Neural Network: An Art to Mimic Human Brain
It is a bit ambivalent to decide where exactly one should begin in-order to start with a note on the Human brain. Human brain is one of the most sophisticated machines ever known to mankind. Be it’s veracious ability to store everyday activities(data) or monitoring the complex human system in a restive manner, it has always been an inspiration and even more a mystery to us. It controls most of the activities of the body, processing, as well as coordinating the information it receives from the sense organs, and makes the decisions thereafter. A human brain begins to function from the moment of birth and works till the person’s last breath, doing its tasks with utmost efficiency. The sole credit to which humans are regarded as the pinnacle of evolution goes to the human brain.
What makes our brain extremely unique is its salient feature to learn from the previous experiences. A famous quote by Jeff Bezos on the human brain elaborates that the brain is an incredible pattern-matching machine. This unique ability of the human brain laid the foundation to numerous modern day branches which are collectively called Artificial Intelligence. One such domain which is branched under A.I. is Neural Network. The definition of Neural Network is quite ambiguous. One among them regards Neural Network as the study of the internal architecture of the human brain in-order to induce intelligence artificially on machines. If you google the word Neural Network for its definition, this is the result which pops up: “a computer system modelled on the human brain and nervous system”. Since it is exclusively based on modelling the human brain, let us plunge into the structure of the human brain.
Human brain is the central organ of the nervous system. It consists of the cerebrum, the medulla and the cerebellum. But at the cellular level, the brain is a cluster of basic units called neurons. The neuron, the basic working unit of the brain and the nervous system, has a cell body, an axon and a dendrite. The cell body is the powerhouse of a neuron which consists of nucleus and cytoplasm. Arising from the cell body is the axon which further branches out into smaller terminals called nerve terminals. Dendrites extend from the neuron cell body and receive messages from other neurons. The whole activities of the brain are characterized by the movement of nerve impulses in the neurons. This forms a system through which nerve cells communicate. The impulses originate from the dendrites of a neuron, get carried away to the nerve terminal via the cell body and the axon. From the nerve terminal, it is transmitted to the nearby neurons. Nerve impulses from the neighboring neurons are collected by the dendrites, get added up and are passed on to the cell body and axon for further transmission. This process repeats itself throughout the course of time. This transmission and reception of nerve impulses by the neurons form the basics of our characters such as intelligence, actions, reactions and so on.
Drawing up the knowledge on neurons, let me explain the concept of Neural Networks. A Neural Network is basically a dense interconnection of layers, which are further made up of basic units called perceptrons, which draw its basic functionalities from a neuron. A perceptron consists of input terminals, the processing unit and the output terminals. The input terminals of a perceptron are connected to the output terminals of the preceding perceptrons. Like a neuron, a perceptron receives all the input signals from the preceding perceptrons via the input terminals. The processing unit sums up all the signals received from the input terminals, applies an activation to it and transmits the resultant signal to the output terminals.
A perceptron receives a set of input values. The perceptron calculates a weighted average of the values of the vector input, based on a weight vector w and adds bias to the results. The result of this calculation is passed through a non-linear activation function, which forms the output of the unit. A cluster of these perceptrons are arranged to form dense layers which are further interconnected to form what we perceive as a Neural Network. A Neural Network is a collection of layers which are stacked in a sequential manner. The output of one set of layers is the input for the next set. Deep learning, an application of Neural Network is purely based on the construction of neural layers.
The result of the weighted average of the input vectors is passed on to an activation function. Activation function forms one of the key features of a Neural Network. Traditional Machine Learning methods are based on the idea of linearity, and focus predominantly on the linear modelling of a system. However Neural Networks are solely based on non-linearity. The whole rhythm of these networks depends on its ability to model a system non-linearly. Hence in the absence of an activation function, a Neural Network becomes a linear combination of input signals. The non-linearity element allows for greater flexibility and creation of complex functions during the learning process. However during the construction of a Neural Network, care should always be taken in-order to avoid the model into overfitting. For those who want to know more about overfitting and its remedies, please read the following blog written by me.
Neural Network serves as a reference to built a series of algorithms that establishes some relationships in a data. It laid the foundation to the modern day Contextual Natural Language Processing(NLP). A Neural Network is trained by varying the weight vector w of a perceptron while simultaneously keeping a track on it’s performance on a specified dataset. All the pre-processing methods which we applied on traditional Machine Learning are also applicable for Neural Networks for a better performance. Neural Network has already become one of the most widely used Artificial Intelligence tools. They are used for solving many business problems such as forecasting, customer research, data analytics and validation, and risk management.
In an upcoming blog, I will discuss in detail the mathematics behind the Neural Networks. For now, I want you to get a clear picture of what exactly the neural network is and how exactly it works.
In this blog, I have presented you with the basic idea of a Neural Network and how exactly it works. I hope this blog was helpful and would have galvanised you to dive deep into topic
