Showing papers by "Darko Stefanovic published in 2010"

Training a molecular automaton to play a game

Abstract: Research at the interface between chemistry and cybernetics has led to reports of 'programmable molecules', but what does it mean to say 'we programmed a set of solution-phase molecules to do X'? A survey of recently implemented solution-phase circuitry indicates that this statement could be replaced with 'we pre-mixed a set of molecules to do X and functional subsets of X'. These hard-wired mixtures are then exposed to a set of molecular inputs, which can be interpreted as being keyed to human moves in a game, or as assertions of logical propositions. In nucleic acids-based systems, stemming from DNA computation, these inputs can be seen as generic oligonucleotides. Here, we report using reconfigurable nucleic acid catalyst-based units to build a multipurpose reprogrammable molecular automaton that goes beyond single-purpose 'hard-wired' molecular automata. The automaton covers all possible responses to two consecutive sets of four inputs (such as four first and four second moves for a generic set of trivial two-player two-move games). This is a model system for more general molecular field programmable gate array (FPGA)-like devices that can be programmed by example, which means that the operator need not have any knowledge of molecular computing methods.

Shape analysis with reference set relations

Abstract: Tracking subset relations between the contents containers on the heap is fundamental to modeling the semantics of many common programing idioms such as applying a function to a subset of objects and maintaining multiple views of the same set of objects. We introduce a relation, must reference sets, which subsumes the concept of must-aliasing and enables existing shape analysis techniques to efficiently and accurately model many types of containment properties without the use of explicit quantification or specialized logics for containers/sets. We extend an existing shape analysis to model the concept of reference sets. Reference sets allow the analysis to efficiently track a number of important relations (must-=, and must-⊆) between objects that are the targets of sets of references (variables or pointers). We show that shape analysis augmented with reference set information is able to precisely model sharing for a range of data structures in real programs that cannot be expressed using simple must-alias information. In contrast to more expressive proposals based on logic languages (e.g., extensions of first-order predicate logic with transitive closure or the use of a decision procedure for sets), reference sets can be efficiently tracked in a shape analyzer.