neuraldeep
v1.1.5
Published
Create, train, test and compare perceptron neural networks
Downloads
7
Maintainers
Readme
NeuralDeep
Now yo can create, train, test and compare easly diferents secuential neural networks with diferents architectures for select the architecture that better solve your problems.NeuralDeep allow to see the performance and the error rate of differents neural networks and compare these with a ranking table. Also you can to export your neural networks to a json file to use with Synaptic.
- The neural networks only can recive binary inputs and only send binary outputs.
NeuralDeep is a console interface based in the library of deep learning named: Synaptic.
Index
- Requirements
- Install
- Use
- Train and Create a new Neural Network
- Execute a Neural Network
- Test a Neural Network
- Compare Two or More Neural Networks
Requirements
Install
npm install -g neuraldeep
Use
- Create a neuraldeep project:
neuraldeep init <projectName>
- projectName: String with the name of the project (One project can contain multiples neural networks).
- Go to the root of the project:
cd <projectName>
- Now you can start using neuraldeep.
For use this CI, first of all you must create a neural network, making a datasheet more info...
Then you can execute, test and compare the neural networks.
Train and Create a new Neural Network
- First of all you have to create the training data to train the neural network, so you have to create a training data file in the trainingData folder, following the syntax of the training data file example.
- The name of the trainingData file must be the same name of the neural network.
- Now to create the network, you must execute the following command:
neuraldeep create <name> <architecture>
name: The name of the Neural Network and the Training Data File.
architecture: The Architecture of the Neural Network. Input,deep and output neurons.
example:
neuraldeep create neuralNetwork1 20 4 4 2
. This command creates a neural network named neuralNetwork1 with 20 input neurons, two hidden layers with 4 neurons and two output neurons.The Neural Network is saved in the neuralNetwork folder.
Execute a Neural Network
- First, you must have to create a Neural Network with the neuraldeep create command. After that, to execute the Neural Network created you must use the command:
neuraldeep run <name> <binaryInputArray>
- name: The name of the Neural Network and the Training Data File.
- binaryInputArray: The input of the Neural Network in array format, for example If the neural network have 8 input neurons, the input may be [0,0,0,0,0,0,1,1].
Test a Neural Network
- First of all you have to create a neural network (with neuraldeep create) and the test data to test it, so you have to create a test data file in the testData folder, following the syntax of the test data file example.
- The name of the testData file must be the same name of the neural network.
- Now to test the network, you must execute the following command:
neuraldeep test <name>
name: The name of the Neural Network and the Test Data File.
This command has the extensive option (-e or --extensive) whitch showing all failed tests.
Compare Two or More Neural Networks
- First of all you have to create two or more neural network (with neuraldeep create),the neural network must have the same base name: baseName_version and the test data to test its, so you have to create a test data file in the testData folder, following the syntax of the test data file example. (the name of the test data file must be the base name of the neural networks).
- Now to compare the neural networks, you must execute the following command:
neuraldeep compare <baseName>
baseName: The name of the test data file and the base name of the neural networks.
This command has the top option (-t or --top) whitch recive the neural items number and show a ranking table until the last item number.