Neural Network Programming with TensorFlow

Pretraining the DBN

In this section, we look at how a DBN is pretrained:

class RBM(UnsupervisedModel):...  def pretrain(self, train_set, validation_set=None):    """Perform Unsupervised pretraining of the DBN."""    self.do_pretrain = True    def set_params_func(rbmmachine, rbmgraph):    params = rbmmachine.get_parameters(graph=rbmgraph)     self.encoding_w_.append(params['W'])     self.encoding_b_.append(params['bh_'])    return SupervisedModel.pretrain_procedure(     self, self.rbms, self.rbm_graphs, set_params_func=set_params_func,     train_set=train_set, validation_set=validation_set)

This in turn calls SupervisedModel.pretrain_procedure(..), which takes the following parameters:

• layer_objs: A list of model objects (autoencoders or RBMs)
• layer_graphs: A list of model ...