Disordered Brain Modeling Using Artificial Network SOFM

International Journal of Artificial Intelligence and Applications for Smart Devices

Vol.3, No.1 (2015), pp.1-14

http://dx.doi.org/10.14257/ijaiasd.2015.3.1.01

ISSN: 2288-6710 IJAIASD

Copyright ⓒ 2015 SERSC

Disordered Brain Modeling Using Artificial Network SOFM

Md. Syeful Islam1, Ruhul Abedin

2 and Fakrul Hasan

3

1Samsung R&D Institute Bangladesh Ltd.

2BJIT Ltd.

3Dynamic Solution Innovators Ltd.

[email protected], [email protected], [email protected]

Abstract

Autism is known as a neurobiological developmental disorder which affects language,

communication, and cognitive skill. In the case of autism attention shift impairment and

strong familiarity preference are considered to be prime deficiencies. Attention shift

impairment is one of the most seen behavioral disorders found in autistic patients. We

have model this behavior by employing self-organizing feature map (SOFM).

Keywords: Autism, Self Organizing Feature Map, Autism Spectrum Disorder, Feature

Map, Neural Network

1. Introduction

Autism is a disorder of neural development characterized by impaired social interaction

and communication by restricted and repetitive behavior. These signs all begin before a

child is three years old [1]. Autism affects information processing in the brain by altering

how nerve cells and their synapses connect and organize; how this occurs is not well

understood [2].

Autism has a strong genetic basis, although the genetics of autism are complex and it is

unclear whether ASD is explained more by rare mutations, or by rare combinations of

common genetic variants [3]. In rare cases, autism is strongly associated with agents that

cause birth defects. Controversies surround other proposed environmental causes, such as

heavy metals, pesticides or childhood vaccines; the vaccine hypotheses are biologically

implausible and lack convincing scientific evidence. The prevalence of autism is about 1–

2 per 1,000 people worldwide; however, the Centers for Disease Control and Prevention

(CDC) reports approximately 9 per 1,000 children in the United States are diagnosed with

ASD. The number of people diagnosed with autism has increased dramatically since the

1980s, partly due to changes in diagnostic practice; the question of whether actual

prevalence has increased is unresolved.

Currently, there is no reliable evidence as to exactly what are the neural bases for

autism. It is clear that autism is a heterogeneous disorder, even it is genetic. There is no

accepted animal model condition, although infant monkeys with selective brain lesions

show behavioral, features suggestive of autism. Our research goal is to model autism in

such way that is helpful to improve the learning and training process of autistic people

.We used unsupervised learning algorithm that can train autistic people to recognize and

learn object. This model has been implemented using Self Organizing Feature Map

(SOFM) [4].

2. Understanding Autism

Autism is a term used for a number of developmental disabilities called Autism

Spectrum Disorder (ASD). ASD is a life-long neurobiological disorder that affects


Vol. 3, No. 1 (2015)

2 Copyright ⓒ 2015 SERSC

how the person perceives and interprets their world, particularly the social

environment. [5]

The symptoms of ASD vary and can range from mild to severe, but most children

on the spectrum show difficulties with:

1) Social Interaction

2) Verbal and non-verbal communication

3) Repetitive behaviors or limited interests

2.1. Possible Reason of Autism

It has long been presumed that there is a common cause at the genetic, cogniti ve,

and neural levels for autism's characteristic triad of symptoms [6]. However, there is

increasing suspicion that autism is instead a complex disorder whose core aspects

have distinct causes that often co-occur.

Autism has a strong genetic basis, although the genetics of autism are complex

and it is unclear whether ASD is explained more by rare mutations with major

effects, or by rare multi-gene interactions of common genetic variants. Complexity

arises due to interactions among multiple genes, the environment, and epigenetic

factors which do not change DNA but are heritable and influence gene expression.

Studies of twins suggest that heritability is 0.7 for autism and as high as 0.9 for

ASD, and siblings of those with autism are about 25 times more likely to be autistic

than the general population. However, most of the mutations that increase autism

risk have not been identified. Typically, autism cannot be traced to a Mendelian

(single-gene) mutation or to a single chromosome abnormality like fragile X

syndrome, and none of the genetic syndromes associated with ASDs have been

shown to selectively cause ASD. Numerous candidate genes have been located, with

only small effects attributable to any particular gene. The large number of autistic

individuals with unaffected family members may result from copy number

variations—spontaneous deletions or duplications in genetic material during

meiosis. Hence, a substantial fraction of autism cases may be traceable to genetic

causes that are highly heritable but not inherited: that is, the mutation that causes

the autism is not present in the parental genome.

2.2. Over view of Human Brain

The human brain is the center of the human nervous system [8]. Enclosed in the

cranium, the human brain has the same general structure as that of other mammals,

but is over three times larger than the brain of a typical mammal with an equivalent

body size. Most of the spatial expansion comes from the cerebral cortex, a

convoluted layer of neural tissue which covers the surface of the forebrain.

Especially expanded are the frontal lobes, which are associated with executive

functions such as self-control, planning, reasoning, and abstract thought. The

portion of the brain devoted to vision, the occipital lobe, is also greatly enlarged in

human beings.


Vol. 3, No. 1 (2015)

Copyright ⓒ 2015 SERSC 3

Brain evolution, from the earliest shrew-like mammals through primates to

hominids, is marked by a steady increase in en-cephalization, or the ratio of brain to

body size. Estimates vary for the number of neuronal and non-neuronal cells

contained in the brain, ranging from 80 or 90 billion (~85 109) non-neuronal cells

(glial cells) and an approximately equal number of (~86 109) neurons, of which

about 10 billion (1010) are cortical pyramidal cells, to over 120 billion neuronal

cells, with an approximately equal number of non-neuronal cells. These cells pass

signals to each other via as many as 1000 trillion (1015, 1 quadrillion) synaptic

connections. Due to evolution and synaptic pruning, however, the modern human

brain has been shrinking over the past 28,000 years.

Figure 1. Overview of Human Brain

The brain monitors and regulates the body's actions and reactions. It continuously

receives sensory information, and rapidly analyzes this data and then responds

accordingly by controlling bodily actions and functions. The brainstem controls

breathing, heart rate, and other autonomic processes that are independent of

conscious brain functions. The neo-cortex is the center of higher-order thinking,

learning, and memory. The cerebellum is responsible for the body's balance,

posture, and the coordination of movement.

Despite being protected by the thick bones of the skull, suspended in

cerebrospinal fluid, and isolated from the bloodstream by the blood-brain barrier,

the human brain is susceptible to many types of damage and disease. The most

common forms of physical damage are closed head injuries such as a blow to the

head, a stroke, or poisoning by a wide variety of chemicals that can act as

neurotoxins. Infection of the brain, though serious, is rare due to the biological

barriers which protect it. The human brain is also susceptible to degenerative

disorders, such as Parkinson's disease, multiple sclerosis, and Alzheimer's disease.

A number of psychiatric conditions, such as schizophrenia and depression, are

widely thought to be associated with brain dysfunctions, although the nature of such

brain anomalies is not well understood.


Vol. 3, No. 1 (2015)


2.3. Overview of Autistic Brain

Autism is a disorder of neural development characterized by impaired socia l

interaction and communication, and by restricted and repetitive behavior. These

signs all begin before a child is three years old. Autism affects information

processing in the brain by altering how nerve cells and their synapses connect and

organize; how this occurs is not well understood. It is one of three recognized

disorders in the autism spectrum (ASDs), the other two being Asperger syndrome,

which lacks delays in cognitive development and language, and Pervasive

Developmental Disorder-Not Otherwise Specified (commonly abbreviated as PDD-

NOS), which is diagnosed when the full set of criteria for autism or Asperger

syndrome are not met.

Figure 2. Parts of the Brain Affected by Autism

Over the past few years, a number of studies have been published linking

differences in brain structure and function to autism spectrum disorders. For

example, scientists have noted that:

1) At a certain point in post-natal development, autistic brains are larger.

2) Testosterone may be linked to autism.

3) Certain portions of the brain, such as the amygdala, may be enlarged in

autistic brains.

4) Certain parts of the brain may function differently in autistic people.

5) "Minicolumns" in the brain may be formed differently and be more

numerous in autistic brains.

6) The entire brain may function differently in autistic people.

To better understand which of these findings is legitimate and which is most

significant, I interviewed Dr. Nancy Minshew of the University of Pittsburgh.

Minshew is one of the most prolific and best-known researchers in the field of

autism and the brain. According to Dr. Minshew, "These different theories are not

all so different."


Vol. 3, No. 1 (2015)


2.4. The Autistic Brain Is Differently Wired

What all of these brain findings have in common, Dr. Minshew explains, is that

they point to autism as a disorder of the cortex. The cortex is the proverbial "gray

matter": the part of the brain which is largely responsible for higher brain functions,

including sensation, voluntary muscle movement, thought, reasoning, and memory.

Figure 3. Children’s Brain (Autistic Brain vs. Normal Brain)

Figure 4. Major Brain Structure Implicated in Autism

In many autistic people, the brain develops too quickly beginning at about 12

months. By age ten, their brains are at a normal size, but "wired" atypically. "The

brain is most complex thing on the planet," says Dr. Minshew. "So its wiring has to

be very complex and intricate. With autism there's accelerated growth at the wrong

time, and that creates havoc. The consequences, in terms of disturbing early

development, include problems within the cortex and from the cortex to other

regions of the cortex in ways that compromise language and reasoning abilities."

Minicolumns, which are small structures within the cortex, are also different

among autistic people. Dr. Manuel Casanova, a researcher at the University of


Vol. 3, No. 1 (2015)


Kentucky, has found that autistic people have more minicolumns which include a

greater number of smaller brain cells. In addition, the "insulation" between these

minicolumns is not as effective as it is among typically developing people. The

result may be that autistic people think and perceive differently and have less of an

ability to block sensory input.

2.5. Treatment of Autism

No cure is known. Children recover occasionally, so that they lose their diagnosis

of ASD; this occurs sometimes after intensive treatment and sometimes not. It is not

known how often recovery happens; reported rates in unselected samples of children

with ASD have ranged from 3% to 25%. Most autistic children can acquire language

by age 5 or younger, though a few have developed communication skills in later

years. Most children with autism lack social support, meaningful relationships,

future employment opportunities or self-determination. Although core difficulties

tend to persist, symptoms often become less severe with age. Few high-quality

studies address long-term prognosis. Some adults show modest improvement in

communication skills, but a few decline; no study has focused on autism after

midlife. Acquiring language before age six, having an IQ above 50, and having a

marketable skill all predict better outcomes; independent living is unlikely with

severe autism. A 2004 British study of 68 adults who were diagnosed before 1980 as

autistic children with IQ above 50 found that 12% achieved a high level of

independence as adults, 10% had some friends and were generally in work but

required some support, 19% had some independence but were generally living at

home and needed considerable support and supervision in daily living, 46% needed

specialist residential provision from facilities specializing in ASD with a high level

of support and very limited autonomy, and 12% needed high-level hospital care. A

2005 Swedish study of 78 adults that did not exclude low IQ found worse prognosis;

for example, only 4% achieved independence. A 2008 Canadian study of 48 young

adults diagnosed with ASD as preschoolers found outcomes ranging through poor

(46%), fair (32%), good (17%), and very good (4%); 56% of these young adults had

been employed at some point during their lives, mostly in volunteer, sheltered or

part-time work. Changes in diagnostic practice and increased availability of

effective early intervention make it unclear whether these findings can be

generalized to recently diagnosed children [9].

3. Modeling of Authentic Learning

A self-organizing map (SOM) or self-organizing feature map (SOFM) is a type of

artificial neural network that is trained using unsupervised learning to produce a low-

dimensional (typically two-dimensional), discredited representation of the input space of

the training samples, called a map. Self-organizing maps are different from other artificial

neural networks in the sense that they use a neighborhood function to preserve the

topological properties of the input space. This makes SOMs useful for visualizing low-

dimensional views of high-dimensional data, akin to multidimensional scaling. The model

was first described as an artificial neural network by the Finnish professor Teuvo

Kohonen, and is sometimes called a Kohonen map [11, 12].

For our working purpose the self-organizing map is described by dividing it into 3

parts:

1) p-dimensional input space

2) l-dimensional feature space and

3) Competitive net for measuring minimum distance.


Vol. 3, No. 1 (2015)


There a SOM is discussed about where m neurons, each with p synapses are organized

in an l-dimensional lattice (grid) representing the feature space. Such a neural network

performs mapping of a p-dimensional input space into the l-dimensional feature space. In

Figure 5 we present an example of a self-organizing map consisting of m =12 neurons in

which the input space is 3-dimensional (p = 3) and the feature space is 2-dimensional (l =

2). The first section of the network is a distance-measure layer consisting of m = 12

dendrites each containing p = 3 synapses ex-cited by p–dimensional stimuli x and

characterized by the p–dimensional weight vector wi, i =1,...,m. The distance-measure

layer calculates the distances d¬i between each input vector x and every weight vector wi.

This distance vector di = [d1,...,dm)] is passed to the competition layer, the Min

Net,which calculates the minimal distance di = min di in order to establish the position of

the winning neuron k.

Figure 5. A 2-D SOM with p=3; m=[3 4]; l=2

3.1. Detail of the SOM Learning Algorithm

The complete algorithm can be described as consisting of the following steps-

Initialize:

(a) The weight matrix W with a random sample of m input vectors.

(b) The learning gain and the spread of the neighborhood function.

For every input vector, x(n), n = 1, . . . ,N:

(a) Determine the winning neuron, k(n), and its position V (k, :) as

k (n) = argmin = |xT(n) −W(j, :)|

(b) Calculate the neighborhood functions -

Λ(n, j) = exp(−ρ2(j)/2σ2)

; Where ρ (j) = |V (j, :) − V (k(n), :)| for j = 1, . . . ,m.

(c) Update the weight matrix as -

∆W = η (n) • Λ(n) • (xT(n) −W(j, :))


Vol. 3, No. 1 (2015)


0 0.5 1 1.5 2 2.5 3 3.5 40.5

1

1.5

2

2.5

3

3.5

4

w11

w21

w31

w12

w22

w32

w13

w23

w33

w14

w24

w34

x1

x2

All neurons (unlike in the simple competitive learning) have their weights modified

with a strength proportional to the neighborhood function and to the distance of their

weight vector from the current input vector (as in competitive learning).The step (2) is

repeated E times, where E is the number of epochs.

3.2. Feature Map

Self-Organizing Feature Maps (SOFM or SOM) also known as Kohonen maps or

topographic maps were first introduced by von der Malsburg (1973) and in its present

form by Kohonen (1982) [7]. A typical Feature Map is a plot of synaptic weights in the

input space in which weights of the neighboring neurons are joined by lines and illustrates

the mapping from the input space to the feature spaces. For simplicity, we restrict our

attention here to two-dimensional input and feature spaces (p, l = 2).As an illustrative

example let us consider a SOM with p = 2 inputs and m = 12 neurons organized on a 3 ×

4 lattice. An example of the weight W and position V matrices and the resulting feature

map is given in Figure 6.

Figure 6. Example of Weight and Position Matrix and their Feature Map for p, l =2

In our modeling we will be using similar SOMs with p, l = 2 and neurons organized in

either a 2 × 2, or 3 × 3 mesh.

3.3. The Autistic Learning Model

Based on the SOM neural network a model of a learning autistic model can be build.

The block diagram of the learning model is described below:

Figure 7. A Block Diagram of the Model of Autistic Learning

K W V 1

2

3

4

5

6

7

8

9

10

11

12

0.73 0.87

0 .62 1.01

0.18 2.93

3.07 2.06

1.83 2.81

1.47 2.28

3.38 1.27

0.6 2.27

3.51 0.61

3.26 1.85

2.92 3.05

3.16 3.90

1 1

2 1

3 1

1 2

2 2

3 2

1 3

2 3

3 3

1 4

2 4

3 4


Vol. 3, No. 1 (2015)


The block-diagram of the model of autistic learning which includes source familiarity

filter and attention shift mechanism is presented in Figure 7. The central part is the SOM

neural network as presented in Figure 5, together with the learning section implementing

the learning law, or map formation algorithm. At each learning step a stimulus is

randomly generated from one of the sources, S1...Sc. The attention shifting mechanism

determines if that stimulus is presented to the map for learning. For modeling autistic

learning we have used two learning mode:

In the normal, or novelty seeking learning mode, attention is shifted to another source

if the new stimulus originates from that source.

(a) (b)

Figure 8. A 4×4 Feature Map in the 2-D Input Space Developed in the: (a) Novelty Seeking, and (b) Attention Shifting Restricted by Familiarity

Preference Learning Modes

In the attention shifting restricted by familiarity preference learning mode attention is

shifted to another source if that source presents the next new stimulus, but conditionally,

depending on the map’s familiarity with that source. The map familiarity to a particular

source is measured by the time averaged value of the distance between map nodes and the

stimuli. When both sources are unfamiliar to the map, i.e., in the beginning stage of self-

organization, attention is shifted to an alternate source if that source presented the next

stimulus as in the novelty seeking mode. As the map develops some familiarity with the

sources, i.e., the node weights begin to resemble the data; attention is shifted with a higher

probability to the source which is most familiar to the map. If the map becomes familiar

to two or more sources (the average deference between node weights and the data from

the sources becomes smaller than a predetermined small value) then attention is

unconditionally shifted.

The feature map presented in Figure 8 (b) is the result of learning when the attention

shifting is restricted by familiarity preference.

4. Result Analysis

In order to model autistic behavior we arrange the two-dimensional training data

and plot them into feature map using SOFM. In this section I have taken training

data for random behavioral neuron and we presented some of experimental output of

training data plot into feature map.

Here Training data (synaptic weights in the input space in which weights of the

neighboring neurons are joined by lines) is plot into map which illustrates the

mapping from p-dimensional input space to l-dimensional feature space. In training

data K denotes the serial no. of neuron and V is the position and W represent the


Vol. 3, No. 1 (2015)


1 1.5 2 2.5 3 3.5 4 4.51.5

2

2.5

3

3.5

4

4.5

w11

w21

w31

w12 w

22

w32

w13

w23

w33

x1

1 1.5 2 2.5 3 3.5 41.5

2

2.5

3

3.5

4

4.5

5

w11

w21

w31

w12

w22

w32

w13

w23

w33

w14

w24

w34

x1

weight of corresponding neuron. Each neuron, yv in the feature map is characterized

by its position in the lattice specified by a 2-D vector v = [v1 v2], and by a 3-D

weight vector wv = [w1v w2v w3v]. In the feature map the point representing the

weight vector, wk, is joined by line segments with points representing weights wk −

1 and wk +1 an so no because neurons wk − 1, k, and wk + 1 are located in the

adjacent positions of the 1-D neuronal lattice.

4.1. Second-order Headings

(a) (b)

Figure 9. (a) Table Containing Weight and Position Vector of a 3x3 Matrices, (b) Feature Map of a 3x3 Matrices

In Figure 9 (a) denote the no. of nine neurons and the weight W and position V of

corresponding neurons. After that we plot them into 2-D input space into a 2-D

neuronal space. Here 9 neurons are organized into 3×3 grid and it should be plot

into 3×3 output lattice. Consider a neuron 5 which located at the central vertex of

the 3×3 neuronal output lattice. The neuron has four neighbors: 4, 6, 2 and 8.

Therefore, in the feature maps the nodes w12, w32, w21 and w23 will all be joint

with a line to the node w22.

Similarly we have plot several neurons and there behavioral training data those

are organized at 4×4, 4×5 and 5×5 grid and map into corresponding feature map

[Figure 10-13].

4.2. Computing Result of 3x4 Matrices

(a) (b)


k W V

1

2

3

4

5

6

7

8

9

1.2112 1.9029

2.9771 2.1452

3.5297 1.9243

2.2040 2.4788

3.1951 2.4056

3.8310 2.4777

1.6952 3.7477

3.2597 4.0180

4.1503 3.4330

1.0000 1.0000

2.0000 1.0000

3.0000 1.0000

1.0000 2.0000

2.0000 2.0000

3.0000 2.0000

1.0000 3.0000

2.0000 3.0000

3.0000 3.0000

1

2

3

4

5

6

7

8

9

10

11

12

1.0214 1.9538

3.0455 1.5313

3.6231 2.1645

2.3045 2.7039

2.6524 2.9933

3.5861 2.6004

2.1847 3.4265

2.7352 3.2655

3.2837 3.2708

1.9410 4.9551

3.1734 4.4239

3.0275 4.7583

1.0000 1.0000

2.0000 1.0000

3.0000 1.0000

1.0000 2.0000

2.0000 2.0000

3.0000 2.0000

1.0000 3.0000

2.0000 3.0000

3.0000 3.0000

1.0000 4.0000

2.0000 4.0000

3.0000 4.0000


Vol. 3, No. 1 (2015)


1 1.5 2 2.5 3 3.5 4 4.5 5 5.51

1.5

2

2.5

3

3.5

4

4.5

5

w11

w21

w31

w41

w12

w22

w32

w42

w13

w23

w33

w43

w14

w24

w34

w44

x1

x2

1 1.5 2 2.5 3 3.5 4 4.5 5 5.51

2

3

4

5

6

7

w11

w21

w31

w41

w12

w22

w32

w42

w13

w23 w

33 w43

w14

w24

w34

w44

w15

w25

w35

w45

x1

x2

4.3. Computing Result of 4x4 Matrices

(a) (b)


4.4. Computing Results of 4x5 Matrices

(a) (b)

Figure 12. (a) Weight and Position Vector of a 4x5 Matrices, (b) Feature Map of a 4x5 Matrices

K W V

1 2

3

4 5

6 7

8

9 10

11 12

13

14 15

16

2.1739 2.2260 2.7953 2.0322

3.5186 1.1911

4.9838 1.0165 1.7652 3.2515

2.6228 2.2788 3.9724 2.4182

4.8698 2.9260

2.1127 3.3982 3.3396 3.6569

3.7316 3.0907 5.2322 4.3837

1.2421 4.8159

3.3716 4.5929 3.3800 4.7217

4.3533 4.4675

1.0000 1.0000 2.0000 1.0000

3.0000 1.0000

4.0000 1.0000 1.0000 2.0000

2.0000 2.0000 3.0000 2.0000

4.0000 2.0000

1.0000 3.0000 2.0000 3.0000

3.0000 3.0000 4.0000 3.0000

1.0000 4.0000

2.0000 4.0000 3.0000 4.0000

4.0000 4.0000

K W V

1

2

3

4

5 6

7

8 9

10 11

12

13 14

15 16

17

18 19

20

1.5744 1.2837

3.2511 1.9410

3.0810 2.1734

4.4940 1.0275

2.1384 2.9538 2.0138 2.5313

3.1944 3.1645

4.2839 2.7039 1.2782 3.9933

2.8453 3.6004 3.3811 3.4265

4.2783 3.2655

1.0214 4.2708 3.0455 4.9551

3.6231 4.4239 5.3045 4.7583

1.6524 5.2112

2.5861 5.9771 4.1847 5.5297

4.7352 6.2040

1.0000 1.0000

2.0000 1.0000

3.0000 1.0000

4.0000 1.0000

1.0000 2.0000 2.0000 2.0000

3.0000 2.0000

4.0000 2.0000 1.0000 3.0000

2.0000 3.0000 3.0000 3.0000

4.0000 3.0000

1.0000 4.0000 2.0000 4.0000

3.0000 4.0000 4.0000 4.0000

1.0000 5.0000

2.0000 5.0000 3.0000 5.0000

4.0000 5.0000


Vol. 3, No. 1 (2015)


1 2 3 4 5 6 71

2

3

4

5

6

7

w11

w21

w31

w41 w

51

w12 w

22

w32 w

42

w52

w13

w23

w33

w43

w53

w14

w24

w34

w44

w54

w15

w25

w35

w45

w55

x1

x2

4.5. Computing Results of 5x5 Matrices

(a) (b)

Figure 13. (a) Weight and Position Vector of a 5x5 Matrices, (b) Feature Map of a 5x5 Matrices

5. Conclusion

Our works do not help autistic children to learn but it only focuses on the

modeling of one of the autistic behavioral problem, attention shift impairment, with

the conjunction of familiarity preference. In our resulting feature maps for the

attention shift impairment restricted by familiarity preference, the map shrinks by it

area because of the reduced learning capability. This shrinkage of the map

represents the reduced learning capability due to attention shift impairment.

Self-organization feature map shows if attention shift is very low (for the case of

autism) and the learning rate is very low too. And hence the autistic patient whose

attention shift is restricted by familiarity preference shows the behavior of doing

something repetitively.

References

[1] S. M. Myers and C. P. Johnson, “Management of children with autism spectrum disorders”, Pediatrics,

vol. 120, no. 5, pp. 1162-82.

[2] G. A. Stefanatos, “Regression in autistic spectrum disorders”. Neuropsychol Rev., vol. 18, no. 4, (2008),

pp. 305-19.

[3] E. B. Caronna, J. M. Milunsky and H. Tager-Flusberg, “Autism spectrum disorders: clinical and

research frontiers”, Arch Dis Child, vol. 93, no. 6, pp. 518-23.

[4] T. Kohonen, “Self-Organisation and Associative Memory”, Berlin: Springer-Verlag, 3rd ed., (2001).

[5] E. B. Caronna, J. M. Milunsky and H. Tager-Flusberg, “Autism spectrum disorders: clinical and

research frontiers”, Arch Dis Child, vol. 93, no. 6, (2008), pp. 518-23.

[6] Diagnostic and statistical manual of mental disorders. 4th ed. American Psychiatric Association, (1994).

[7] I. Cohen, “An artificial neural network analogue of learning in autism”, Biol. Psychiatry, no. 36, (1994),

pp. 5-20.

[8] D. G. Amaral, C. M. Schumann and C. W. Nordahl, “Neuroanatomy of autism”, Trends Neurosci., vol.

31, no. 3, pp. 137–45.

K W V 1

2 3

4

5 6

7 8

9

10 11

12 13

14

15 16

17 18

19

20 21

22 23

24

25

2.1951 2.2322

2.8310 1.2421 3.6952 2.3716

5.2597 1.3800

6.1503 1.3533 1.9029 3.2260

3.1452 3.0322 3.9243 2.1911

4.4788 2.0165

5.4056 3.2515 1.4777 3.2788

2.7477 3.4182 4.0180 3.9260

4.4330 3.3982

6.1739 3.6569 1.7953 4.0907

2.5186 5.3837 3.9838 4.8159

4.7652 4.5929

5.6228 4.7217 1.9724 5.4675

2.8698 5.6061 4.1127 5.3163

5.3396 5.8117

5.7316 6.0645

1.0000 1.0000

2.0000 1.0000 3.0000 1.0000

4.0000 1.0000

5.0000 1.0000 1.0000 2.0000

2.0000 2.0000 3.0000 2.0000

4.0000 2.0000

5.0000 2.0000 1.0000 3.0000

2.0000 3.0000 3.0000 3.0000

4.0000 3.0000

5.0000 3.0000 1.0000 4.0000

2.0000 4.0000 3.0000 4.0000

4.0000 4.0000

5.0000 4.0000 1.0000 5.0000

2.0000 5.0000 3.0000 5.0000

4.0000 5.0000

5.0000 5.0000


Vol. 3, No. 1 (2015)


[9] H. Cass, “Visual impairment and autism: current questions and future research”, Autism, vol. 2, no. 2,

(1998), pp. 117–38.

[10] S. Geman, E. Bienenstock and R. Doursat, “Neural networks and the bias/variance dilemma”, Neural

Computation, vol. 4, (1992), pp. 1-58.

[11] T. Kohonen and T. Honkela, “Kohonen Network”, (2007).

[12] M. Oja, S. Kaski and T. Kohonen, “Bibliography of SelfOrganizing Map (SOM) Papers: 1998-2001

Addendum”, Neural Computing Surveys, vol. 3, (2003), pp. 1-156.

Authors

Md. Syeful Islam, he obtained his B.Sc. and M.Sc. in Computer

Science and Engineering from Jahangirnagar University, Dhaka,

Bangladesh in 2010 and 2011 respectively. He is now working as a

Senior Software Engineer at Samsung R&D Institute Bangladesh.

Previously he worked as a software consultant in the Micro-Finance

solutions Department of Southtech Ltd. in Dhaka, Bangladesh. His

research interests are in Natural Language processing, AI,

embedded computer systems and sensor networks, distributed

Computing and big data analysis.

Ruhul Abedin, he obtained his B.Sc. in Computer Science and

Engineering from Jahangirnagar University, Dhaka, Bangladesh in

2010. He is now working as a Software Engineer at BJIT ltd.

previously he worked as Software Engineer Sonali Polaris Ltd. His

research interests are in Natural Language processing, AI and

Neural Network.

Fakrul Hasan, he obtained his B.Sc. in Computer Science and

Engineering from Jahangirnagar University, Dhaka, Bangladesh in

2010. He is now working as a Software Engineer at Dynamic

Solution Innovators Ltd. His research interests are in Natural

Language processing, AI and Neural Network.


Vol. 3, No. 1 (2015)


Documents

Disordered Brain Modeling Using Artificial Network SOFM