BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20131121T001000Z
DTEND:20131121T003000Z
LOCATION:Booth 3947
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The Automaton Processor (AP) directly implements a non-deterministic finite automaton substrate, but provides extra features that make it capable of much more. In fact, it is Turing-complete. This presentation illustrates some additional capabilities of the AP. Aluru will present ongoing work at Georgia Tech for solving bioinformatics applications. Numerous applications in bioinformatics deal with sequence data and finding approximate matches. This talk will focus on the (l,d) motif finding problem, the problem of finding a motif of length l that is present in at least q of the n given sequences, with each occurrence different from the motif in at most d positions. This problem is NP-hard and the largest instance reported to date is (26,11). Using the resources available in a single Automata Processor board, we have been able to design a parallel algorithm to solve many larger instances, including (39,18) and (40,17). Skadron will explain work at the University of Virginia showing how to achieve the fundamental concepts of universal computing - local memory, data/instruction stream and sharing units for different operation on AP. We have implemented two Turing-Machine-equivalent Cellular Automata, Elementary Cellular Automata (ECA) and Game of Life (2D CA), on AP and obtained significant speedup, showing that the AP is Turing-complete. We also show the AP implementations of three other applications: run-length encoding, graph connectivity, and shortest path.
SUMMARY:
PRIORITY:3
END:VEVENT
END:VCALENDAR