JP3296609B2 - Neural network that outputs membership function - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a neural network (hereinafter referred to as NN) for outputting a membership function.

[0002]

2. Description of the Related Art In recent years, neurocomputers have been developed and various research results have been announced. A neurocomputer focuses on neurons (neural cells), which are basic elements of the brain, and attempts to achieve a function similar to that of the brain by using NNs formed as a result of combining these neurons as hints. The feature of these NNs lies in the parallel information processing between the neurons and the ability to learn. In addition, the actual network model has an input layer, an intermediate layer, and an output layer provided with cells corresponding to each neuron, and a cell in each layer is connected to all cells in the layers before and after it. ing. In addition, a coupling coefficient called a load is introduced for coupling between the cells, and these coefficients are modified by learning.

[0003]

As described above, according to the prior art, the cells belonging to each layer constituting the NN are connected to all cells in the layers before and after that. Therefore,
In the conventional hierarchical NN learning program, the calculation is performed assuming that one cell is connected to all cells in the layers before and after it, and the NN type is inflexible and the output waveform has one piece. There is a drawback that deviation occurs. SUMMARY OF THE INVENTION The present invention has been made to solve the above-described drawbacks, and has as its object to provide an NN which has flexibility and outputs a membership function in a non-biased arrangement.

[0004]

In order to achieve the above object, the present invention provides a cell for outputting an input signal, a bias cell, and a plurality of first cells connected to each other via a predetermined coupling coefficient. An intermediate cell, a plurality of second intermediate cells connected to the intermediate cells by a predetermined coupling coefficient, and two output cells, and a constant value teacher signal is applied through one of the output cells. Configuration.

[0005]

An embodiment will be described below with reference to the drawings. FIG.
1 is a configuration diagram of an embodiment of an NN according to the present invention. In the figure, reference numerals 1 to 11 denote cells, of which cell 1 is a special cell, called a bias cell, and always outputs 1. W
_{1 to} W ₇ and WB _{1 to} WB ₄ are coupling coefficients. Cell 3 ~
The input / output function 6 is a sigmoid function shown in equation (2), and the input / output function of cells 7 to 11 is an identity function shown in equation (1). The coupling coefficients of cells 4 and 8 and cells 6 and 9 are fixed at -1 as shown in the figure, and no learning is performed. Cell 7 has coupling coefficient 1 from cell 3 and bias cell 1
, A coupling coefficient -1 from the cell 6 and a coupling coefficient 1 from the bias cell 1 are also input to the cell 9. The cell 10 is a cell for learning so that the sum of the outputs of the cells 7 to 9 approaches 1. Therefore, the cell 10 is supplied with the teacher signal 1, and the cells 10 and 7
The coupling coefficient with .about.9 is fixed at 1 and learning is not performed.

[Equation 1] y = x (1) y = 1-exp (−x) / 1 + exp (−x) (2) where y = 1−exp (−x) / 1 + exp (−x) x: sum of inputs to the cell, y: output of the cell. FIG. 3 shows the outputs of cells 7, 8, and 9 shown in FIG. Cell 7
With respect to the sigmoid function, the sigmoid function from the cell 3 is input as it is (coupling coefficient 1), and is shifted rightward by the coupling coefficient from the bias cell 1. The output of the cell 8 is output as the sum of the inverted output of the sigmoid function by the coupling coefficient (-1) from the cell 4 and the sigmoid function itself from the cell 5, and the output of the cell 9 is the coupling from the cell 6. This is due to the inverted output of the sigmoid function by the coefficient (−1) and the shift by the coupling coefficient 1 from the bias cell 1. In addition, the coupling coefficient W ₁ ~W ₄ is determines the smoothness of the membership function of cell 7~9, WB ₁ ~WB ₄ cell 7
The offset of the membership function in the X-axis direction is determined.

FIG. 2 is a block diagram of another embodiment. The difference from FIG. 1 is that two cells for handling a sigmoid function are provided from an NN (reference numeral A in the figure) having a membership function function. The connection between them is as shown in FIG. As in the previous embodiment, the cell 8 receives the simple addition output from the cells 5 to 7 as an input, supplies the teacher signal 1, and outputs the cell 8 and the cells 5 to 5.
The coupling coefficient with 7 is fixed at 1 and learning is not performed. FIG. 4 shows the membership function waveforms of cells 5 to 7, and their creation is the same as that of FIG. In this embodiment, the same effect as that of FIG. 1 can be obtained.

[0007]

As described above, according to the present invention, since a teacher signal having a constant value is provided to a cell which outputs a simple addition, learning can be performed so that the total sum of the outputs of the membership functions becomes constant. , A membership function can be obtained in a non-biased arrangement.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an embodiment of an NN according to the present invention.

FIG. 2 is a configuration diagram of another embodiment.

FIG. 3 is an output characteristic diagram of cells 7, 8, and 9 in FIG.

FIG. 4 is an output characteristic diagram of cells 5, 6, and 7 of FIG. 2;

[Explanation of symbols]

1-11 cell _{W 1 ~W 7, WB 1 ~WB} 4 coupling coefficient

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06G 7/60 G06F 9/44 554 G06F 15/18 G06G 7/12

Claims (1)

(57) [Claims] 1. A cell for outputting an input signal and a bias cell, a plurality of first intermediate cells connected to the signal of the cell via a predetermined coupling coefficient, and a plurality of first intermediate cells connected to the intermediate cell and a predetermined coupling coefficient. A neural network having a plurality of second intermediate cells connected thereto and two output cells, and outputting a membership function, wherein a teacher signal having a constant value is applied through one of the output cells. .