Seminar DSCY: Hrabáček R.: Evolutionary Algorithms for Approximate Computing (cvičná obhajoba tezí disertační práce)

Computers or computer based systems play a crucial role in people's everyday lives, embedded systems can be found almost everywhere. More and more applications area able to tolerate inaccurate or incorrect computations to a certain extent due to imperfections of human senses, statistical data processing, noisy input data etc. At the same time, power efficiency is becoming increasingly important property of computing platforms, especially because of limited power supply capacity of embedded devices. Approximate computing, an emerging paradigm, takes advantage of relaxed functionality requirements to make computer systems more efficient in terms of energy consumption, computing speed or complexity.
Error resilient applications can achieve significant savings while still serving their purpose with the same or a slightly degraded quality.
The complexity of computer systems is permanently growing and thus, automated design tools have to deal with more and more complex problems specified on higher level of abstraction than before. The same holds true for approximate computing. Even though new methods are emerging, there is a lack of methods for approximate computing offering a numerous set of trade-off solutions. Evolutionary algorithms (EAs) have been confirmed to bring innovative solutions to complex problems. Recently, complex digital circuits have been optimized by means of EAs while the scalability of the methods has been improved substantially. The goal of this report is to analyze existing approximate computing techniques and evolutionary computation methods, identify open problems that need to be solved, isolate the topic of the Ph.D. thesis and set up a plan how to solve the problems.