My name is Sepide Saeedi and I have joined the APROPOS project on November 2021 at Politecnico di Torino, Italy. I received a BSc degree in Computer Engineering from the Iran University of Science and Technology, Tehran, Iran, in 2016 and an MSc degree in Computer Architecture from the University of Tehran, Iran, in 2019. My research interests include approximate computing, embedded system design, high performance computing and neural networks.
The title of my project is: “Fast and Accurate Prediction of the Impact of Approximate Operators on a Complex Computation.” The focus of my research is on providing an Interval Arithmetic-based model to predict the impact of Approximation on Neural Network computations.
Overall Research Plan is to design and build a complete Approximate Computing -enabled RISC-V Simulator. In order to achieve this goal, we Implement the RISC-V ISA based on the gem5 simulator. Then the customization happens by introducing Approximate Components/Instructions for Hardware and Software techniques (Adders, Multipliers, Conditional Increment (for loop perforation-based code), etc.) Next step is to guarantee a power model (for keeping the energy requirements under investigation) and to study and implement techniques to dynamically control the approximation at runtime, based on prediction.
