Neural processor comprising means for normalizing data

• US 5,625,753 A
• Filed: 04/29/1993
• Issued: 04/29/1997
• Est. Priority Date: 04/29/1992
• Status: Expired due to Fees

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Deville, Yannick
• Primary Examiner(s): Davis, George B.
• Application Number:
  US08/055,431
• Time in Patent Office: 1,461 Days
• Field of Search: 395/27, 395/22, 395/24
• US Class Current: 706/26
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

A neural processor, including neural calculation apparatus (30, NQ, RQ) which normalize an input data X with respect to another input data Y. The calculation apparatus performs a division of X by Y in order to determine a quotient Q. The calculation apparatus is trained to calculate by iteration a series of contributions ΔQ_i which are used to update a partial quotient QP which becomes the quotient Q at the end of calculation. Calculation can be performed on an arbitrary arithmetic base which determines the number of neurons utilized and also the accuracy of the calculation. It is also possible to utilize a partial remainder RP.

Claims

1. A neural processor comprising neural calculation means (11-15) for normalizing first input data X with respect to second input data Y by calculating a quotient q of a division of X by Y, said neural calculation means including:

at least one first neuron for recursively calculating a series of contributions Δ

Q_i =q_i ·

Bⁱ which together form the quotient Q on an arbitrary arithmetic base B, where q and i are integers; and

at least one second neuron fed by said at least one first neuron for recursively updating a partial quotient QP by summing said contributions Δ

Q_i in order to produce the quotient Q.

2. The neural processor of claim 1, wherein the partial quotient QP is initially zero, and said processor performs the following operations:

calculating a plurality of quantities SD_j according to the following equation;

space="preserve" listing-type="equation">SD.sub.j =X-(QP.sub.i+1 +j·

B.sup.i)·

Y

said plurality of quantities resulting from B-1 operations performed by at least one neuron for an index j varying from 1 to B-1, i being an integer index initially equal to a predetermined maximum number;

first determining a value j=q_i for which;

space="preserve" listing-type="equation">sgn(SD.sub.j)≠

sgn(SD.sub.j+1)

where SD₀ ≧

0, SD_B <

0, and sgn (0)=+1;

second determining a contribution Δ

Q_i =q_i ·

Bⁱ ;

third determining a new partial quotient such that;

space="preserve" listing-type="equation">QP.sub.i =QP.sub.i+1 Δ

Q.sub.i ;

and

decrementing i;

iterating through the operations until a minimum value of i is reached so as to determine the quotient Q to a predetermined accuracy.

3. The neural process or of claim 2, wherein the neural calculation means further calculates a partial remainder RP_i =X-QP_i ·

Y by performing the following additional operations;

setting the partial remainder initially to X; and

updating the partial remainder by iteration so that;

space="preserve" listing-type="equation">RP.sub.i =RP.sub.i+1 -Y·

Δ

Q.sub.i

where

i is an integer index which is initialized to a maximum value and decremented to a predetermined minimum value which yields a desired precision for Q.

4. The neural processor of claim 3, wherein the neural calculation means further calculates a plurality of quantities SD_j according to the following equation:

space="preserve" listing-type="equation">SD.sub.j =RP.sub.i+1 -j·

B.sup.i ·

Y.

where j is an integer index varying from 1 to B-1.

5. The neural processor of claim 3, wherein for B=2, the neural processor operates with three neurons, a first neuron executing the calculating operations, first determining operations and second determining operations and a second neuron (NQ) and a third neuron (NR) executing the updating of the partial quotient and the partial remainder, respectively.

6. The neural processor of claim 2, wherein the second and third determining operations are executed by at least one neuron.

7. The neural processor of claim 2, wherein for B=2, the neural processor operates with two neurons, a first neuron executing the calculating operations, first determining operations, and second determining operations and a second neuron executing the third determining operation.

8. The neural processor of claim 1, wherein the execution of the B-1 operations is distributed between several groups of at least one neuron each.

9. The neural processor of claim 1, wherein the neural calculation means further performs a division at one instant and performs neural resolving and/or learning tasks at another instant.

10. A neural processor comprising neural calculation means for normalizing first input data X with respect to a second input data Y by calculating a quotient Q of a division of X by Y, said calculation means comprising:

a first neuron;

calculating a quantity SD₁ according to the following equation;

space="preserve" listing-type="equation">SD.sub.1 =X-(QP.sub.i+1 +2.sup.i)·

Y

where i is an index initially set to a maximum value, and QP₁ is a partial quotient which is initially zero; and

first determining a value q_i =j for which

space="preserve" listing-type="equation">sgn(SD.sub.j)≠

sgn(SD.sub.j+1)

where 0≦

j≦

1, SD₀ ≧

0, SD₂ <

0, and sgn(0)=+1; and

a second neuron

second determining a contribution Δ

Q=q_i ·

2ⁱ ; and

third determining a new partial quotient QP_i such that

space="preserve" listing-type="equation">QP.sub.i =QP.sub.i+1+Δ

Q.sub.i

the processor then decrementing i and iterating through the operations so as to determine the quotient Q to a predetermined accuracy.

11. The neural processor of claim 10 comprising at least one third neuron for calculating a partial remainder RP_i =X-QP_i ·

Y, the partial remainder, initially being equal to X, being updated by iteration so that

space="preserve" listing-type="equation">RP.sub.i =RP.sub.i+1 -Y·

Δ

Q.sub.i.