Artificial Neural Network – Introduction , Learning Process & Applications

Artificial Neural Networks , also termed as ANNs is a fundamental part of Artificial Intelligence . ANNs mainly deals with data processing , analyzing and predicting the best result possible . It is supposed to function just like animal brain or animal nervous system with hundreds of thousands of neurons interconnected to each other hence the name neural network . Let’s talk about Artificial Neural Networks in detail .

What is Artificial Neural Network ?

Artificial Neural Network (ANN) is information processing computational system that is inspired by living neural network of animal nervous system which constitutes animal brain .

ANN is composed of large number of processing elements called neurons which are interconnected to each other and function in unison to solve problems . ANNs are unlike other computer programs which work on specific problem but they are trained with huge data set and they learn by training just like humans do . As they are not task specific , they learn progressively improving performances each time they perform . ANN is configured for certain application purpose such as Image Recognition , Pattern Recognition , Data Classification through Learning Processes . For example ,  Image Recognition technology with ANN might identify images with tigers analyzing example images with label ‘tiger’ or ‘no tiger’ and using the analytic results to identify tigers in other images.

Like animal nervous system is composition of neurons ( axons and dendrites ) , ANN is also formed with artificial neurons , interconnected to each other at node . Each neuron is able to transmit signal to others and others can receive signals transmitted to them and vice versa . Neurons have state which is typically represented by real numbers between 0 and 1 .

ANNs V/S Conventional Approach

ANN is completely different in approach than traditional/conventional computational system . Traditional or conventional computers work with algorithmic approach which is just meant to solve specific problem . They can not do other than what they are instructed to do .  That way , we can not get our problem solved  if the problem is unknown to us . They can only do what we already know how to solve . This restricts conventional computers to solve problems that humans know to solve . The problem must be known and instructed to computer . The instruction is converted to high level programming language and then converted to machine level code which computer can understand . In this way specific problem is solved in conventional approach . But the thing is that computers would be worthy if they could solve problems we can’t . Neural Network works just like human brain functions . The network is composed of a large number of highly interconnected processing elements(neurons) working in parallel to solve a specific problem. Neural networks learn by example. They cannot be programmed to perform a specific task. The examples must be selected carefully otherwise useful time is wasted or even worse the network might be functioning incorrectly. The disadvantage is that because the network finds out how to solve the problem by itself, its operation can be unpredictable.

Goal Of Artificial Neural Network

The goal is obvious . ANNs are intended to solve problems same as human mind does . The main purpose of Artificial Neural Network is to learn from data and make decisions to various scenarios regardless the types of data set , just like animal brain functions . As ANN is the computational model based on structure and functions of animal neural network (biological neural network ) , the network is affected by each and every data information that flows through it . This is because , Neural Network is capable of learning by inputs and outputs made each time  .

Learning Process Of Artificial Neural Network

As ANN is inspired by the structure and functions of human brain , learning process of ANN is similar to that of human brain . To understand how ANN learns , we must understand how our brain works or trains itself .

How Does Human Brain Learn ?

Human brain is composed of hundred of thousands of neurons interconnected to each other . There are two major structures by which neuron is complete . One is Axon by which neuron sends out spikes of electrical activity and another is dendrites from which neuron collects signals . Axon splits into thousands of branches and at the end of each branch , a structure called a synapse converts the activity from the axon into electrical effects that inhibit or excite activity in the connected neurons . 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.

neuron

Components Of Neuron : Image Source

Basic structure And Working Of Artificial Neural Network

Since the working of human brain (nervous system ) is still unclear , we are limited to make advanced ANN models . We have to stick with what we have been taught from past decades . The base model is just similar as I wrote above about human brain . In each neuron of ANN , there are many inputs and one output as dendrites and axon in human brain . An ANN functions in two different modes . One is Training Mode in which neuron can be trained with example data whether it should fire or not for given pattern . Another mode is Using Mode . In this mode , when  a taught pattern is detected , the output is obvious and if pattern is not found in taught list , Firing Rule is applied to determine whether to fire or not .

neuron model

The Neuron Model : Image Source

Learning Paradigms

The major learning paradigms of Artificial Neural Network are Supervised Learning , Unsupervised Learning and Reinforcement Learning .

  1. Supervised Learning
    Supervised Learning is a learning process in which ANN uses example sets and then intends to find the output of given problem that matches example sets .It uses external teacher to get trained for future problems . Paradigms of supervised learning include error-correction learning, pattern recognition , regression , stochastic learning…etc .
  2. Unsupervised Learning
    Unsupervised Learning is based on the local information not on external teacher . It is also referred to as self-organisation, in the sense that it self-organizes data presented to the network and detects their emergent collective properties. Paradigms of unsupervised learning are Hebbian Learning and competitive learning.
  3. Reinforcement Learning
    Reinforcement is another learning process of ANN which is completely different from supervised and unsupervised learning . It seeks for a result data which has already been received after performing certain tasks . The paradigm of reinforcement learning are control problems, games and other sequential decision making tasks.

Applications Of Artificial Neural Network

ANNs are very applicable because they are capable of solving very complex problems even humans could feel tough or may not be able to solve . As ANNs mainly deal with pattern in data and prediction making , they are very helpful at data analysis and forecasting . The major applications of ANN are :-

Forecasting

  • sales forecasting
  • industrial process control
  • customer research
  • data validation
  • risk management
  • target marketing

Medicine

  • Disease Recognition From Scans
  • Neuroscience
  • Modelling and Diagnosing the Cardiovascular System
  • Electronic Noses ( Telemedicine )
  • Instant Physician

Business

  • Marketing
  • Credit Evaluation

Technology

  • Recognition (Speech , Hand Writing , Face )
  • Image Processing
  • Language Processing