Skip to main content

Neural Networks

                 

                            

                        Neural Networks



An artificial neural network is a collection of smaller units called neurons, which are computing units modeled on the way the human brain processes information. 


Artificial neural networks borrow some ideas from the biological neural network of the brain, in order to approximate some of its processing results. These units or neurons take incoming data like the biological neural networks and learn to make decisions over time. Neural networks learn through a process called backpropagation. Backpropagation uses a set of training data that match known inputs to desired outputs. First, the inputs are plugged into the network and outputs are determined. Then, an error function determines how far the given output is from the desired output. Finally, adjustments are made in order to reduce errors. A collection of neurons is called a layer, and a layer takes in an input and provides an output. Any neural network will have one input layer and one output layer. It will also have one or more hidden layers which simulate the types of activity that goes on in the human brain. Hidden layers take in a set of weighted inputs and produce an output through an activation function. A neural network having more than one hidden layer is referred to as a deep neural network. 


Perceptrons are the simplest and oldest types of neural networks. They are single-layered neural networks consisting of input nodes connected directly to an output node. Input layers forward the input values to the next layer, by means of multiplying by a weight and summing the results. Hidden layers receive input from other nodes and forward their output to other nodes. Hidden and output nodes have a property called bias, which is a special type of weight that applies to a node after the other inputs are considered. Finally, an activation function determines how a node responds to its inputs. The function is run against the sum of the inputs and bias, and then the result is forwarded as an output. Activation functions can take different forms, and choosing them is a critical component to the success of a neural network. Convolutional neural networks or CNNs are multilayer neural networks that take inspiration from the animal visual cortex. CNNs are useful in applications such as image processing, video recognition, and natural language processing. A convolution is a mathematical operation, where a function is applied to another function and the result is a mixture of the two functions. Convolutions are good at detecting simple structures in an image, and putting those simple features together to construct more complex features. In a convolutional network, this process occurs over a series of layers, each of which conducts a convolution on the output of the previous layer. CNNs are adept at building complex features from less complex ones. Recurrent neural networks or RNNs, are recurrent because they perform the same task for every element of a sequence, with prior outputs feeding subsequent stage inputs. In a general neural network, an input is processed through a number of layers and an output is produced with an assumption that the two successive inputs are independent of each other, but that may not hold true in certain scenarios. 


For example, when we need to consider the context in which a word has been spoken, in such scenarios, dependence on previous observations has to be considered to produce the output. RNNs can make use of information in long sequences, each layer of the network representing the observation at a certain time.




Avinash C. Pillai

Technology Director

syniverse® 

The world’s most connected company™ 

Website / Twitter / LinkedIn/ connected company™  


Comments

Popular posts from this blog

Seven Personal Qualities Found In A Good Leader

Whether in fact a person is born a leader or develops skills and abilities to become a leader is open for debate. There are some clear characteristics that are found in good leaders. These qualities can be developed or may be naturally part of their personality. Let us explore them further. Seven Personal Qualities Found In A Good Leader: 1. A good leader has an exemplary character. It is of utmost importance that a leader is trustworthy to lead others. A leader needs to be trusted and be known to live their life with honestly and integrity. A good leader “walks the talk” and in doing so earns the right to have responsibility for others. True authority is born from respect for the good character and trustworthiness of the person who leads.   2.A good leader is enthusiastic about their work or cause and also about their role as leader. People will respond more openly to a person of passion and dedication. Leaders need to be able to be a source of inspiration, and b...

Defining AI Ethics

                           Defining AI Ethics Welcome to Defining AI Ethics. Humans rely on culturally agreed-upon morals and standards of action — or ethics — to guide their decision-making, especially for decisions that impact others. As AI is increasingly used to automate and augment decision-making, it is critical that AI is built with ethics at the core so its outcomes align with human ethics and expectations. AI ethics is a multidisciplinary field that investigates how to maximize AI's beneficial impacts while reducing risks and adverse impacts. It explores issues like data responsibility and privacy, inclusion, moral agency, value alignment, accountability, and technology misuse …to understand how to build and use AI in ways that align with human ethics and expectations.  There are five pillars for AI ethics: explainability, fairness,...

Command to Find Version of Linux OS

Command to Find Version of Linux OS # cat /etc/redhat-release Regards Avinash Pillai URL : http://avinashpillai.blogspot.com Email: avinashp[AT]aztecsoft[DOT]com, avinashp25[AT]gmail[DOT]com