We will initialize various parameters that are needed by the DBN class defined earlier:
finetune_act_func = tf.nn.relu rbm_layers = [256] do_pretrain = True name = 'dbn' rbm_layers = [256] finetune_act_func ='relu' do_pretrain = True rbm_learning_rate = [0.001] rbm_num_epochs = [1] rbm_gibbs_k= [1] rbm_stddev= 0.1 rbm_gauss_visible= False momentum= 0.5 rbm_batch_size= [32] finetune_learning_rate = 0.01 finetune_num_epochs = 1 finetune_batch_size = 32 finetune_opt = 'momentum' finetune_loss_func = 'softmax_cross_entropy' finetune_dropout = 1 finetune_act_func = tf.nn.sigmoid
Once the parameters are defined, let's run the DBN network on the MNIST dataset:
srbm = dbn.DeepBeliefNetwork(
