Compiler development for concurrent procedural programming languages: This is the traditional approach in which the syntax of a procedural code is translated into an optimized assembly code that can be executed on a machine.
Synthesis based on behavioral VHDL descriptions (digital hardware description languages):
Hardware/software co-design using VHDL processes:
Analog and mixed-signal synthesis using VHDL-AMS descriptions: Analog and mixed-signal (AMS) systems are traditionally described at the structural level, such as the circuit or system structure. However, high-level descriptions of such systems used mainly for modeling and simulation expresses AMS systems using sets of ordinary differential equations (ODEs), possibly controlled by digital signals, like in multi-mode descriptions. For example, state-space equations are popular in filter design. In our work, we assumed that AMS ODEs are expressed using VHDL-AMS language. There is a semantic gap between the descriptions as sets of differential equations used in modeling and simulation and circuit and system structures used in design. Our work devised a set of constraints for VHDL-AMS programs and a set of rules that correctly maps the ODE descriptions in the constrained VHDL-AMS language into system structures based on analog and mixed-signal circuits, like OpAmps, comparators, integrators, and FSM controllers. The parameterized structures include design variables, which are then used for trade-off exploration and optimization.
Trade-off optimization:
Topology synthesis for high-level analog and mixed-signal descriptions:
Understanding the semantics of design creativity:
Understanding the behavior and dynamic of teams:
Computational methods applied to understanding artwork: Our goal is to study new knowledge structures and methods to understand abstraction in modern painting. While visual features have been essential in classical painting, modern painting mainly relies on semantic abstractions and ambiguities to communicate a message in which the observer becomes an active participant to the dialog with the artist (not just a passive actor, like for classical art). Our work focuses on understanding the capabilities and limitations of current Deep Neural Networks in analyzing artwork. Moreover, we are pursuing a stochastic-symbolic approach that attempts to parallel the human cognitive activities in understanding art, including memory organization, attention, concept associations, and reasoning.