pydbm.synapse.recurrenttemporalgraph package¶
Subpackages¶
Submodules¶
pydbm.synapse.recurrenttemporalgraph.lstm_graph module¶
-
class
pydbm.synapse.recurrenttemporalgraph.lstm_graph.
LSTMGraph
¶ Bases:
pydbm.synapse.recurrent_temporal_graph.RecurrentTemporalGraph
Long short term memory(LSTM) networks based on Complete Bipartite Graph.
The shallower layer is to the deeper layer what the visible layer is to the hidden layer.
In relation to do transfer learning, this object is-a Synapse which can be delegated to LSTMModel.
-
create_node
¶ Set links of nodes to the graphs.
Override.
Parameters: - shallower_neuron_count – The number of neurons in shallower layer.
- deeper_neuron_count – The number of neurons in deeper layer.
- shallower_activating_function – The activation function in shallower layer.
- deeper_activating_function – The activation function in deeper layer.
- weights_arr – The weights of links. If this array is not empty, ParamsInitializer.sample_f will not be called and weights_arr will be refered as initial weights.
- scale – Scale of parameters which will be ParamsInitializer.
- params_initializer – is-a ParamsInitializer.
- params_dict – dict of parameters other than size to be input to function ParamsInitializer.sample_f.
-
create_rnn_cells
¶ Create RNN cells for a LSTMModel.
Parameters: - input_neuron_count – The number of units in input layer.
- hidden_neuron_count – The number of units in hidden layer.
- output_neuron_count – The number of units in output layer.
- scale – Scale of parameters which will be ParamsInitializer.
- params_initializer – is-a ParamsInitializer.
- params_dict – dict of parameters other than size to be input to function ParamsInitializer.sample_f.
getter
getter
-
diff_rnn_visible_bias_arr
¶ getter
-
forget_gate_activating_function
¶ getter
getter
getter
-
get_diff_rnn_visible_bias_arr
¶ getter
-
get_forget_gate_activating_function
¶ getter
-
get_hat_weights_arr
¶ getter
getter
getter
getter
getter
-
get_input_gate_activating_function
¶ getter
-
get_lstm_bias_arr
¶ getter
-
get_observed_activating_function
¶ getter
-
get_output_activating_function
¶ getter
-
get_output_bias_arr
¶ getter
-
get_output_gate_activating_function
¶ getter
getter
getter
getter
getter
-
get_v_hat_weights_arr
¶ getter
-
get_visible_bias_arr_list
¶ getter
-
get_weights_forget_cec_arr
¶ getter
-
get_weights_input_cec_arr
¶ getter
getter
-
get_weights_lstm_observed_arr
¶ getter
-
get_weights_output_arr
¶ getter
-
get_weights_output_cec_arr
¶ getter
-
hat_weights_arr
¶ getter
getter
getter
getter
getter
-
input_gate_activating_function
¶ getter
-
lstm_bias_arr
¶ getter
-
observed_activating_function
¶ getter
-
output_activating_function
¶ getter
-
output_bias_arr
¶ getter
-
output_gate_activating_function
¶ getter
getter
getter
getter
getter
setter
setter
-
set_diff_rnn_visible_bias_arr
¶ setter
-
set_forget_gate_activating_function
¶ setter
-
set_hat_weights_arr
¶ setter
setter
setter
setter
setter
-
set_input_gate_activating_function
¶ setter
-
set_lstm_bias_arr
¶ setter
-
set_observed_activating_function
¶ setter
-
set_output_activating_function
¶ setter
-
set_output_bias_arr
¶ setter
-
set_output_gate_activating_function
¶ setter
setter
setter
setter
setter
-
set_v_hat_weights_arr
¶ setter
-
set_visible_bias_arr_list
¶ setter
-
set_weights_forget_cec_arr
¶ setter
-
set_weights_input_cec_arr
¶ setter
setter
-
set_weights_lstm_observed_arr
¶ setter
-
set_weights_output_arr
¶ setter
-
set_weights_output_cec_arr
¶ setter
-
v_hat_weights_arr
¶ getter
-
visible_bias_arr_list
¶ getter
-
weights_forget_cec_arr
¶ getter
-
weights_input_cec_arr
¶ getter
getter
-
weights_lstm_observed_arr
¶ getter
-
weights_output_arr
¶ getter
-
weights_output_cec_arr
¶ getter
-
pydbm.synapse.recurrenttemporalgraph.rnn_graph module¶
-
class
pydbm.synapse.recurrenttemporalgraph.rnn_graph.
RNNGraph
¶ Bases:
pydbm.synapse.recurrent_temporal_graph.RecurrentTemporalGraph
Recurrent Neural Network Restricted Boltzmann Machines (RNN-RBM) based on Complete Bipartite Graph.
The shallower layer is to the deeper layer what the visible layer is to the hidden layer.
-
create_node
¶ Set links of nodes to the graphs.
Override.
Parameters: - shallower_neuron_count – The number of neurons in shallower layer.
- deeper_neuron_count – The number of neurons in deeper layer.
- shallower_activating_function – The activation function in shallower layer.
- deeper_activating_function – The activation function in deeper layer.
- weights_arr – The weights of links. If this array is not empty, ParamsInitializer.sample_f will not be called and weights_arr will be refered as initial weights.
- scale – Scale of parameters which will be ParamsInitializer.
- params_initializer – is-a ParamsInitializer.
- params_dict – dict of parameters other than size to be input to function ParamsInitializer.sample_f.
getter
-
diff_visible_bias_arr_list
¶ getter
getter
-
get_diff_visible_bias_arr_list
¶ getter
-
get_hat_weights_arr
¶ getter
getter
getter
-
get_v_hat_weights_arr
¶ getter
-
hat_weights_arr
¶ getter
getter
getter
setter
-
set_diff_visible_bias_arr_list
¶ setter
-
set_hat_weights_arr
¶ setter
setter
setter
-
set_v_hat_weights_arr
¶ setter
-
v_hat_weights_arr
¶ getter
-