Backpropagation

 - Standard algorithm
     encoder_nb.sce      : 4-3-4 encoder on ANN without biases
     encoder.sce         : 4-2-4 tight encoder on ANN with biases

 - Momentum
     encoder_m_nb.sce    : 4-3-4 encoder on ANN without biases
       compare with encoder_nb.sce
     encoder_m.sce       : 4-2-4 tight encoder on ANN with biases
       compare with encoder.sce
     enc848_m_nb.sce     : 8-4-8 encoder on ANN without biases
     enc838_m.sce        : 8-3-8 encoder on ANN with biases

 - SuperSAB
     enc858_ssab_nb.sce  : 8-5-8 encoder on ANN without biases
     enc848_ssab.sce     : 8-4-8 encoder on ANN with biases
     (Note that the more tight encoders
      are very difficult to train with this algorithm)

 - Conjugate Gradients
     encoder_cc.sce      : 4-2-4 tight encoder on ANN with biases
                           uses a mixed standard/conjugate gradients method