JPH0556864A - Cooker - Google Patents

Abstract

PURPOSE:To perform an automatic cooking with an improved finish state of cooking by estimating the temperature of food under cooking in real time through detecting the amount of smoke generated from the food under cooking. CONSTITUTION:A cooker is provided with a heating compartment 1 where food to be cooked is placed, heaters 2 which heat the food to be cooked, a smoke amount detector 5 which detects the amount of smoke in the heating compartment 1, an atmospheric temperature detector 8 which detects the atmospheric temperature in the heating compartment 1, a temperature estimating device 14 that estimates the temperature of food under cooking based on the outputs from the atmospheric temperature detector 8, and a controller 15 which controls the heaters 2 based on the outputs from the smoke amount detector 5 and the temperature estimating device 14. The temperature estimating device 14 has a built-in neural network simulator having two or more fixed combined weighting factors that have been determined through learning in the actual cooking environment. Therefore, the temperatures of food under cooking (surface, central and rear face temperatures) can be estimated regardless of the initial temperature in the heating compartment 1 and the initial temperature and amount of food to be cooked, so that an automatic cooking with an improved finish state can be performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cooking utensil for automatically cooking in a microwave oven, a gas oven, a roaster or the like.

[0002]

2. Description of the Related Art Conventionally, a cooking utensil of this kind, for example, an electric roaster has been constructed as shown in FIG. The configuration will be described below.

As shown in the figure, a heating chamber 1 is for cooking food by putting it in it. A heating supply means (heater) 2 for heating the food is provided and a temperature detecting means such as a thermistor for detecting the internal temperature. 3 is provided. The control means 4 controls the heating supply means 2 based on the information from the temperature detection means 3. In order to automatically cook with such a configuration, it is necessary to know the weight of the food and the initial temperature. Therefore, the output voltage gradient of the temperature detecting means 3 is measured for several minutes after the power is turned on, and if the gradient is steep, it is determined that the weight of the food is light, and if the gradient is gentle, the weight is heavy, and the voltage gradient is determined. The optimum cooking time was determined by multiplying the constant K by The output voltage characteristic of the temperature detecting means 3 is shown in FIG. Figure 1
2 (a) has a light weight, and FIG. 12 (b) has a heavy weight. And I did a lot of cooking experiments and decided the constant.

[0004]

In such a conventional cooking utensil (here, a roaster), the gradient of the ambient temperature in the heating chamber 1 is detected, and the cooking product (for example, fish grill) is detected based on the gradient. The cooking time was decided by judging the weight of), so there was considerable variation in the completion of cooking. For example, the initial temperature in the heating chamber 1 is not always low, and when cooking is performed immediately after finishing the cooking, the initial temperature in the heating chamber 1 becomes very high. In this case, when a heavy food is cooked, the output voltage characteristic of the temperature detecting means 3 becomes as shown in FIG. 13, and the temperature in the heating chamber 1 drops for a moment. This is because even if cooking is started, the temperature in the heating chamber 1 is high, and thus the heat in the heating chamber 1 is absorbed by the food. In such a case, it was difficult to determine the optimum cooking time by the above method. Also, the heating supply means 2
Is a heater, it has a considerable effect on the completion of cooking due to fluctuations in commercial power supply voltage. In other words, the environment at the time of starting cooking (type of food, initial temperature of heating chamber 1,
There was a problem that the finished cooking varied considerably depending on the initial temperature of the cooked food, the power supply voltage, etc.). Furthermore, in terms of grilling fish, it is important that the surface is baked, but the temperature inside the fish will rise by 60-
70 ° C is said to be the best. In consideration of this point, there is a problem that it is very difficult to detect the completion of cooking from the aspect of surface burning and temperature rise inside.

The present invention solves the above problems by estimating the temperature of cooked food in real time from the physical quantity of the environment inside the heating chamber that can actually measure and detect the temperature of the cooked food and detect the amount of smoke emitted from the cooked food as the cooking progresses. The purpose is to indirectly detect the completion of cooking and improve the cooking completion.

[0006]

In order to achieve the above-mentioned object, the present invention has a heating chamber for storing food for cooking, a heating supply means for heating the food, and an amount of smoke in the heating chamber. At least one of the surface temperature, the central temperature, and the back surface temperature of the cooking object based on the output of the atmosphere amount detecting means for detecting the smoke amount detecting means for detecting the atmosphere temperature in the heating chamber, and the atmosphere temperature detecting means. A temperature estimating means for estimating one of them, and a control means for controlling the heating and supplying means based on the output of the smoke amount detecting means and the output of the temperature estimating means. It is provided with a hierarchical neural network model means having a plurality of connection weight coefficients of a fixed neural network for estimating the temperature of the food obtained by a method simulating the constructed neural network. Task It is set to determine means.

[0007]

According to the present invention, by the above-mentioned means for solving the problems, by inputting the ambient temperature information in the heating chamber from the ambient temperature detecting means to the temperature estimating means, all the actual cooking environment to be actually cooked is learned and internally The temperature estimation means having the neural network pattern means having the fixed coupling weight coefficient is
At least one of the surface temperature, the central temperature, and the back surface temperature of the cooked food is estimated momentarily. Further, the smoke amount detecting means detects the actual amount of smoke emitted from the food as cooking progresses. The control unit controls the heating supply unit (heater in the electric roaster) based on the actual amount of smoke and the output of the temperature estimation unit that indirectly detects the temperature, and can recognize the optimum completion of cooking.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An electric roaster according to an embodiment of the present invention will be described below with reference to FIGS. In addition,
The same components as those of the conventional example are designated by the same reference numerals and the description thereof will be omitted.

As shown in the figure, the smoke amount detecting means 5 detects the amount of smoke generated as the cooking inside the heating chamber 1 progresses, and is composed of an infrared light emitting diode 6 and a phototransistor 7. There is. The ambient temperature detecting means 8 is the heating chamber 1
The temperature of the inside atmosphere is detected. In this embodiment, the thermistor is used. The voltage level detection means 9 detects the voltage level of the commercial power supply voltage. The timekeeping means 10 counts the time since the power was turned on. The operation means 11 comprises a category selection key 12 for selecting a category of food and a cooking key 13 for starting / stopping cooking. The temperature estimating means 14 estimates the surface temperature, center temperature, and back surface temperature of the cooked food based on the outputs of the ambient temperature detecting means 8, the voltage level detecting means 9, the time measuring means 10 and the category selection key 12, and the control means. 15
Controls the heating supply means 2 based on the output of the smoke amount detection means 5 and the output of the temperature estimation means 14. The heating means 2 is composed of a heater and is arranged in the heating chamber 1. The display means 16 comprises a fluorescent display tube and displays the remaining cooking time and the like. Further, the A / D conversion means 17 and 18 respectively convert the outputs of the ambient temperature detection means 8 and the voltage level detection means 9 into digital quantities. Operation means 11 and display means 1
6 is configured as shown in FIG.

Since the solving means used in the conventional control method cannot be applied to the means constituting the temperature estimating means 14, the means for simulating an optimum neural network as a multidimensional information processing method is used. There is. In the method that imitates the neural network, a method of using a plurality of connection weight coefficients of the neural network for estimating the temperature (surface temperature, central temperature, back surface temperature) of the cooked food as a fixed table and leaving the learning function There are ways to adapt to the environment and users. The present embodiment is provided with a temperature estimating means 14 having a neural network model means having therein a fixed coupling weighting coefficient for estimating the temperature of a food product obtained by a method simulating a neural network.

Factors that affect the completion of the cooked food include the initial temperature in the heating chamber, the initial temperature of the cooked food, the type (category) of the cooked food, and the voltage level of the commercial power supply. Due to these factors, the completion of the cooked food will vary greatly.

The fixed weighting coefficient fixed in the neural network for estimating the temperature of the cooked food is the surface temperature of the cooked food when cooked in the actual cooking environment (environment in which various factors described above are combined). Collect data on how the center temperature, backside temperature, and ambient temperature in the heating chamber change, and correlate environmental data, ambient temperature data in the heating chamber, and cooking product temperature (front, center, backside) data. It can be obtained by learning the neural network model means. As a neural network schematic means to be used, reference 1 (DE Ramelhart et al., 2 authors, "PDP model" translated by Shunichi Amari, 1989), reference 2 (Kaoru Nakano, 7 authors, "Basics of Neurocomputer") Published by Corona, P102, 1990), JP-B-63-55106
Some of them are disclosed in the official gazette. Hereinafter, the configuration and operation of a concrete neural network schematic means will be described by taking a multilayer perceptron using an error backpropagation method as the most well-known learning algorithm described in Document 1 as an example.

FIG. 3 is a conceptual diagram of a neural element which is a constituent unit of the neural network model means. In FIG. 3, 21 to 2
N is a pseudo synapse coupling converter simulating synaptic coupling of nerves, and 2a is pseudo synapse coupling converters 21 to 2N.
2b is a set nonlinear function, for example, a sigmoid function whose threshold is h, f (y, h) = 1 / (1 + exp (-y + h)) (Equation 1 ) Is a non-linear converter for non-linearly converting the output of the adder 2a. Although omitted because the drawing is complicated, an input line for receiving a correction signal from the correction means is connected to the pseudo synapse coupling converters 21 to 2N and the non-linear converter 2b. Further, the pseudo synapse coupling converters 21 to 2N serve as coupling weight coefficients of the neural network schematic means. This neural element has two types of operation modes, a signal processing mode and a learning mode.

The operation of each mode of the neural element will be described below with reference to FIG. First, the operation of the signal processing mode will be described. Neural elements are N inputs X1 to X
It receives n and outputs one output. i-th side input signal Xi
Is the i-th pseudo-synaptic coupling converter 2 shown by a square
i is converted to Wi · Xi. N signals W1 · X1 converted by the pseudo synapse coupling converters 21 to 2N
Wn · Xn enters the adder 2a, the addition result y is sent to the nonlinear converter 2b, and becomes the final output f (y, h). Next, the operation of the learning mode will be described. In the learning mode, the conversion parameters W1 to Wn and h of the pseudo synapse coupling converters 21 to 2N and the non-linear converter 2b are received, and the correction signals representing the correction parameter correction amounts ΔW1 to ΔWn and Δh are received from the correction means. , Wi + ΔWi; i = 1, 2, ..., N h + Δh (formula 2).

FIG. 4 is a conceptual diagram of a signal converting means constituted by connecting four neural elements in parallel. Needless to say, the following description does not specify that the number of neural elements constituting this signal converting means is four. In FIG. 4, 211 to 244 are pseudo synapse coupling converters,
201 to 204 are addition nonlinear converters that combine the adder 2a and the nonlinear converter 2b described in FIG. 4, the illustration is omitted because the drawing is complicated as in FIG. 3, but the input lines for receiving the correction signal from the correction means are pseudo synapse coupling converters 211 to 244 and an addition nonlinear converter 20.
It is connected to 1-204. The pseudo synapse coupling converters 211 to 244 also serve as coupling weight coefficients. Regarding the operation of this signal conversion means, the operation of the neural element described in FIG. 3 is performed in parallel.

FIG. 5 is a block diagram showing the configuration of the signal processing means when the error back-propagation method is adopted as the learning algorithm, and 31 is the above-mentioned signal converting means. However,
Here, M neural elements for receiving N inputs are arranged in parallel. Reference numeral 32 is a correction means for calculating the correction amount of the signal conversion means 31 in the learning mode. Hereinafter, the operation when learning the signal processing means will be described with reference to FIG. The signal converting means 31 has N inputs S
Upon receiving _in (X), it outputs M outputs S _out (X).
The correction means 32 includes an input signal S _in (X) and an output signal S _out.
Upon receiving (X), the process waits until M error signals δ _i (X) are input from the error calculating means or the signal converting means in the subsequent stage. The error signal δ _i (X) is input and the correction amount is set to ΔW _ij = δ _i (X) · S _iout (X) · (1−S _iout (X)) · S _jin (X) (i = 1 to N, j = 1-M) (Equation 3) is calculated and a correction signal is sent to the signal conversion means 31. The signal conversion means 31 modifies the conversion parameters of the internal neural elements according to the learning mode described above.

FIG. 6 is a block diagram showing the structure of a multilayer perceptron using a neural circuit retina type means.
X, 31Y, and 31Z are signal conversion means composed of K, L, and M neural elements, respectively, and are 32X, 32Y, and 3X.
2Z is a correction means, and 33 is an error calculation means. The operation of the multi-layer perceptron configured as described above will be described with reference to FIG. Signal processing means 3
In 4X, the signal conversion means 31X receives the input S
_{Receives iin} (X) (i = 1 to N) and outputs S _jout (X)
(J = 1 to K) is output. The correction means 32X uses the signal S
_The error signal δ is received by receiving _iin (X) and the signal S _jout (X).
Wait until _j (X) (j = 1 to K) is input. The same processing is performed in the signal processing means 34Y and 34Z, and the final output _Shout (Z) from the signal converting means 31Z.
(H = 1 to M) is output. Final output _Shout (Z)
Is also sent to the error calculation means 33. Error calculation means 33
, The error with the ideal output T (T1, ..., TM) is calculated based on the squared error evaluation function COST (Equation 4), and the error signal δ _h (Z) is corrected by the correction means. 32Z
Sent to.

[0018]

[Equation 1]

However, η is a parameter that determines the learning speed of the multilayer perceptron. Next, when the evaluation function is a square error, the error signal is δh (Z) = − η · (S _hout (Z) −T _h ) (Equation 5). The correction unit 32Z calculates the correction amount ΔW (Z) of the conversion parameter of the signal conversion unit 31Z according to the procedure described above, calculates the error signal to be sent to the correction unit 32Y based on (Equation 6), and the correction signal ΔW (Z) is sent to the signal converting means 31Z, and the error signal δ (Y) is corrected by the correcting means 32.
Send to Y. The signal conversion means 31Z uses the correction signal ΔW (Z).
Modify internal parameters based on. The error signal δ (Y) is given by (Equation 6).

[0020]

[Equation 2]

[0021] Here, W _ij (Z) signal converting means 31Z
Is a conversion parameter of the pseudo synapse coupling converter of. Hereinafter, similar processing is performed in the signal processing means 34X and 34Y. By repeating the following procedure called learning, the multi-layer perceptron produces an output that closely approximates the ideal output T when given an input. In the above description, the three-stage multi-layer perceptron is used, but this may be any number of stages. In addition, reference 1
The modification method of the conversion parameter h of the non-linear conversion means in the signal conversion means and the method of accelerating learning known as the inertia term are omitted for simplification of the description, but this omission is described below. It does not bind the invention.

In this way, the neural network model means uses the environmental data (the initial temperature in the heating chamber, the initial temperature of the cooking product, the commercial power supply voltage level, the type of the cooking product, etc.) when cooking and the ambient temperature data in the heating chamber. By learning the relationship with the temperature (front surface, center, back surface temperature) data of the food, it is possible to express in a natural way the control method that is not easy to describe with simple rules. In this embodiment, the temperature estimation means 11 is configured by incorporating the information thus obtained. Specifically, the temperature estimating means 11 is configured by using only the signal converting means 31X, 31Y, 31Z of the multilayer perceptron after sufficiently learning as a neural network schematic means. The data actually learned will be described.

FIG. 7 shows the characteristics when cooking is performed when the initial temperature of the heating chamber 1 is low, the commercial power supply voltage is 100 V, the type of food is one horse mackerel, and the initial temperature of the food is about 10 ° C. It is a thing. FIG. 7 (a) shows changes in the ambient temperature detecting means 8 (ambient temperature in the heating chamber 1), FIG. 7 (b) is the surface temperature of the cooking product, FIG. 7 (c) is the central temperature of the cooking product, and FIG. 7
(D) has shown the change of the back surface temperature of a cooked material. The temperature of the cooked food is measured with a thermocouple or the like. Figure 8
Indicates that the initial temperature of the heating chamber 1 is high and the commercial power supply voltage is 100
V, the type of food is one horse mackerel, the initial temperature of the food is about 10
It shows the characteristics when cooked at ℃.
FIG. 9 shows that the initial temperature of the heating chamber 1 is low and the commercial power supply voltage 10
0V, the type of food is 4 horse mackerel, the initial temperature of the food is about 1
It shows the characteristics when cooked at 0 ° C. FIG. 10 shows the characteristics when cooking is performed when the initial temperature of the heating chamber 1 is high, the commercial power supply voltage is 100 V, the type of food is 4 horse mackerels, and the initial temperature of the food is approximately 10 ° C. is there.

FIGS. 8 (a) to 8 (d), 9 (a) to 9 (d), and 10 (a) to 10 (d) are shown in FIG. 7 (a).
~ Corresponds to Fig. 7 (d). It can be seen that the output voltage change (atmosphere temperature in the heating chamber 1) of the atmosphere temperature detecting means 8 varies depending on the initial temperature in the heating chamber 1 and the amount of food. Similarly, when the power supply voltage is changed, or when the type of food is changed, the output changes differently. Such an experiment was carried out in the same manner for all combinations of actual cooking environments. Then, the experimental data was input to the neural network model means for learning. That is, the heating chamber 1 of the ambient temperature detecting means 8 is connected to the neural network model means.
Atmosphere temperature information inside, atmosphere temperature information as one minute before the current time as atmosphere temperature gradient information, commercial power supply voltage level information of voltage level detecting means 9, and elapsed time information from power-on time obtained from clocking means 10. And 5 pieces of category information obtained from the category selection key 12 and 3 pieces of information of the surface temperature, the center temperature and the backside temperature of the cooked food as ideal outputs are input and learned, and the neural network model means The signal conversion means 31X, 31Y, 31Z are established, and they are incorporated in the surface temperature estimation means 14 as a neural network schematic means.

Next, the operation will be described based on the system configuration diagram shown in FIG. First, the food to be cooked is placed in the heating chamber 1, and the cooking category is selected by the category selection key 12 of the operation means 11. Then, cooking is started with the cooking key 13. The category information is input to the temperature estimation means 14 via the control means 15. The control means 15 outputs a signal to start timing to the timing means 10 and also outputs a heating start signal to cause the heating supply means 2 to generate heat.
The timing information of the timing means 10 is input to the temperature estimation means 14. The output of the ambient temperature detecting means 8 is digitally converted by the A / D converting means 17 and the ambient temperature information in the heating chamber 1 is input to the temperature estimating means 14 every moment. Further, the voltage level information of the commercial power supply voltage from the voltage level detecting means 9 is digitally converted by the A / D converting means 18 and input to the temperature estimating means 14. The temperature estimating means 14 momentarily estimates the surface temperature, the center temperature, and the back surface temperature of the cooking product based on these input signals and information,
The information is output to the control means 15. Further, in the smoke amount detecting means 5, infrared light is constantly emitted from the infrared light emitting diode 6 and is received by the phototransistor 7. When smoke is generated from the cooking product and fills the heating chamber 1,
In response to the smoke amount, the infrared light from the infrared light emitting diode 6 is blocked by the smoke in the phototransistor 7, and the amount of light received is reduced. The phototransistor 7 inputs the output as a voltage to the control means 15. The control means 15 controls the heating supply means 2 based on the smoke amount information and the estimated temperature information.

Further, the time measuring means 10, the temperature estimating means 14,
Although all the control means 15 are composed of 4-bit microcomputers, it is of course possible to form them by one microcomputer. It should be noted that the temperature estimation means 14 includes temperature gradient information of the ambient temperature detection means 8 (two pieces of information at the current time point and one minute ago), voltage level information of the commercial power supply voltage obtained from the voltage level detection means 9, and the time counting means 1.
Five pieces of information, namely, elapsed time information after power-on obtained from 0 and category information of cooking obtained from the category selection key 12 are inputted, but this limitation does not restrict the present invention. Further, in this embodiment, the electric roaster is used as the cooking utensil, but a microwave oven, a gas oven or the like may be used.

As described above, according to the present embodiment, the temperature estimation incorporating the neural network model means having a plurality of fixed coupling weight coefficients of the neural network already learned in the environment of the heating chamber where the cooking is actually performed Since the device is provided with a smoke amount detecting device for detecting the amount of smoke generated from the cooked product, the finished state of the cooked product can be detected by the surface temperature, the center temperature, the back surface temperature and the amount of smoke. Therefore, the cooking state can be improved as compared with the conventional example, and optimum automatic cooking can be realized.

[0028]

As is apparent from the above embodiments, according to the present invention, there is provided a heating chamber for storing food to be cooked,
Based on the output of the heating supply means for heating the cooking product, the smoke amount detection means for detecting the amount of smoke in the heating chamber, the ambient temperature detection means for detecting the ambient temperature in the heating chamber, and the ambient temperature detection means. Since the temperature estimation means for estimating the temperature of the cooked food and the control means for controlling the heating supply means based on the output of the smoke amount detection means and the output of the temperature estimation means are provided, the completion state of the cooked food is recognized. Therefore, it is possible to indirectly detect the actual amount of smoke generated from the cooked food and the actual temperature of the cooked food, and compared with the conventional method that determines the optimum cooking time from the gradient such as the ambient temperature.
The finished state of cooking can be improved.

Further, since the temperature estimating means estimates at least one of the surface temperature, the central temperature and the back surface temperature of the cooked food, the whole finished state of the cooked food from the surface to the inside can be known. ..

Further, the temperature estimation means is a fixed nerve for estimating the temperature (surface temperature, central temperature, back surface temperature) of the food obtained by a method simulating a neural network composed of a plurality of neural elements. Since the neural network model means having a plurality of coupling weight coefficients of the network therein is provided, or the hierarchical neural network model means constructed by combining layers composed of a plurality of neural elements in multiple layers is provided. The temperature of the cooked food can be estimated regardless of the initial temperature in the heating chamber, the initial temperature of the cooked food, the amount of the cooked food, etc., and automatic cooking is possible.

Further, the environmental physical quantity detecting means for detecting the environment in the heating chamber is provided with a temperature detecting means for detecting the ambient temperature in the heating chamber, and the temperature detecting means indirectly estimates the temperature of the cooking product. It is not necessary to actually use a sensor for detecting the temperature of the food (for example, a pyroelectric infrared sensor that measures the surface temperature of the food in a non-contact manner, or a temperature sensor that directly contacts the food), and the cost can be reduced.

[Brief description of drawings]

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

FIG. 2 is a front view of an operation unit and a display unit of the cooking utensil.

FIG. 3 is a conceptual diagram of a neural element that is a constituent unit of a neural network schematic means used in the cooking utensil.

FIG. 4 is a conceptual diagram of a signal conversion means composed of neural elements used in the cooking utensil.

FIG. 5 is a block diagram of a signal processing unit that employs an error backpropagation method as a learning algorithm used for the cooking utensil.

FIG. 6 is a block diagram showing a configuration of a multilayer perceptron using a neural network schematic means used in the cooking utensil.

FIG. 7 is a diagram showing an example of experimental data of the cooking utensils (a) to (d).

FIG. 8 is a view showing another example of experimental data of the cooking utensils (a) to (d).

9A to 9D are diagrams showing another example of the experimental data of the cooking utensil.

FIG. 10 is a diagram showing another example of experimental data of the cooking utensils of (a) to (d).

FIG. 11 is a system configuration diagram of a conventional cooking utensil.

12A and 12B are views showing an example of experimental data of the cooking utensil.

FIG. 13 is a diagram showing another example of experimental data of the cooking utensil.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Heating chamber 2 Heating supply means 5 Smoke amount detection means 8 Atmosphere temperature detection means 14 Temperature estimation means 15 Control means

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Nakason Son 10-1 Higashisanda, Tama-ku, Kawasaki-shi, Kanagawa Matsushita Giken Co., Ltd.

Claims (4)

[Claims] 1. A heating chamber for storing food for cooking, a heating supply means for heating the food, a smoke amount detecting means for detecting the amount of smoke in the heating chamber, and an ambient temperature in the heating chamber. For detecting the temperature of the cooking product based on the output of the ambient temperature detecting means, and the heating supply based on the output of the smoke amount detecting means and the output of the temperature estimating means. And a control means for controlling the means. 2. The cooking utensil according to claim 1, wherein the temperature estimating means estimates at least one of a surface temperature, a center temperature and a back surface temperature of the food. 3. The temperature estimating means internally includes a plurality of coupling weight coefficients of a fixed neural network for estimating a temperature of a food obtained by a method simulating a neural network composed of a plurality of neural elements. 2. The cooking utensil according to claim 1, further comprising a neural network model means included in. 4. The cooking utensil according to claim 1, wherein the temperature estimation means comprises a hierarchical neural network schematic means constructed by combining layers composed of a plurality of neural elements in multiple layers.