Neural Network NNTOOL


Explore how the training accuracy of neural network depneds on: - Number of features used to describe the pattern - Number of classes - Number of training samples

Necessary: to select the task you want the neural network will solve and to collect / to generate training samples Training mode: supervised training

Advised tool: Matlab NNTOOL


  • Select the task neural network to solve
  • Generate training samples

Real Life scenario

Looking througth real life scenario each dependency has impact for the final result

More questions, more answers More data given, the more questions answered. More training, greater results.

To test actually value of Data in Machine Learning #TODO


Input Data First Row gender 0 - male 1 - female Second Row age numeric

Target consumption 0 - very_low 1 - low 2 - normal 3 - high 4 - very_high


Neural Network Configuration

Network type: Feed-forward backprop Training function: TRAINLM Adaptation learning function: LEARNGDM Performance function: MSE

Training Results

Networkprevie.PNG NetworkTraining.PNG Performance.PNG Regression.PNG TrainingState.PNG


The results after exploring the Neural Network by increasing and decreasing given data to the network changed significantly. But each dependency had its role to play

  • Number of features used to describe the pattern:
    Amount of features, mostly had impact on <…> as a <…> where <…>
  • Number of classes:
    Amount of classes, acted as a <…> on <…> where <…>
  • Number of training samples:
    Number of samples, changed <…> where <…> as a <…>