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

Rohit Agarwal, Krishna Agarwal, Alexander Horsch, Dilip K. Prasad

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