|
An overlooked need of High-Throughput-Screening (HTS) technologies is the ability to rapidly classify incoming events in real-time. Proper real-time classification of events frequently requires more than simple threshold, or even multiple-threshold, gating logic. Usually for HTS the desired event is ultra-rare and requires more complex decision making than simple gating on raw parameters. Otherwise, the number of false positives can lead to unacceptable results. Similarly, the number of false negatives can lead to unacceptable loss of ultra-rare events. Storage of all data for off-line analysis can lead to requirements for huge off-line storage capabilities with the need to still replay all the data. A better approach is a real-time, multi-queuing, multistage decision-making process that keeps up with the incoming data stream and stores only data of probable interest. A multistage classification system permits faster simpler decisions in early stages leaving more time for more sophisticated decision-making at later stages when the remaining event data stream is much slower. Complex linear and nonlinear mathematical and multi variate statistical classifier systems can be implemented in real-time using linked lookup tables operating at memory speeds. Overall event classifications rate in excess of 100,000 events per second can be achieved for HTS of billions of total events.
|
