Context-Dependent Pre-Trained Deep Neural Networks for Large-Vocabulary Speech Recognition | Synapse