NNproject

• This is a neural network implementation in MATLAB.
• It is designed as a term project in Intro to Artificial Neural Networks class.
• This network is currently applied to a specific task of predicting age by one's normalized fMRI data.
• But it is applicable to other tasks as well.

Key features

1. Define the model

• Task type
  • Classification
  • Regression
• Cost function
  • MSE (Mean Square Error)
  • Cross-entropy

2. Cross-validation

• K-fold cross-validation
  • (nfolds) without inner CV for parameter optimization
• K-fold nested cross validation
  • (nfolds) by (vfolds)

3. Not to overfit

• L1 & L2 regularization
  • and its corresponding lambda
  • or a candidate list of lambdas (if nested cv applied)
• Dropout
  • and dropout rate
• Resampling methods
  • None
  • Bayesian
  • GaussianNoise