Introduction to Neural Networks

Here you will get an introduction to neural networks in the field of data science.

Neural networks are similar to biological neural network. Biological neural network is collection of biological neurons in human brain similarly Neural network is collection of nodes called Artificial neurons.

Neural networks are based on non-task specific programming concepts like in image recognition they learn to resemble images by analyzing sample images labeled with name as “car” or “no car” and by using such sample example they identify car in other images. They need not to acquire any knowledge regarding car like it has engine, four wheels, shape and so on. They generate their own relevant characteristics from the process of their learning material.

Neural Network is a feature of artificial intelligence that efforts to copy the way human brain works. Neural network perform its operation by connecting the processing elements rather than doing all computations that manipulate zeros and ones in digital model

Neural network is used to represent relationships between complex input/output and also it is capable to capture data the same way human brain works. This idea for the development of neural network technology arises from the desire to perform all task intelligently similar to human brain and develop an artificial system to perform all this task.

Neural network gains knowledge through learning. Synaptic weight i.e. inter-neuron connection strengths are used to store a neural network’s knowledge.

Basic Organization of Neural Network

In Neural network, each connection called synapse between nodes or artificial neurons is used to transmit signal from one another and the receiving neuron called postsynaptic can process the signal and information and thereafter signal neurons connected to it.

Generally, neurons are arranged in layers. Different layers may perform different transformation with their inputs. Signals travelling from input i.e. the first layer to output i.e. the last layer, possibly have to traverse the layers multiple times.

In neural networks the sending signal (synapse) is real number and output is calculated by non-linear function. Output of each neuron is calculated by non-linear function of the sum of its inputs. Synapses and artificial neurons have weights that are adjustable as learning proceeds. The weight of neurons increases or decreases the signal strength that it has to send over synapse .

Neural network solve the problem the same way human brain do. Neural network have been used in computer speech recognition, video games, machine translation and medical diagnosis.

Introduction To Neural Networks

Image Source

Types of Neural Network

1. Feedforward Neural Network

Feedforward neural network is simplest of all neural network. It has no cycles or loops in its network that is it moves information in one direction only. The data in feedforward network moves from input to output nodes passing through hidden layer if there is any hidden layer in between the input and output node.

2. Radial Basis Function Neural Network

Radial basis function neural network is highly spontaneous neural network. It is the first choice in multidimensional space during interpolation. In radial basis function neural network, each neuron saves an example from training set as a “prototype”. Radial basis function has advantage that it doesn’t suffer from local minima offered by linearity involved in functioning of it.

3. Kohonen Self-Organizing Neural Network

Kohonen self-organizing neural network is used to visualize high dimensional data as low dimensional views. It is invented by Teuvo Kohonen. Kohonen self-organizing  neural network is used to describe hidden structures in it by performing functions on unlabeled data. competitive learning is applied by self-organizing neural network to a set of input data not error correction learning applied by other neural network.

 4. Recurrent Neural Network

Recurrent Neural Network is used for parallel and sequential computation it is used to compute each and every thing similar to traditional computer. Recurrent neural  network  works  similar to human brain, it is a large feedback network of connected neurons that can translate a input stream into a sequence of motor outputs. A recurrent neural network (RNN) use their internal memory to process sequencing of inputs. In recurrent neural network, connections between units form a directed cycle. Recurrent neural network model each vector from sequence of input stream vectors one at time. This allows  the network to retain its state during modeling of each input vector across the window of input vectors.

5. Modular Neural Network

Modular neural network consist of series of independent neural networks that are operated by intermediary. Each independent neural network act as a module and works with separate input that network hopes to perform by accomplishing subtask. The intermediary accepts the output of each module and helps them to process the final output for modular neural network. Also the independent neural network i.e. the modules do not interact with each other.

6. Physical Neural Network

The physical neural network aims to focus on physical hardware along with the software while simulating the neural network. A resistance material that is electrically adjustable is used for emulating the function of neural synapse. When the physical hardware emulates the neurons, the software emulates the neural network.

Applications of Neural Networks

  • Speech Recognition
  • Character Recognition
  • Signature Verification Application
  • Human Face Recognition

About Author:

Shubham Sharma, currently working as Analytics Engineer in Data Science Domain. Has around 2+ years of experience in Data Science. Skilled in Python, Pandas, Anaconda, Tensorflow, Keras, Scikit learn, Numpy, Scipy, Microsoft Excel, SQL, Cassandra and Statistical Data Analysis, Hadoop, Hive, Pig, Spark, Pyspark. Connect with him at [email protected].

Comment below if you found any information incorrect or have doubts related to above tutorial for introduction to neural networks.

Leave a Comment

Your email address will not be published. Required fields are marked *