Assignment Choice #2 : Using Neural Networks to Classify People ( Please Explain

Question

Assignment Choice #2: Using Neural Networks to Classify People ( Please Explain in Detail Thank you)

Some have suggested that neural networks could be applied to people to indicate how likely they are to develop a disease or even become criminals. The idea is to design and “teach” a neural network-based DSS (Decision Support System) with a child’s personal characteristics, demographics, and genealogy, so that the neural network can recognize patterns in that data and suggest if that youngster is at risk for a specific disease or aberrant behavior. For this assignment, choose either susceptibility to disease or to criminal behavior. Explain thoroughly, in-depth what a neural network is and how it could work in the classification.

Include the following in your analysis:

What is a neural network?

How can we “teach” a neural network to recognize the specific patterns we’re interested in?

Select three personal characteristics of a child --demographic or hereditary-- that could influence in the future adult the chosen characteristic (disease or aberrant behavior.)

What DSS tool would you recommend/use to track those three characteristics you identified earlier?

Explanation / Answer

Explain thoroughly, in-depth what a neural network is and how it could work in the classification.

Include the following in your analysis:

What is a neural network?

How can we “teach” a neural network to recognize the specific patterns we’re interested in?

Select three personal characteristics of a child --demographic or hereditary-- that could influence in the future adult the chosen characteristic (disease or aberrant behavior.)

What DSS tool would you recommend/use to track those three characteristics you identified earlier?

Neural network-

1.1 What is a Neural Network?

An Artificial Neural Network (ANN) is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurones) working in unison to solve specific problems. ANNs, like people, learn by example. An ANN is configured for a specific application, such as pattern recognition or data classification, through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurones. This is true of ANNs as well.

How the Human Brain Learns?

Much is still unknown about how the brain trains itself to process information, so theories abound. In the human brain, a typical neuron collects signals from others through a host of fine structures called dendrites. The neuron sends out spikes of electrical activity through a long, thin stand known as an axon, which splits into thousands of branches. At the end of each branch, a structure called a synapseconverts the activity from the axon into electrical effects that inhibit or excite activity from the axon into electrical effects that inhibit or excite activity in the connected neurones. When a neuron receives excitatory input that is sufficiently large compared with its inhibitory input, it sends a spike of electrical activity down its axon. Learning occurs by changing the effectiveness of the synapses so that the influence of one neuron on another changes.

Pattern Recognition - an example

An important application of neural networks is pattern recognition. Pattern recognition can be implemented by using a feed-forward neural network that has been trained accordingly. During training, the network is trained to associate outputs with input patterns. When the network is used, it identifies the input pattern and tries to output the associated output pattern. The power of neural networks comes to life when a pattern that has no output associated with it, is given as an input. In this case, the network gives the output that corresponds to a taught input pattern that is least different from the given pattern.

How can we “teach” a neural network to recognize the specific patterns we’re interested in?

The human visual system is one of the wonders of the world. Consider the following sequence of handwritten digits:

Most people effortlessly recognize those digits as 504192. That ease is deceptive. In each hemisphere of our brain, humans have a primary visual cortex, also known as V1, containing 140 million neurons, with tens of billions of connections between them. And yet human vision involves not just V1, but an entire series of visual cortices - V2, V3, V4, and V5 - doing progressively more complex image processing. We carry in our heads a supercomputer, tuned by evolution over hundreds of millions of years, and superbly adapted to understand the visual world. Recognizing handwritten digits isn't easy. Rather, we humans are stupendously, astoundingly good at making sense of what our eyes show us. But nearly all that work is done unconsciously. And so we don't usually appreciate how tough a problem our visual systems solve.

The difficulty of visual pattern recognition becomes apparent if you attempt to write a computer program to recognize digits like those above. What seems easy when we do it ourselves suddenly becomes extremely difficult. Simple intuitions about how we recognize shapes - "a 9 has a loop at the top, and a vertical stroke in the bottom right" - turn out to be not so simple to express algorithmically. When you try to make such rules precise, you quickly get lost in a morass of exceptions and caveats and special cases. It seems hopeless.

Person Recognition

Person recognition is a challenging task in computer vision with applications such as surveillance, autonomous driving, robotics and virtual reality. The task of person recognition comprises classification, detection and segmentation problems and is a special case of the more general object recognition. The person class is specifically difficult to model, since pedestrians or faces exhibit high intraclass variability, because of variations in, for example, pose, clothing and background. The main difference between person detection and person classification is that in detection only the person class is looked for and people have to be differentiated from the rest of the world, whereas in classification the class is one of a select pool of fixed classes. Person detection has a long history; In 1985, for example, Tsukiyama et al. detect people in image sequences. In the following, an overview on types of pedestrian detection methods will be given, including the state of the art.

Pedestrian Detection

The methods of pedestrian detection can be split roughly into static and dynamic methods. Static methods are applied on still images or single frames, whereas dynamic methods require video sequences for motion information. Another type of method is the use of combinations of models and features to produce a more robust detector. Finally, there are part-based methods, which aim to handle partial occlusions. These method types are explained in the following. 3.1.1 Static Pedestrian Detection Static methods detect people on a single image by either searching the entire image with a model or by calculating features, which are then classified. Model-based methods, therefore, require a robust pedestrian model able to deal with problems.

Person Recognition variation and occlusion.

Examples of model-based methods include histogram of human appearance, where localization is formulated using the basin of attraction of a smoothed similarity function, template-trees, where a variety of shape exemplars is matched using trees that are clustered with stochastic optimization, and statistical field models, which capture local non-rigidity. Most pedestrian detection methods use feature extraction and classifiers on regions of interest. A detection window is selected from the entire image, often by the use of sliding window. Features are computed for this window, which are then fed into classifiers. The classifiers are trained on sample windows labelled as person and nonperson. An example of a classifier is support vector machines. Histogram of Oriented Gradients (HOG) is one such feature commonly used for feature-based detection. The detection window is tiled into overlapping blocks. In these blocks, computed gradients are aggregated into spatial and orientation cells and are additionally contrast-normalized. Another common feature is the Scale Invariant Feature Transform (SIFT), since it is robust and scale invariant. SIFT computes high-dimensional vectors representing image gradients similar to HOG. It can be combined with principle component analysis to reduce computational costs. Furthermore, other important features include Local Binary Patterns (LBP), where binary codes are computed by thresholding local information, and shape context, where points inside the shape and on its boundary are used to measure shape similarity.

Dynamic Pedestrian Detection

Assuming that the camera has a fixed position and orientation, i.e. there is no relative motion between the camera and the background, motion features are a strong indicator for pedestrian movement. Thus, multiple dynamic methods have been developed based on video data. For example, Histogram of Oriented Flow (HOF) combines motionbased descriptors with HOG appearance descriptors to detect moving and standing people.

Combined Methods The combination of complementary features can improve detection performance. Walk et al. combine HOG with a self-similarity feature and motion features derived from optical flow to construct a robust detector, showing that feature combinations handle detection challenges better than single features.

Part-Based Pedestrian Detection

A main challenge in pedestrian detection are occlusions, since pedestrians in, for example, street scenes occlude each other or are occluded by cars, street signs, animals, or other objects. The idea of part-based detection is to separate the object in question into parts, which are detected separately.

State of the Art known geometrical constraints, such as that one part has to be connected to another specific part in a certain angle. In the case of pedestrians, arms, legs, torsos and heads can be detected separately to robustly handle partial occlusions.

Basic principles have emerged that allow application to educational practice, especially in the early years from birth to five years that place great responsibility for brain development in the hands of parents and early childhood teachers. These principles include the following:

1. The human brain develops from conception to the early twenties from the bottom up with vital and autonomic functions and control coming first and cognitive-motor sensory and perceptual processes later and integration and decision making last.

2. The child's brain is influenced by the combined roles of genetics and experience.

3. The brain's capacity for change decreases with age.

4. Cognitive, emotional, and social capacities are inextricably intertwined throughout the life course.

5. Motor and cognitive functions interact with our brains, being the direct result of bipedalism.

6. Toxic stress damages developing brain architecture, which can lead to life-long problems in learning, behavior, and physical and mental health.

7. The child's environment directly affects synaptogenesis and allows for neurological optimization.

The Effect of Environmental Enrichment on the Child's Brain: Playing with the Genetics

Early life events can exert a powerful influence on both the pattern of brain architecture and behavioral development. Both early as well as later experiences contribute to the wiring diagram of the child's brain, but experiences during critical periods establish the basis for development beyond the early years. The role of the kindergarten and nursery teachers becomes critical in establishing the solid functional footing of the developing child and the neurological adult.

The foundations of brain architecture are established early in life through a continuous series of dynamic interactions between genetic influences, environmental conditions, and experiences. We have come to learn that the child's environment significantly impacts the timing and nature of gene expression directly affecting the child's brain architecture.

Because specific experiences potentiate or inhibit neural connectivity at key developmental stages, these time points are referred to as critical periods. Brain, cognitive, sensory, and perceptual development does not occur simultaneously but rather at different developmental stages. Each one of our perceptual, cognitive, and emotional capabilities is built upon the scaffolding provided by early life experiences. Examples can be found in both the visual and auditory systems, where the foundation for later cognitive architecture is laid down during sensitive periods for basic neural circuitry.

Genetic and Environmental Interplay in the Developing Brain

The uniquely large number of cells and their potential for association as well as asymmetry is directly the result of bipedalism along with genetic mutations. Once the large cell assemblies were established and pressures for bi-symmetry were released, humans then could develop asymmetric functions in their brains that were not directly tied to motor or autonomic control. Hemispheric specialization then could develop different control centers consistent with the previous function of that hemisphere, creating most of the unique human characteristics. The other demand that bipedalism would place on the brain would be the need to be more precise and complex in the synchronization of muscles to be able to walk, run, and jump. This increased synchronization would require greater frequency of oscillation of control centers within the inferior olive and cerebellum and their feedback to the intralaminar nucleus of the thalamus and its reciprocal thalamo-cortical projections. This increase in oscillation into the 40-Hertz range is thought to be required to achieve binding within various cortical sites into one continuous conscious percept of the world. This appears to be the foundation of human consciousness, which is thought to be unique in humans and due to unique connectivities in the human brain.

Therefore, a proposal of an increase in neuroblast proliferation in the human brain is consistent with the concept of neoteny in the human evolution. This concept states that certain characters are delayed in their development with respect to others. This resulted in changes in adult morphology during evolution. This is thought to be the process in the human skull in which infantile dimensions are comparable to other primates. This first factor explains the increase in cell size that concurs with minimal genetic change. However, the maintenance of these cells would not continue without the appropriate activity, presynaptically and postsynaptically. In essence, they require a power source as well as they would, in turn, require connections to expanded areas sub-cortically. Bipedalism would provide both by increasing exponentially the amount of temporal and spatial summation within sensory motor networks, especially cerebellum, thalamus, and cortex. This would require expanded areas of cerebellum and thalamus that would evolve in parallel with the expanded areas of cortex and could provide a site for connection to these increased numbers of neurons. This would take place because although the genetic change would increase cell number, it would do so with a non-directional force, which would not specify any specific shape. Posterior epigenetic reorganization (synaptic stabilization) would determine the shape and configuration of the networks within the brain itself. Therefore, genetic factors would produce the density of cells required but environmental factors would trim and shape it in a specific fashion.

Therefore, it can be speculated that there are no genes specifying particular types of neuronal networks involved in higher cognitive function. The human brain is about four times larger than those of primates, because the brain cells or neurons are spread about. The thicker cortices of the large mammals and humans as well seem to be primarily a function of larger nerve cell bodies, more extensive dendritic and axonal systems, and more numerous glial cells. Although neurons do not reproduce after birth, glial cells can. They reproduce based on increased metabolic demand of the neurons or increased stimulation. This increase in growth of glial cells allows the neurons to make more connections, which increases the ability and speed of the cell to transmit signal. The increase in size and strength of connections allows both to happen more efficiently. The growth in size and complexity of the human brain comes from the number of supporting cells which in essence feeds the neurons with more fuel and encourages the growth of new connections. It is not the increased number of neurons, but the increase in connections between the cells and the increase in separating and supporting cells that accounts for the large growth of the human brain. This is the very definition of “plasticity.”

Plasticity is the ability of the brain to grow and whether it is growing on a short-term basis or on a long-term basis in the case of evolution, the facts of plasticity are consistent. This can only mean that there was some increase in the frequency, duration, and intensity of stimulation of the human brain over time for it to have evolved as uniquely as it has. There are two things that make humans unique among other organisms: 1) we have a larger cortex and, 2) we stand upright (bipedal).

Refinements in the neural circuits that mediate sensory, emotional, and social behaviours are driven by experience. Specifically, postnatal experiences drive a protracted process of maturation at the structural and functional level, but the very ability of such developmental processes to occur successfully is dependent in large part on the prenatal establishment of the fundamental brain architecture that provides the basis for receiving, interpreting, and acting on information from the world around us.

While the term “blueprint” has been utilized in the past to describe a fixed set of genes with inflexible interactions, the term is used here as an analogy to a rough draft, or design – the framework from which a more defined structure will evolve, alternatively, an operating system in which programs have yet to be laid down. The emergence of the architecture in all vertebrate species begins early; in humans, this occurs within the first two months post-fertilization.

The cerebral cortex has garnered substantial attention in defining key developmental features across species. This is due in part to the technical advantages of studying a well-organized, layered structure, and the functional relevance of linking typical and atypical maturation of complex behaviors and neurodevelopment.

The neocortex in all mammalian species is comprised of six layers of neurons, the architecture, connectivity and chemistry of which are distinct depending upon their location. The neocortex is organized to receive information from the organism's surrounding environment, typically through connections with the thalamus. It does so by integrating information within and across architecturally distinct functional domains, and then relays this information to other brain centers that generate an appropriate functional response.

There are two major organizing principles of the neocortex influenced by gradients of gene networks that have developed evolutionarily. First, the precursors of different functional areas emerge during roughly the first and second trimester of pregnancy in the human. Regional specification is not absolute, but involves networks controlling the expression of axon guidance molecules that control the initial input and output wiring plan. Expansion of the size of the neocortex during evolution (e.g., 1000-fold between mouse and human) occurs mostly in this period.

The ‘inside-out’ pattern of neuron production and migration provides the basis for building cell connectivity forming functional areas, with small variations in the ratio of excitatory to inhibitory neurons in different regions. In fact, this organization provides a framework for later-developing refinement of circuits influenced extensively by patterns of physiological activity through experience and training.

Experiments in genetically manipulated mice demonstrate that altering the expression of just one genetic transcription factor, cortical regions can be changed. For example, the genetic factor emx2 controls the expression of the Fgf8 factor near the anterior end of the cerebrum. Fgf8 alone is sufficient to specify the cortical regions that will eventually receive connections that are typical of frontal and somatosensory cortices. This type of early genetic re-specification is functionally relevant. For example, the Fgf17 is responsible for initial patterning of different frontal cortex areas.

It is not our function here to pursue this notion in detail other than to indicate that the early specification and re-specification of the neocortex by genetic factors is powerful because additional axon guidance molecules serve as important chemical cues for getting axons to grow into their correct target region prior to beginning the extended process of synapse formation (cf. Alcamo et al., 2008). Gene regulatory networks also can influence the initial size of cortical areas by modulating the number of neurons produced. The long-distance circuit projections that help to define functional cortical areas, and even functional differences in superficial and deep projecting neurons, are altered when the disruption of early gene networks modifies guidance cues so that atypical connections are made.

Though we tend to think that genetic mechanisms are immutable, it is important to stress that expression of early gene networks can be perturbed not only by catastrophic genetic mutations that disrupt important regulatory genes, but also by prenatal environmental influences, such as drugs, alcohol, toxins, and inflammatory responses. These may have less profound impacts on brain patterning, but nonetheless can result in long-term disruption of cellular differentiation and behavioral development.

In all mammalian species, this early period of specified patterning to generate a unique architecture is followed by an extended period of synapse formation, adjustment, and pruning that typically extends from the last quarter of gestation through puberty.

Experience-based Adjustments to Neural Architecture in Early Childhood

Although genetics provides an important foundation for early development, it is only a framework upon which the early childhood environment can influence future structure and function. This can best be illustrated through studies of the sensory systems, which demonstrate the crucial role of environment in the early development and maintenance of the nervous system. Such work also demonstrates the need for patterned physiologic activity during development, as well as refinement and maintenance of detailed sensory maps.

Synaptic reorganization takes place most predominantly during childhood and adolescence (Blakemore, 2012). During these periods the brain becomes sensitive to change which allows it to develop in unique ways dependent upon the individual age, gender, and environment along with many other variables (Andersen, 2003). The concept of “self-organization” indicates that the brain actually organizes itself based on the individual's experiences. Environment stimulation and training can affect how the brain develops and at what pace (Andersen, 2003; Leisman, 2011). The environment can include factors like location and surroundings, home, parenting, and of course the classroom, as well as circumstances in each of those environments (Blakemore, 2012; Tau & Peterson, 2010). Environment can also be identified as a child's emotions or responses to certain stimuli, in this case, the concept of self-organization which postulates that the brain organizes itself based on each child's unique experiences.

The fact that humans have a greater capacity than rats or even chimps for self-organizing, plastic, or flexible behavior provides no implication that we are either all stereotyped or flexible in our behavior and brain organization. Stereotypy creates for efficiencies but plasticity or flexibility allow for adaptation due to the exigencies of one's environment. We, given the notions of stability and flexibility (Leisman, 1980), have a basis for rehabilitation and effective adaptive function. The concept of the interplay between stability and flexibility and its implications for the education of the normally developing child's brain needs to be viewed as a relativistic notion, viewed against the features of the organism that are not plastic. In order to identify flexibility or plasticity, one must be able to identify the invariant and constant. The identification of plasticity requires us to be able to know the constraints of the system. The fact, however, that we are more plastic than other organisms is expressed even in our adult lives as organisms. This suggests that our capacity for systematic change and the fact that we retain flexibility across our later developmental periods allows application of rehabilitation thinking and the measurement of optimization throughout the life span.

Hebb had postulated in 1949 that when one cell excites another repeatedly, a change takes place in one or both cells such that one cell becomes more efficient at firing the other (Hebb, 1949). It is this view that is not only limited to a particular cell and its arborized neuronal connections but to definable anatomical regions. It is this notion that forms the basis of our concept of plasticity.

Hebb was the first to propose the ‘enriched environment’ as an experimental concept. He reported anecdotally that laboratory rats that nurtured at home as pets were behaviorally different than their littermates kept at the laboratory. Hebb was not the only one who conceptualized the effects of enriched nurturance having an effect on nervous system structure and function. Hubel and Wiesel examined the effects of selective visual deprivation during development on the anatomy and physiology of the visual cortex and Rosenzweig and colleagues introduced enriched environments as a testable scientific concept by measuring the effects of environment on ‘total brain weight,’ ‘total DNA or RNA content,’ or ‘total brain protein’. Numerous researchers have demonstrated a significant linkage between enrichment and neurological plasticity that have included biochemical changes, gliogenesis, neurogenesis, dendritic arborization, and improved learning and memory.

Related Questions

Navigate

• Browse (All)
• Browse A
• Subjects

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.