The Result of Experiment on Neural Network Leaning of Human Art Drawing

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

J. Environ. Treat. Tech.

ISSN: 2309-1185

Journal web link: http://www.jett.dormaj.com

The Result of Experiment on Neural Network

Leaning of Human Art Drawing

1

*

2

3

Svetlana V. Veretekhina , Maxim A. Kudryavtsev , Vladimir L. Simonov , Elena V.

4

5

Potekhina , Tatyana V. Karyagina

¹Department of Information Systems, Networks and Security, Russian State Social University, Moscow, Russia

2

Postgraduate Study Mathematical Support of Computers, Complex and Computer networks, Russian State Social University, Moscow, Russia

3

Department of Information Systems, Networks and Security, Russian State Social University, Moscow, Russia

4

5

Department of Mathematics and Computer Science, Faculty of Information Technologies, Faculty of Information Technologies, Russian State Social

University, Moscow, Russia

Department of Mathematics and Computer Science, Faculty of Information Technologies, Faculty of Information Technologies, Russian State Social

University, Moscow, Russia

Received: 13/09/2019

Accepted: 22/11/2019

Published: 20/12/2019

Abstract

In the article, the authors consider one of the most pressing problems of our time. In the first part of the article authors describes

the most modern and effective heuristic and multi-heuristic algorithms that can be used in NP-hard problems solving. The most

effective algorithms considered in order to produce answers that are as close as possible to the correct solutions in a satisfactory time.

In the second part, the authors develop the idea of us in intelligent algorithms and methods, giving examples of machine learning of

neural networks. In particular, it raises the question about assessing the quality of the neural networks in both on classic templates

and in the most difficult situations concerning the creative process. A large number of experiments were conducted as a part of the

work, and the data was collected in the process of the neural network training such as its reaction to individual objects. The results

are briefly sorted and presented in the form of tables intended for the further analyze, in order to develop mathematical models and

algorithms for assessing the quality of the learning process for modern neural networks.

Keywords: Artificial intelligence, Neural networks, Machine learning, Deep learning, Heuristic algorithms, Cuckoo algorithm,

Agent approach, Multi-heuristic approach, Learning optimization, Particle swarm optimization

1

spiritual values or the result of creating objectively new. The

1

Introduction

main criterion that distinguishes creativity from

manufacturing (production) is the uniqueness of its result.

The result of creativity can`t be directly derived from the

initial conditions. No one, except perhaps the author, can get

exactly the same result if you create the same initial situation

for him. Thus, in the process of creativity, the author invests

in the material, in addition to labor, some irreducible to labor

operations or the logical conclusion of the possibility,

expresses in the end some aspects of his personality. It is this

fact that gives the products of creativity additional value in

comparison with the products of production. In creativity, not

only the result, but the process itself is valuable.

.1 Artificial Intelligence and Creativity

The history of Artificial intelligence showed several

stages of its development. First stage: Early enthusiasm and

high expectations. The second stage: Machine learning is a

period of development of artificial intelligence with the

advent of large amounts of data. The third stage: Deep

training of artificial intelligence is the next stage of

development, new opportunities. England mathematician and

philosopher Alan Turing is the founder of artificial

intelligence. He developed a theoretical model of a computer.

Modern programming languages, such as C, Pascal, Java, are

equivalent to a Turing machine. Modern technologies are

considered intelligent if they allow to solve creative

problems. Definition from Wikipedia: "Creativity is a process

of activity that creates qualitatively new materials and

There are different kinds of creativity: technical,

scientific, social, philosophical, cultural, pedagogical, artistic,

musical, sports, game, etc. All kinds of creativity correspond

to the types of spiritual human activity. All kinds of human

creativity give

a NEW output (discovery, invention,

Corresponding autor: Svetlana V. Veretekhina, Department

of Information Systems, Networks and Security, Russian

State Social University, Moscow, Russia. E-mail:

veretehinas@mail.ru.

innovation). Creativity and personality are inextricably

linked. N.A. Berdyaev (2003) (4) writes: "Personality is not a

substance, in the creative aspect".

1

050

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Neurobiological bases of creativity are connected with

the work of neural networks of the nervous system (9).

Creativity and personality are inextricably linked. Reflecting

on the nature of unconscious creativity described in (1).

Roem Wallace in 1926 identified four stages of creative

thinking: preparation - incubation - insight - test (6, 20).

Insight - a poly-semantic term from the field of zoo-

psychology, psychology and psychiatry, describing the

complex mental, the essence of which is unexpected, partly

intuitive to understand the problem and "sudden" finding its

solution. Insight is not an Epiphany of logic. Research in this

area belongs to scientists: Karl Dunker, Max Wertheimer,

Professor of the University of Gottingen - Wolfgang Koehler.

Artificial intelligence is a property of intelligent systems

to perform creative functions, which are traditionally

considered the prerogative of man; science and technology of

creating intelligent machines, especially intelligent computer

programs (2).

NATO experts give the following definition of Artificial

intelligence: "the Capacity provided by the optimal or

suboptimal choice from a wide space of opportunities to

achieve goals by implementing strategies, based on training

or adaptation to the environment" (19).

Modern artificial intelligence - is the use of various

algorithms, software analysis of big data (input) and the final

solution (output),

Artistic creativity is made on the basis of the laws of artistic

and figurative reflection of reality.

Can artificial intelligence draw anything NEW? To begin

with, we will discern whether artificial intelligence is able to

recognize a human figure. The research was conducted at the

Russian state social University.

Students performed

laboratory work on the training of artificial intelligence on

the site https://quickdraw.withgoogle.com. "Can a neural

network learn to recognize patterns?". The main task of the

study is to create the largest set of data in the form of

drawings and help in the development of machine learning

technologies. The neural network offered a person to draw an

object in 20 seconds. For example: mailbox, glass of wine,

kitchen stove, palm, line, square, face, belt, ocean, mobile

phone, etc. - an infinite number of items. The study involved

44 people, students of higher education (compiled a list of

participants of the experiment). Each student completed 10

tasks. One task included 6 drawings. The students were

making drawings. The neural network recognized some of the

drawings, but did not recognize some of the drawings. During

the experiment, the reasons why the neural network could not

recognize the pattern were revealed. 375 unique objects were

analyzed. The most common drawings are square, bear,

butterfly, camel. More than 10 times students used 73

objects. Unique objects (participated in the experiment 1-2

times) – 35 pieces. As a result, it was necessary to determine

the level of thinking of the neural network. According to the

results of the experiment, a report was formed on the

recognized figures (excellent figure).

The most productive algorithms for planning are:

1

2

3

4

5

6

.

Genetic algorithm (3, 10, 16).

Two-phase neural networks (14).

Differential evolution (11, 12, 24).

Particle swarm optimization (5, 8, 13, 15, 23, 25).

The sampling similarity algorithm (21).

Hybrid differential evolution and sequential

quadratic programming algorithm.

7

8

9

.

Artificial bee colony algorithm (22)

Optimization learning algorithms (7)

Cuckoo algorithm for solving problems in the

energy sector (18, 26).

Cuckoo algorithm is one of the most modern meta-

heuristic algorithms, which has become popular due to

optimization processes in the field of energy. The cuckoo

algorithm is widely and successfully applied to various

technical optimization problems, such as economic load

optimization, hydrothermal planning, and distributed network

reconfiguration.

With the use of artificial intelligence, mathematical

calculations are carried out according to the algorithms

incorporated in the program and performed by various

software. Artificial intelligence, equipped with mathematical

tools, is able to solve complex science-intensive problems.

The authors in their study put forward a hypothesis about

how much we train artificial intelligence in the field of

artistic creativity (on the example of the analysis of human

drawings).

Figure 1: Neural network result of recognizing the man drawings.

So, the subjects performed 10 iterations (approaches) of 6

figures. As a result of the research, conclusions are drawn.

The analysis of unrecognized drawings is carried out.

According to the results of 25 experiments, the data are

summarized. Each subject drew conclusions on the basis of

mental (speculative) conclusions and equated the level of

training of the neural network to the initial drawing skills of

the child, thereby determining the level of thinking of the

neural network, equating it to the age of the child.

2

Methodological Framework

Artistic creativity is a special human activity that

Experiment 1. In the experiment, 60 pictures, among the

generates a qualitatively new work and is distinguished by

uniqueness, originality and socio-historical uniqueness.

unidentified (not witnesses) of figures

Unrecognized description of the drawings:

– 13 pieces.

1

051

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

1

2

3

4

5

6

7

8

9

1

.

Squirrel – poorly drawn

3)stitches – could not portray

4)snorkel with the mask depicted, the system didn't

recognize

Baseball bat– the system is not yet trained

Matches - the system is not yet trained

Nail - system not yet trained

The aircraft system is not yet trained

Crocodile - the system is still not trained

The tube with the mask is poorly drawn

Foot – badly drawn

5) bat– the system did not recognize

6)ladder – the system does not recognize

7)the dog – the system didn't recognize

8)fireplace – the system does not recognize

9)bed – the system didn't recognize

10) rabbit - the system didn't recognize

11)basketball was not able to portray

12)pie – system not recognized

The key system is still not trained

0. Watermelon - system is still not trained

1. The aircraft system is not yet trained

2. Avocado - system is still not trained

3. Bandage - badly drawn

13)the steak was not able to portray

Result: at the moment the system corresponds to a 5-year-

old child. In my opinion, AI corresponds to the development

of a child aged 5 years. Most of the drawings are not

recognized due to the fact that I am very bad with a computer

mouse, the consequence of this is the inability to determine

the figure of AI. System thinking is appropriate to a 5 y.o.

child.

Result: as a result of the experiment, 60 drawings were

made, among the unidentified (not understood drawings) – 13

pieces. The reason for the system error is that the system is

not yet trained or the person did not draw well.

System thinking is appropriate to a 2-4 y.o. child.

Experiment 2. In the experiment, 60 pictures, recognized

-

39, unrecognized - 21. Basically, the problem of not

Experiment 8. As a result of the experiment, 60

drawings were made that were not recognized by the system -

17. Unidentified drawings: Sink, Candlestick, Train, Green

beans, Trombone, Telephone, Crab, Fish, air conditioning,

Foot, Cat, Stairs, Letter, Fan, Raccoon, Cruise liner, Balloon.

Describe the problem: Some pictures were not recognized.

Several figures the system could not recognize due to the

small number of samples in the database.

recognizing the picture was in poor-quality drawings of a

person. Intelligence corresponds to a child of about seven (7)

years.

Result: System thinking is appropriate to a 7y.o.child.

Experiment 3. As a result of the experiment, 16

iterations of 6 figures and 1 iteration with 4 figures (the total

number of 100 figures) were made. Among unidentified (not

understandable drawings)

–

41 units. Description of

Result: artificial intelligence corresponds to the

development of the child at the level of 14-18 years. The

system is not yet trained, maybe the person painted badly.

System thinking is appropriate to a 14 y.o. child.

Experiment 9. As a result of the experiment, 10

iterations were carried out. Among the 60 drawings, the

system did not recognize 19 drawings. The main reason is the

lack of skill of the artist or not enough time to portray the

object.

Result: the neural Network is not sufficiently trained, as

many simple, accurately drawn objects, the neural network

did not recognize. In my opinion, artificial intelligence

corresponds to the development of a child at the age of 6

years. System thinking is appropriateto a 6 y.o. child.

Experiment 10. As a result of the experiment, 10

iterations were carried out, the total number of drawings – 60.

Not recognized by the system – 9 (dishwasher, trombone,

angel, lantern, Scorpion, bracelet, leaf, foot).

Result: the Problem of non-recognition, in most cases,

was the “human problem”, that is, the wrong interpretation of

the picture, although we should not exclude the fact that the

system is not fully trained in the recognition of individual

elements. System thinking is appropriate to a 5 y.o. child.

Experiment 11. As a result of the experiment, 10

iterations were carried out. Of the total number of drawings

44 are recognized, not recognized by the system 16.

Unrecognized pictures: BlackBerry (the problem of man), the

hand (the problem ISS), a calculator(problem ISS), air

conditioning (man's problem), Wallet (the problem of man),

toe (problem ISS), bandage (the problem ISS), the bulb (the

problem ISS), the balloon (the problem ISS), the bed (the

problem ISS), sleeping bag(man's problem), nail (problem

unrecognized drawings - the system is not yet trained or

poorly drawn. Thinking corresponds to a child of 6 years. The

system didn't recognize 41 out of 100.

Result: System thinking is appropriate to a 6 y.o.child.

Experiment 4. As a result of the experiment, 10

iterations were carried out. Among the 60 drawings, 14 were

not recognized, among which the first problem is the problem

of the information system, and 13 poorly drawn images.

Artificial intelligence corresponds to a 3-year-old child

Result: System thinking is appropriate to a 3 y.o.child.

Experiment 5. As a result of the experiment, 10

iterations were carried out, where 6 figures were drawn in

each. Among 60 drawings, 38 images were recognized, 22

images were not recognized. Some images were not

recognized due to the fault of the person, namely: 3 images

could not be performed by the person from 22 unrecognized.

Artificial intelligence corresponds to a child – 5 years.

Result: System thinking is appropriate to a 5 y.o. child.

Experiment 6. The experiment was carried out 10

iterations of 60 drawings unrecognized 6 figures (10%), the

cause of the problem person and 1 picture in error AI (1.6%).

AI corresponds to a child under 10 years of age.

Result: System thinking is appropriate to a 10 y.o. child.

Experiment 7. As a result of the experiment, 10

iterations were carried out, among 60 drawings. Not

recognized - 16 drawings – because of the low skill of

drawing a person. Of the 100 drawings, the system did not

recognize 42. The system is not perfect and needs to be

improved. Of the 60 drawings made, the system could not

recognize 13 drawings, namely:

1

2

)the toe is not able to portray

)bird – the system didn't recognize

1

052

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

ISS), goatee(the problem of man), avocado (the problem

ISS).

subjects; 3) incorrect drawings. In my opinion, the neural

network corresponds to the development of a 10y.o. child.

Experiment 17. When training the neural network, as a

result of the experiment, 10 iterations were carried out the

total number of drawings – 60, not recognized by the system -

11. Unrecognized drawings and description of the problem:

Car – inability of the system to recognize the pattern

Bird - the inability to portray a figure for 20 seconds

Car pickup - inability to draw a picture

Result: Most of the drawings are not recognized because

of the limitations of artificial intelligence. In my opinion, IIS

corresponds to the development of a 4y.o. child.

Experiment 12. When training the neural network, as a

result of the experiment, 10 iterations were performed. Of the

total number of drawings (60); not recognized by the system

14. List of unrecognized pictures: whale (the problem of

man), wax crayon (the problem of the neural network), fire

hydrant (the problem of man), tractor (the problem of the

neural network), green beans (man's problem), ceiling fan

Truck – inability of the system to recognize the symbol

Sock – the inability of the system to recognize a slightly

inaccurate character

(

problem people),

a

postcard (the problem of man),

Bottle cap– inability to draw a picture in 20 seconds

Teapot - inability of the system to recognize the symbol

Underwear – inability of the system to recognize the

symbol

Bird - the inability to portray a figure for 20 seconds

Envelope – inability of the system to recognize the

pattern

watermelon (the problem of the neural network), camera (the

problem of man), stitch (the problem of the neural network),

pencil (the problem of man), Zebra (problem people), the

bulb (the problem of the neural network), the hand (man's

problem). In 9 out of 14 cases, the human factor, which

means the wrong picture, was the fault of the incorrect

definition of the figure.

Bracelet – incorrect question and the inability of the

Result: in my opinion, the neural network corresponds to

the development of a 13 y.o. child.

Experiment 13. When training the neural network, as a

result of the experiment, 10 iterations were carried out. Of the

total number of drawings - 60, not recognized by the system –

system to isolate the desired element.

Result: the neural network corresponds to the

development of a child -10 years. Since some complex

characters are recognized, true, and some are not.System

thinking is appropriate to a child of 10 years.

Experiment 18. When training the neural network, as a

result of the experiment, 10 iterations were carried out.

Among unidentified (not understandable drawings) – 29. The

system is not trained or the person drew badly.

Conclusion: the Thinking of the neural network

corresponds to a child 2-4 years old, because the neural

network did not recognize 29 of the 60 drawings. System

thinking is appropriate to a 2-4 y.o. child.

Experiment 19. When training the neural network, as a

result of the experiment, 10 iterations were carried out. The

system did not recognize 13 of the 60 drawings. The system

did not recognize:

bird - system determined that it is a diamond

purse - 2 times - the system determined that it is a socket

and toaster

12. Unrecognized drawings – Crab (the neural network is not

trained), rabbit (it is not clear drawn), sleeping bag (it is not

clear drawn), violin, key and Dolphin (the neural network is

not trained), Inbox (dropped twice, was drawn in two

different ways, the neural network is not trained), it is unclear

drawn - eraser, peanuts, mermaid and protein. Description of

the problem – in 6 cases, the neural network is not trained -

could not recognize the pictures, in the other 6 cases there

was not enough time or an error was made.

Result: in my opinion, the neural network corresponds to

the development of a 14 y.o. child.

Experiment 14. When training the neural network, as a

result of the experiment, 10 iterations were carried out. Of the

total number of drawings are not recognized by the system

19.

Result: in my opinion, the neural network corresponds to

the development of a 8y.o. child.

baseball - the system has determined that it is a tennis

racket

Experiment 15. When training the neural network, as a

result of the experiment, 10 iterations were carried out. Of the

total number of drawings in 60, not recognized by the system

snail - the system determined that the drawn spoon

postcard - the system has determined that it is an

envelope

-

17. Unidentified: headphones, bandage, pie, bulldozer,

hurricane - the system determined that the tornado is

drawn

ocean - the system determined that the pool is drawn

violin - the system determined that the cello is drawn

toilet - the system determined that the dishwasher is

drawn

anvil, sword, purse, Cup, BlackBerry, sea turtle, anvil, green

beans, cruise ship, guitar, calculator, wax chalk, asparagus.

Result: the Problem: 1) "did Not know how to draw an

object"; 2) Incorrectly drew the object; 3) Not enough time

for the execution of the figure. In my opinion, the IIA

corresponds to the development of a 7-8y.o. child.

Experiment 16. When training the neural network, as a

result of the experiment, 10 iterations were carried out. Of the

total number of drawings – 60, unrecognized by the system-

blueberries - the system has determined that the painted

peas

the laptop system has determined that the painted oven

raccoon system determined that the dog is drawn

Result: the Intelligence of this system is at the level of 5-

6 year old child. Some of the drawings, the system failed to

recognize their fault, as was the figure the user is drawing

bad. System thinking is appropriate to a 5-6 y.o. child.

1

2. Unrecognized drawings: Zebra, knee, lighthouse, goatee,

foot, purse, BlackBerry, watermelon, ice cream, basketball,

map, scissors.

Result: the Reasons for the failures: 1) could not

complete the picture in 20 seconds; 2) ignorance of some

1

053

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Experiment 20. The neural network did not recognize

from 60 figures 19. From observations, it can be concluded

that the trained neural network corresponds to a child of 8

years.

Result: system thinking is appropriate to a child of 8

years.

Experiment 21. Thinking of IIS is running at 58%. The

system did not recognize 25 of the 60 drawings.

Result: it is not possible to determine the level of thinking

of the neural network.

Experiment 22. When training the neural network, as a

result of the experiment, 10 iterations were carried out.



Human factor. The input data sent by the user

does not correspond to the actual image of the object;

Neural network memory. The images of objects

recorded in the neural network memory have different ways

of representation (a three-dimensional object, two-



a

dimensional object, a minimalist style of representation, etc.),

which prevents the comparison of the input data with the data

in memory.

Result: it is impossible to determine the level of thinking

of the neural network.

Experiment 25. The results Produced by 10 iterations of

communication with the artificial intelligence of the neural

network. Unrecognized drawings – 15. The system didn't

recognize: the Boat – the system does not recognize the Tree

– the system does not recognize the Grapes – badly drawn,

the Rake – the system does not recognize the Torch – the

system does not recognize Omar– poorly drawn, Swing-

Board – incorrectly drawn, the Crocodile is poorly drawn, the

Great wall of China – incorrectly drawn, the BlackBerry

system does not recognize the Owl is poorly drawn, the

Ladder system didn't recognize, Green beans – the system

does not recognize Jacuzzi – the system does not recognize

the Nail system is not recognized.

Among unidentified (incomprehensible) drawings

– 3.

Description of unrecognized drawings: 1. Cloud. (the neural

network is not trained.)2. Giraffe. (Man poorly painted).3.

Water slide. (Man poorly painted. The level of neural

network development corresponds to the level of a 12-year-

old child.

Result: System thinking is appropriate to a 12y.o. child.

Experiment 23. When training the neural network, as a

result of the experiment, 10 iterations were carried out.

Among the unidentified (not clear) of figures – 13 pieces.

The system is not yet trained or the person has drawn badly.

Thinking of ICI corresponds to a child of 2 years. The neural

network did not recognize 13 of the 60 drawings.

Result: System thinking is appropriate to a 2 y.o. child.

Experiment 24. When training the neural network, as a

result of the experiment, 10 iterations were carried out.

Among the unidentified (not clear) of figures -7. The system

is not yet trained or the person has drawn badly.

Result: the Level of development of artificial intelligence

of the neural network is at the level of a 5-year-old child. In

most cases, the drawings were not recognized due to the lack

of training of the neural network. Thinking corresponds to a

5y.o. child.

Summarizing the results of the experiments, the

information processing automated.

Cause of error:

Table 1: Presents statistical data of 25 experiments.

The number of correct answers

+

The number of wrong answers

(empty)

1

0

1

+

-

Result

Picture description

Avocado

Bus

(empty)

-

1

+

1

-

_

+

4

-

5

8

7

4

8

3

4

5

4

2

3

6

9

6

4

1

12

8

10

7

14

5

7

8

3

7

Car

Pickup car

Shark

1

2

1

2

1

2

1

2

1

Barn

Rollercoaster

Pineapple

Angel

Peanut

Watermelon

Harp

Butterfly

Banana

Paintcan

Drum

1

2

1

3

1

5

1

2

7

14

10

6

1

8

12

10

9

11

3

1

Drums

6

5

1

2

Basketball

Swimming pool

Baseball

Baseballbat

Squirrel

Binoculars

Bat

3

2

3

5

2

1

2

1

6

5

1

3

1

4

5

Wineglass

Hospital

4

3

1

054

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Beard

Boot

Bracelet

Diamond

Broccoli

Alarm

Bulldozer

Boomerang

Bottleofwine

Vase

3

6

3

7

4

9

4

3

5

8

6

2

6

7

2

6

3

2

3

9

1

2

1

3

4

1

10

8

11

7

5

9

6

8

9

6

3

8

13

7

6

7

13

1

4

6

9

6

7

8

3

5

4

11

8

7

4

6

5

3

10

8

4

6

5

11

8

6

10

5

12

4

2

8

1

3

1

3

Tub

Bucket

GreatChinesewall

Bike

1

Fan

Camel

Helicopter

Fork

Grape

1

4

2

1

3

1

2

1

Cello

1

2

Waterslide

Waterslides

Balloon

Waxchalk

Octagon

Bow tie

Hamburger

Dumbbell

Guitar

Eye

Mountain

Pea

Rake

Mushroom

Truck

Pear

Door

Dolphin

Tree

Jacuzzi

Sofa

Rain

House

Dragon

Drill

Oven

2

6

8

4

3

6

7

3

1

10

6

2

5

1

2

9

7

1

6

4

5

10

3

8

4

2

4

1

2

1

2

1

2

1

2

3

1

Hedgehog

BlackBerry

Raccoon

Giraffe

Fence

1

3

1

2

1

Brewingteapot

Curl

2

3

1

Lock

Castle

Star

Zebra

Greenbeans

Beans

Zigzag

Snake

Stop” sign

Umbrella

Tooth

1

2

1

3

6

3

4

2

4

2

1

11

5

2

1

“

1

3

1

055

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Toothpaste

Toothbrush

Yoga

6

5

2

3

2

8

6

2

4

5

6

1

3

4

1

14

1

10

6

3

5

3

10

11

7

6

9

1

4

10

6

7

16

6

1

6

1

8

1

6

12

9

10

7

12

10

6

10

4

1

7

5

11

3

Cactus

1

Calendar

Calculator

Camera

Stone

Camouflage

Canoe

Pencil

Map

Potato

Frenchfries

Boat

2

3

1

2

1

2

1

4

1

2

1

2

Swing

SwingBoard

Square

kangaroo

Brush

paintbrush

Brush

1

2

1

2

3

2

3

Whale

5

1

3

8

5

4

10

7

4

WhaleWall

Keyboard

Clarinet

Strawberry

Key

GolfClub

Book

Knee

1

2

1

3

2

1

Wheel

1

Mosquito

Houseplant

Dresser

Computer

Compass

Computer

Envelope

Conditioner

Basket

Cow

Crown

Fire

CoffeeCup

Purse

3

7

3

1

2

1

3

1

4

5

3

4

1

4

1

2

1

2

1

4

6

2

4

11

4

1

4

2

1

2

1

3

9

Cat

Crab

Bed

8

2

4

1

10

9

1

2

1

Bed

Crocodile

Rabbit

Circle

Mug

CruiseShip

CruiseLiner

Cover

1

8

3

6

10

4

1

5

11

3

8

12

1

2

1

2

1

BottleCap

Bush

Palm

Bulb

Eraser

1

4

1

6

3

4

9

4

6

5

1

2

4

1

5

1

3

1

Swan

Lion

3

1

056

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Candy

Stairs

Years. Mouse

FlyingSaucer

Bat

Line

Sheet

Persons

Person

Spoon

Elbow

Shovel

Horse

Bulb

Moon

Frog

Ambulance

Lighthouse

Bear

5

3

1

6

11

1

10

5

12

1

11

6

4

5

14

7

9

7

2

15

5

6

8

6

4

1

7

8

11

7

6

8

3

7

6

8

6

2

8

3

2

7

6

7

6

3

1

7

12

4

1

10

7

3

12

8

3

1

2

1

4

1

6

4

3

1

8

5

3

4

9

2

7

1

4

1

2

3

4

1

2

1

3

1

3

2

1

2

1

2

4

3

2

4

8

1

Mill

Broom

Sword

1

2

1

2

3

AnimalMigration

Microwave

MicrowaveOven

Microphone

MobilePhone

Brain

1

4

7

6

4

3

4

3

2

5

3

5

2

4

2

4

1

2

3

1

2

1

Lightning

Hammer

Mona

2

1

2

3

2

Carrot

IceCream

SeaTurtle

Most

Motorcycle

Ant

1

3

1

Mouse

Anvil

2

1

2

1

WristWatch

Headphones

Skyscraper

Underwear

Foot

Nail

Scissors

Nose

1

2

1

2

1

2

1

2

1

2

1

Socks

Sock

4

7

2

3

1

Rhinoceros

Laptop

Monkey

Marmoset

Cloud

Sheep

Necklace

Ocean

Lobster

Octopus

Screwdriver

Glasses

Tent

1

2

1

7

6

3

5

1

5

1

4

6

2

1

2

3

1

6

3

6

8

1

Finger

1

9

1

057

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Toe

Palm

Panda

Parachute

SailingShip

Pliers

Passport

Pliers

Spider

Pen

Plumage

Hourglass

Cookie

4

7

5

6

1

7

2

7

5

1

8

1

5

2

3

1

7

9

7

9

1

9

7

8

7

1

9

3

9

2

9

6

1

3

2

1

2

1

2

Jacket

Pickup

Saw

Penguin

Pie

2

1

2

1

2

1

2

1

2

4

Letter

Pizza

1

2

1

2

1

5

11

4

7

2

8

7

1

3

5

13

1

10

8

6

7

11

8

13

7

4

8

9

8

4

5

11

8

3

1

6

4

5

2

1

TeddyBear

Beach

Bandage

Submarine

Candlestick

Pillow

3

1

5

1

5

1

7

1

3

6

1

7

5

3

5

4

7

1

3

5

2

4

2

4

2

9

Pillow

Train

1

FireEngine

FireHydrant

PoliceCar

Pomade

Doughnut

Parrot

Dishwasher

CeilingFan

PostCard

Mailbox

Belt

AccessLadder

Pond

Bird

Console

Gun

5

1

2

1

2

1

2

1

2

1

2

1

2

1

2

1

2

1

2

3

1

2

1

2

1

2

1

2

1

Bee

Radio

Rainbow

Sink

PictureFrame

River

Socket

RollerSkates

Roth

5

1

2

1

5

2

9

7

1

4

3

4

7

4

3

1

2

10

12

4

8

6

7

8

7

3

2

1

Hand

1

Mouthpiece

Mermaid

Fish

Backpack

Garden

GardenHose

Saxophone

Aircraft

1

3

2

1

2

1

2

1

7

5

1

058

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

TrafficLights

2

5

2

10

5

8

9

4

1

3

7

8

2

4

8

4

3

1

7

3

2

1

2

7

6

14

10

9

11

8

1

7

12

2

5

11

5

11

7

8

1

6

5

1

10

8

10

3

8

Candle

Pig

Sweater

Heart

Bench

Skateboard

Pan

Griddle

Ambulance

Scorpion

Clip

Violin

Remand

Elephant

Snowman

Snowflake

Dog

Owl

Sun

SleepingBag

Asparagus

Match

1

3

1

2

1

2

1

3

1

2

3

1

3

1

2

5

2

1

2

4

1

2

Matches

Quiltings

Stitch

Steak

Stereo

2

1

2

1

2

6

5

7

3

4

5

6

1

Stethoscope

WashingMachine

Table

2

1

2

1

Chair

Foot

Ties

1

11

1

Sandwich

Table

Tv

Phone

TennisRacket

Tigris

Axe

Tornado

1

7

4

8

3

7

6

5

2

7

10

4

8

3

4

3

4

5

2

5

9

3

5

2

1

2

1

2

Cake

BirthdayCake

Torchere

Toaster

1

Grass

Tractor

1

2

1

3

1

8

Trombone

Springboard for jumping the water

Triangle

1

5

10

4

2

1

5

6

10

6

9

5

9

1

Trombone

Pipe

1

Trumpet (MusicalInstrument)

Pipe(Muses. Tool)

Trumpet(MusicalInstrument)

TubeAndMask

TubeWithMask

Toilet

1

2

8

6

7

1

5

1

2

Snail

1

StreetLamp

SmilingFace

Smile

1

10

Hurricane

2

1

059

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

Moustache

Duck

2

5

2

1

5

9

2

Ear

Flag

7

2

5

6

3

4

6

1

9

2

5

7

4

10

7

10

7

Flamingo

FeltPen

Lantern

Piano

FruitIce

Van

1

2

1

2

1

Jersey

6

1

Football

Bread

HockeyStick

HockeyPuck

Hot Dog

Flower

4

8

6

5

4

5

7

1

2

1

6

10

9

7

6

8

1

2

1

2

1

Church

Kettle

Clock

Cup

Suitcase

Skull

Turtle

Blueberry

Hexagon

SchoolBus

Helmet

FlipFlops

Hat

Shorts

Pants

Brush

EiffelTower

Goatee

3

2

7

5

2

4

3

1

4

1

5

8

7

6

3

7

6

8

6

3

6

3

6

1

7

4

1

2

1

2

1

2

1

2

1

2

Apple

4

5

The overall result

1491

220

4

225 305

4

133 41

3

9

76

1

2512

As a result of data processing the following dependencies

are obtained. Figure 2. - The ratio of unrecognized neural

network drawings and recognized objects (pictures). From

figure 2 it can be seen that the recognized neural network

drawings - 59%, which is 18% more than unrecognized.

From figure 3 it is seen that is not recognized iteration -27%

recognized iteration -73%. Figure 4 shows that the human

factor is the cause of errors in pattern recognition by the

neural network, which is explained by the following reasons:



not every person (subject) is an artist, so most

people do not have the ability to draw, do not have the skills

to image objects and objects;



not enough time to complete the picture, the more

that are in need of training the neural network includes

training complex pictures such as: Green beans, Mona Lisa,

migration, anvil, nose, jail, detention facility, stress time,

trombone (musical instrument) and clarinet (musical

instrument), a goatee.

Figure 2: Not recognized and recognized objects (in %).

1

060

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

4

Discussion

October 11, 2019. Russian President Vladimir Putin

approved a national strategy for the development of artificial

intelligence until 2030. “In order to accelerate the

development of artificial intelligence in the Russian

Federation, conduct scientific research in the field of artificial

intelligence, increase the availability of information and

computing resources for users, improve the training system in

this area, it is necessary to apply a national strategy for the

development of artificial intelligence until 2030,” - the

document says (17).

«Artificial intelligence in

a

strategy refers to

technological solutions that allow you to simulate the

cognitive functions of a person and get results when

performing certain tasks that are comparable, at least, with

the results of intellectual human activity. Artificial

intelligence includes IT infrastructure, software (including

that using machine learning methods), as well as processes

and services for processing data and finding solutions, the

strategy says. Countries with the development of artificial

intelligence will receive advantages not comparable to

nuclear weapons, Russia has every chance of succeeding in

this, said Russian President Vladimir Putin earlier» Putin

V.V. approved the development strategy of artificial

intelligence until 2030 (17).

Figure 3: Not recognized and recognized iterations (in %).

5

Conclusion

The article presents the most modern and actual heuristic

and multi-heuristic algorithms that can give solutions as close

as possible to the correct ones by the satisfying time (solution

of NP-difficult problems). It also touches upon the issue of

assessing the quality of neural networks both on standard

templates and for the most difficult situations related to the

creative process. Within the framework of the work, twenty-

five experiments were conducted and data on more than a

hundred proposed images were collected and neural network

reaction was noted to each of the described object. The

obtained results are carefully sorted and presented in the form

of tables intended for further analyze in order to develop

mathematical models and algorithms for assessing the quality

of the learning process of modern neural networks. The

authors put forward the thesis about the need to develop

mathematical models to assess the quality of machine

learning algorithms. The authors classify the results obtained

on the different stages of the machine learning. In the article

discusses both typical examples of the modern neural

network and the most complex ones related to the creative

Figure 4: Cause of error (in iterations in %).

3

Results

It should be noted that students in the course of the

experiment noted do not know the true meaning of some

words. It is determined that the neural network learning

algorithm requests an image of similar words, for example:

1

2

3

4

5

.

Teddy Bear and Teddy bear.

Cruise ship and Cruise liner

Potatoes and French fries

French fries and Chips

Television

Training of the neural network was carried out

simultaneously by 45 students, performing many iterations.

The first experiments on visual observation showed that

neural network training corresponds to the thinking of a child

aged 2 to 4 years. As the neural network was being trained,

the neural network learning ability assessment came to the

age of 7-8 years, then 12-14 years, sometimes up to 18 years

and fell again to the development of a 5-year-old child. This

suggests that the neural network is trainable. Participants of

the experiment in 79% of cases considered guilty of a bad

result of the Information system human factor – the inability

of a person to draw an object correctly. Indeed, not everyone

has artistic abilities.

process. The authors conducted

a huge number of

experiments on different objects recognition and in general at

different stages of machine learning. The authors also tried to

assess the possibility of simulating the creative process

through the deep machine learning. The main purpose of the

publication is the publishing the results for the aim of the

further analyze and developing methods and algorithms for

assessing the quality of machine learning.

References

1

.

Adamar Zh. Psychology process study in the field of mathematic

science. Moscow: Annex III. 1970.

1

061

Journal of Environmental Treatment Techniques

2019, Special Issue on Environment, Management and Economy, Pages: 1050-1062

2

.

Averkin AN, Gaaze-Rapoport MG, Pospelov DA. Artificial

intelligence dictionary. Moscow: Radio and

communication.1992.

the national technology initiative of Russian Federation. Modern

journal of language teaching methods.2018;8(6):84-100.

22. Wallas G. The Artof Thought. New York. 1926.

3

4

.

Basu M. Economic environmental dispatch of hydrothermal

power system. Int J Elec Power.2010;32:711-720.

Berdyaev NA. Experience urges Catholics in logical

metaphysics. Creativity and objectification. Moscow: AST.

23. Xiang L, Jianzhong Z, Shuo O, Rui Z, Yongchuan Z. An

adaptive chaotic artiﬁcial bee colony algorithm for short-term

hydrothermal

generation

scheduling.

Int

J

Elec

Power.2013;53:34-42.

2

003.

24. Ying W, Jianzhong Z, Chao Z, Yongqiang W, Hui Q. An

improved self-adaptive PSO technique for short-term

hydrothermal scheduling. Expert Syst Appl.2012;39:2288-2295.

25. Youlin L, Jianzhong Z, Hui Q, Ying W, Yongchuan Z. An

adaptive chaotic differential evolution for the short-term

hydrothermal generation scheduling problem. Energ Convers

Manage.2010;51:1481-1490.

5

6

7

8

.

Binghui Y, Xiaohui Y, Jinwen W. Short-term hydro-thermal

scheduling using particle swarm optimization method. Energ

Convers Manage.2007;48:1902-1908.

Galushkin AA. Organizations of the sphere of education and

science methods of competitiveness improvement. European

Research Studies Journal.2017;20(4B):421-431

Hota PK, Barisal AK, Chakrabarti R. An improved PSO

technique for short-term optimal hydrothermal scheduling. Electr

Pow Syst Res.2009;79:104-117.

Khitskov EA, Veretekhina SV, Medvedeva AV, Mnatsakanyan

OL, Shmakova EG, Kotenev A. Digital transformation of

society: Problems entering in the digital economy. Eurasian

Journal of Analytical Chemistry.2017;7(5):855-873.

Kumar S, Naresh R. Efficient real coded genetic algorithm to

solve the non-convex hydrothermal scheduling problem. Int J

Elec Power.2007;29:738-747.

26. Cherniavskih, V.I., Sidelnikov, N.I., Dumacheva, E.V.,

Borodaeva, Z.A., Glubsheva, T.N., Gorbacheva, A.A.,

Vorobyova, O.V. and Korolkova, S., 2019. Biological resources

of natural forage grassland of the cretaceous south of the

European Russia. EurAsian Journal of BioSciences, 13(2),

pp.845-849.

9

1

.

0. Lakshminarasimman L, Subramanian S. Short-term scheduling

of hydrothermal power system with cascaded reservoirs by using

modiﬁed differential evolution. IEE Proc - General Transmission

Distribution.2006;153:693-700.

1

1. Lakshminarasimman L, Subramanian S. A modiﬁed hybrid

differential evolution for short-term scheduling of hydrothermal

power systems with cascaded reservoirs. Energ Convers

Manage.2008;49:2513-2521.

1

2. Mandal KK, Basu M, Chakraborty N. Particle swarm

optimization technique based short-term hydrothermal

scheduling. Appl Soft Comput.2008;8:1392-1399.

3. Naresh R, Sharma J. Two-phase neural network based solution

technique for short term hydrothermal scheduling. IEE Proc-

Gener Transm Distrib.1999;146:657-663.

4. Nima A, Hassan RS. Daily hydrothermal generation scheduling

by

a new modiﬁed adaptive particle swarm optimization

technique. Electr Pow Syst Res.2010;80:723-732.

5. Orero, S.O., Irving, M.R. (1998). A genetic algorithm modeling

framework and solution technique for short termoptimal

hydrothermal scheduling. IEEE T Power Syst, 13, 501-518.

6. Ravindhranathreddy B, Suryakalavathi M. Coua Bird-Inspired

Algorithm for Solving Power System Problem. International

Journal of Swarm Intelligence and Evolutionary Computation

Department of Electrical Engineering, Jawaharlal Nehru

Technological University.2016;DOI: 10.4172 2090-

908.1000141.

7. RIF News. Putin V.V. approved the development strategy of

artificial intelligence until

030.2019;https://ria.ru/20191011/1559659017.html

8. Sivasubramani S, Swarup KS. Hybrid DE–SQP algorithm for

non-convex short term hydrothermal scheduling problem. Energ

Convers Manage.2011;52:757-761.

9. Swain RK, Barisal AK, Hota PK, Chakrabarti R. Short-term

hydrothermal scheduling using clonal selection algorithm. Int J

Elec Power.2011;33:647-656.

0. Veretekhina SV, Karyagina TV, Kornyushko VF, Burlyaeva EV,

Kolybanov KY, Potekhina EV, Shmakova EG. Information

system for monitoring the state of the natural environment

according to the Russian Satellite. Ekoloji.2018;27(106):461-

/

4

1

2

4

69.

2

1. Veretekhina SV, Karyagina TV, Potekhina EV, Nakhratova EE,

Pronkina TV, Makushkin SA. Mathematical methods of an

estimation of economic efficiency of investments and the

sequence of execution of starts of investment on the example of

1

062