Opportunities @ sun Sushant Kumar

Create a neural net in easy five steps

Project Home

When I first used tools like JOONE, it took me quite a while
to understand its usability.Although the tool was nothing less that awesome, i wished it
was more easy to use and then thought, why not make a tool with complete automation
support with is as you’ll see is done in 5 easy steps.

- ragha

• Step 1: Create a
  network
  

Set up the num of inputs and outputs to the neural network.Add desired num of hidden layers (In the future we are going to add an auto detect mode which
will be able to decide the neural network configuration based on given train
data)

• Step 2: Set up the training data

Create a txt file with training data and load it.The format is as follows:

Ex - For a network with 2 inputs and outputs

1.2,2.2;0.5,0.43

Char ‘;’ separated the input and output set and ‘,’ separates the values within the set

• Step 3: set up validation data (optional)

Loading the data and its format is similar to the train data If used, the validation data will be used to determine the
performance of the network…

Otherwise conventional ways are used. Validation data can be used to prevent over-fitting of the
data

• Step 4: Train the network

Here you can set the required parameters and start training
the network. A graph is shown to indicate the current error.The graph could be used determine if the network is already
saturated and then you could decide to stop training to prevent over-fitting

• Step 5: Analyze the performance

Here you could test and evaluate the network to see how well
it performs on unknown inputs (This part is yet to be constructed) The network could also be saved and loaded at any time.

Future plans for the tool include automating of the complete
procedure ie, you just have to supply the train data and rest of the
parameters are auto detected (although this mode will consume more time, it will help
save all your trials and errors)

We’ll keep you posted with more improvements

- JNeuralNet Team