Auxiliary Network: Scalable and Agile Online Learning for Dynamic System with Inconsistently Available Inputs

Streaming classification methods assume the number of input features is fixed and always received. But in many real-world scenarios, some are reliable while others unreliable or inconsistent. We propose a novel online deep learning-based model called Auxiliary Network (Aux-Net), which scalable agile can handle any inputs at each time instance. The Aux-Net based on hedging algorithm gradient descent. It employs varying depth an setting using single pass learning. foundational work towards neural network for dynamic complex environment dealing ad hoc inconsistent inputs. efficacy shown Italy Power Demand dataset.