difference between fitcnet and patternnet functions

I am not able to get difference between fitcnet and patternnet functions; when to use which one and what change happens in the result, if one replaced by other?

Answer 1

[Conor Daly]

fitcnet and patternnet can both be used to solve tabular classification problems.

patternnet is used to define a network architecture which can then be passed to the train function, along with training data, to train a network. fitcnet defines the network architecture and trains the network based on training data in a single line of code.

There are some differences between the two approaches. For example, fitcnet uses the L-BFGS optimizer to train the model. patternnet defaults to the scaled conjugate gradient optimizer -- though others are available. In addition, the ClassificatioNeuralNetwork object returned by fitcnet has properties and methods common to the other fitc* functions for tabular classification -- for example predict, loss and edge.

Finally, note that fitcnet is available in the Classification Learner app, which facilitates easy comparison of multiple machine learning models for tabular classifcation problems.

Answer 2

[Girijashankar Sahoo]

1. FITNET for regression (MATLAB calls it curve fitting) which is supposed to be a replacement for NEWFF)

2. PATTERNNET for pattern recognition and classification ( which were previously achieved using NEWFF)

[Yogini Prabhu]

okay thanks

[Yogini Prabhu]

fitcnet

Train neural network 'classification' model

Description

Use fitcnet to train a feedforward, fully connected neural network for classification. The first fully connected layer of the neural network has a connection from the network input (predictor data), and each subsequent layer has a connection from the previous layer. Each fully connected layer multiplies the input by a weight matrix and then adds a bias vector. An activation function follows each fully connected layer. The final fully connected layer and the subsequent softmax activation function produce the network's output, namely classification scores (posterior probabilities) and predicted labels. For more information, see Neural Network Structure.

Answer 3

[pathakunta]

1. FITNET for regression (MATLAB calls it curve fitting) which is supposed to be a replacement for NEWFF) 2. PATTERNNET for pattern recognition and classification ( which were previously achieved using NEWFF)