t)

where δ

A_ij (t) has the form;

space="preserve" listing-type="equation">δ

A.sub.ij (t)=η

(r.sub.i)r.sub.i s.sub.j (t)

where A_ij (t) is the new (modified) transfer function, A_ij (t-1) is the previous transfer function, γ

is a decay constant in the range 0≦

γ

≦

1 and η

is a variable learning constant that is dependent upon the output r_i.

14. The method defined in claim 13, wherein said output response r_i takes the form:

##EQU9##

15. The method defined in claim 13, wherein said modifying step includes changing all (n×

N) of said transfer functions A₁₁, A₁₂, . . . A₂₁, A₂₂, . . . A_nN in accordance with said formula for A_ij (t).

16. The method defined in claim 13, wherein said modifying step includes varying the modification of said transfer functions A_ij.

17. The method defined in any one of claims 13-16 wherein η

is larger than zero for r_i ≦

θ

_M, where θ

_M is the modification threshold, and smaller than zero for r_i<

θ

_M.

18. The method defined in claim 17 wherein δ

A_ij (t) has the form;

space="preserve" listing-type="equation">δ

A.sub.ij (t) =±

η

.sup.±

r.sub.i s.sub.j (t)

where η

is greater or equal to zero having one value η

^- if r_i is less than said threshold value θ

_M and another value η

⁺ if r_i is greater than said threshold value θ

_M.

19. The method defined in claim 18, wherein η

^- and η

⁺ are equal.

20. The method defined in claim 18, wherein η

^- and η

⁺ are unequal.

21. The method defined in claim 17, wherein θ

_M is a variable quantity.

22. The method defined in claim 21 wherein θ

_M is a function of the output, r_i.