Online learning has witnessed an increasing interest over the recent past due to its low computational requirements and relevance a broad range of streaming applications. In this brief, we focus on online regularized regression. We propose novel efficient regression algorithm, called normalized least-squares (ONLS). perform theoretical analysis by comparing total loss ONLS against gradient desc...

We address the phase retrieval problem with errors in sensing vectors. A number of recent methods for are based on least squares (LS) formulations which assume quadratic measurements. extend this approach to handle vectors by adopting total (TLS) framework that is used linear inverse problems operator errors. show how gradient descent and specific geometry can be obtain a simple efficient TLS s...