Abstract: Is news “new?” In a sense, yes. Being first to take advantage of a piece of news has always been essential to traders and investors. Today, however, the overwhelming volume and ever-shrinking time available to react to it profitably has posed mounting pressure on trading desks, portfolio managers and analysts to cope with these challenges. Fortunately, technology is making it feasible to handle more news in less time in ways that were barely conceivable a few years ago. Newly available technologies and techniques are giving traders and investors innovative ways to use automation (such as complex event processing engines) to allow them to incorporate news into their strategies. These enable them to exploit market opportunities and inefficiencies driven by news events and it helps them reduce their exposure to event risks.

Abstract: We describe and analyze an algorithmic framework for online classification where each online trial consists of multiple prediction tasks that are tied together. We tackle the problem of updating the online predictor by defining a projection problem in which each prediction task corresponds to a single linear constraint. These constraints are tied together through a single slack parameter. We then introduce a general method for approximately solving the problem by projecting simultaneously and independently on each constraint which corresponds to a prediction sub-problem, and then averaging the individual solutions. We show that this approach constitutes a feasible, albeit not necessarily optimal, solution of the original projection problem. We derive concrete simultaneous projection schemes and analyze them in the mistake bound model. We demonstrate the power of the proposed algorithm in experiments with synthetic data and with multiclass text categorization tasks.