China
Africa
Explore chapters and articles related to this topic
Supervised Learning
Published in Peter Wlodarczak, Machine Learning and its Applications, 2019
Multilayer perceptrons are a class of feedforward networks that consist of at least three layers: An input layer, one or more hidden middle layers and an output layer, as shown in Figure 4.2. Except for the input layer, the layers consist of perceptrons with nonlinear activation functions. Feedforward networks pass the signal in one direction from the input layer through the middle layer(s) to the output layer without any cycles. This is called forward propagation. The multilayer perceptron can separate data points that are not linearly separable. Typically, a multilayer perceptron uses a sigmoid activation function. Not all neurons within the same neural network need to have the same type of activation function. Typical activation functions in a multilayer perceptron include the sigmoid or logistic activation function, shown in equation 4.4, that is also used in logistic regression, and the hyperbolic tangent, shown in equation 4.5. () y(xi)=(1+e−xi)−1
Evolving Artificial Neural Networks in a Closed Loop
Published in Yi Chen, Yun Li, Computational Intelligence Assisted Design, 2018
The majority of artificial neural networks in use have an architecture of multilayer perceptrons. In such a network, the perceptrons, or neurons, of each layer are not connected to one another but completely connected to the perceptrons in the adjacent layers. This architecture forms a partitioned complete digraph [Rogers and Li (1993), Li and Rogers (1995)]. In the design process for practical applications, the network must be trained to gain an optimal connection‐strength represented by weights in the connecting path. The weighted input to a neuron is further processed by an activation function, together with a linear shift operation determined by another parameter, the threshold level, which also needs to be trained before the network is applied to solve problems.
Multilayer Perceptron
Published in Sing-Tze Bow, Pattern Recognition and Image Preprocessing, 2002
As in a single layer perceptron, each neuron in a multilayer perceptron is also modeled as fi[∑i(wjiOi+bias)] or fj (netj), except that the activation function is a sigmoidal function instead of a hard limiter. Note that Oi here refers to the output of the neuron in the previous layer. It is also the input to this neuron. For the jth neuron of the output layer, the activation function is of the following form: () Ojk=fi(netjk)=fi(∑iwjiOik+θj)
An efficient approach for land record classification and information retrieval in data warehouse
Published in International Journal of Computers and Applications, 2021
C. B. David Joel Kishore, T. Bhaskara Reddy
Summation process with bias can be defined as: The activation process can be used in different types of processes, such as sigmoid activation function and a hyperbolic tangent activation function, which means multilayer perceptron is a sigmoidal activation in the form of a hyperbolic tangent. It is a real mathematical work that denotes for all actual input values and has a non-negative derivative at each point. The hyperbolic tangent is a ratio of corresponding hyperbolic sine and hyperbolic cosine function in Figure 2. The backpropagation algorithm is used for weight changes in hidden layers and output layers. This type of algorithm uses the delta-rule, which computes deltas, such as local gradients in each data going backward direction to reach the input layers. The delta rule needs weight to match the desired output. Adjust the weight for delta’s rule by adding current weight to produce a new weight for the output layer. The difference between the current weights and previous weights are multiplied by a momentum coefficient.
An application of local linear radial basis function neural network for flood prediction
Published in Journal of Management Analytics, 2019
Binaya Kumar Panigrahi, Tushar Kumar Nath, Manas Ranjan Senapati
A multilayer perceptron, shown in Figure 1, is an artificial neural network comprising of an input layer of neurons, at least one or more than one hidden layers and a final layer called output layer. The neurons in the input layer receive the input, computation is done at the hidden layer and the output layer makes decision or prediction about the input. Each connection is associated with some weights and some bias values are added to the neurons (Elsafi, 2014; Panigrahi et al., 2018). The output of neuron j in the hidden layer is given as,where α() is called activation function, N the number of input neurons, the weights, inputs to the input neurons, and the bias value of the hidden neurons.
Exploiting Machine Learning for the Identification of Locomotives’ Position in Large Freight Trains
Published in Applied Artificial Intelligence, 2019
Helder Arruda, Orlando Ohashi, Jair Ferreira, Cleidson De Souza, Gustavo Pessin
The Multilayer Perceptron is a system of interconnected neurons where each neuron is connected to the neurons of previous and next layer. The system is divided into input layer, hidden layers and output layer. For classification purposes, which is the aim of this work, the output layer is represented by each one of the classes as a node. The Multilayer Perceptron is a feed-forward artificial neural network and learns in supervised mode. The algorithm used for training was backpropagation, which is one of the most basic algorithms for this kind of task (Gardner and Dorling 1998).