Main Content

Deploy Shallow Neural Network Functions

Deployment Functions and Tools for Trained Networks

The function genFunction allows stand-alone MATLAB® functions for a trained shallow neural network. The generated code contains all the information needed to simulate a neural network, including settings, weight and bias values, module functions, and calculations.

The generated MATLAB function can be used to inspect the exact simulation calculations that a particular shallow neural network performs, and makes it easier to deploy neural networks for many purposes with a wide variety of MATLAB deployment products and tools.

The function genFunction is used by the Neural Net Fitting, Neural Net Pattern Recognition, Neural Net Clustering and Neural Net Time Series apps. For information on these apps, see Fit Data with a Shallow Neural Network, Pattern Recognition with a Shallow Neural Network, Cluster Data with a Self-Organizing Map, and Shallow Neural Network Time-Series Prediction and Modeling.

The comprehensive scripts generated by these apps includes an example of deploying networks with genFunction.

Generate Neural Network Functions for Application Deployment

The function genFunction generates a stand-alone MATLAB function for simulating any trained shallow neural network and preparing it for deployment. This might be useful for several tasks:

  • Document the input-output transforms of a neural network used as a calculation template for manual reimplementations of the network

  • Use the MATLAB Function block to create a Simulink® block

  • Use MATLAB Compiler™ to:

    • Generate stand-alone executables

    • Generate Excel® add-ins

  • Use MATLAB Compiler SDK™ to:

    • Generate C/C++ libraries

    • Generate .COM components

    • Generate Java® components

    • Generate .NET components

  • Use MATLAB Coder™ to:

    • Generate C/C++ code

    • Generate efficient MEX-functions

genFunction(net,'pathname') takes a neural network and file path, and produces a standalone MATLAB function file filename.m.

genFunction(...,'MatrixOnly','yes') overrides the default cell/matrix notation and instead generates a function that uses only matrix arguments compatible with MATLAB Coder tools. For static networks, the matrix columns are interpreted as independent samples. For dynamic networks, the matrix columns are interpreted as a series of time steps. The default value is 'no'.

genFunction(___,'ShowLinks','no') disables the default behavior of displaying links to generated help and source code. The default is 'yes'.

Here a static network is trained and its outputs calculated.

[x, t] = bodyfat_dataset;
bodyfatNet = feedforwardnet(10);
bodyfatNet = train(bodyfatNet, x, t);
y = bodyfatNet(x);

The following code generates, tests, and displays a MATLAB function with the same interface as the neural network object.

genFunction(bodyfatNet, 'bodyfatFcn');
y2 = bodyfatFcn(x);
accuracy2 = max(abs(y - y2))
edit bodyfatFcn

You can compile the new function with the MATLAB Compiler tools (license required) to a shared/dynamically linked library with mcc.

mcc -W lib:libBodyfat -T link:lib bodyfatFcn

The next code generates another version of the MATLAB function that supports only matrix arguments (no cell arrays). This function is tested. Then it is used to generate a MEX-function with the MATLAB Coder tool codegen (license required), which is also tested.

genFunction(bodyfatNet, 'bodyfatFcn', 'MatrixOnly', 'yes');
y3 = bodyfatFcn(x);
accuracy3 = max(abs(y - y3))

x1Type = coder.typeof(double(0), [13, Inf]); % Coder type of input 1
codegen bodyfatFcn.m -config:mex -o bodyfatCodeGen -args {x1Type}
y4 = bodyfatCodeGen(x);
accuracy4 = max(abs(y - y4))

Here a dynamic network is trained and its outputs calculated.

[x,t] = maglev_dataset;
maglevNet = narxnet(1:2,1:2,10);
[X,Xi,Ai,T] = preparets(maglevNet,x,{},t);
maglevNet = train(maglevNet,X,T,Xi,Ai);
[y,xf,af] = maglevNet(X,Xi,Ai);

Next a MATLAB function is generated and tested. The function is then used to create a shared/dynamically linked library with mcc.

genFunction(maglevNet,'maglevFcn');
[y2,xf,af] = maglevFcn(X,Xi,Ai);
accuracy2 = max(abs(cell2mat(y)-cell2mat(y2)))
mcc -W lib:libMaglev -T link:lib maglevFcn

The following code generates another version of the MATLAB function that supports only matrix arguments (no cell arrays). This function is tested. Then it is used to generate a MEX-function with the MATLAB Coder tool codegen, which is also tested.

genFunction(maglevNet,'maglevFcn','MatrixOnly','yes');
x1 = cell2mat(X(1,:)); % Convert each input to matrix
x2 = cell2mat(X(2,:));
xi1 = cell2mat(Xi(1,:)); % Convert each input state to matrix
xi2 = cell2mat(Xi(2,:));
[y3,xf1,xf2] = maglevFcn(x1,x2,xi1,xi2);
accuracy3 = max(abs(cell2mat(y)-y3))

x1Type = coder.typeof(double(0),[1 Inf]); % Coder type of input 1
x2Type = coder.typeof(double(0),[1 Inf]); % Coder type of input 2
xi1Type = coder.typeof(double(0),[1 2]); % Coder type of input 1 states
xi2Type = coder.typeof(double(0),[1 2]); % Coder type of input 2 states
codegen maglevFcn.m -config:mex -o maglevNetCodeGen ...
                    -args {x1Type x2Type xi1Type xi2Type}
[y4,xf1,xf2] = maglevNetCodeGen(x1,x2,xi1,xi2);
dynamic_codegen_accuracy = max(abs(cell2mat(y)-y4))

Generate Simulink Diagrams

For information on simulating shallow neural networks and deploying trained neural networks with Simulink tools, see Deploy Shallow Neural Network Simulink Diagrams.

Related Topics