CORDIS - Forschungsergebnisse der EU
CORDIS

Approximate Computing for Power and Energy Optimisation

Periodic Reporting for period 1 - APROPOS (Approximate Computing for Power and Energy Optimisation)

Berichtszeitraum: 2020-11-01 bis 2022-10-31

Following the current trend, by 2040, computers will need more electricity than the world's energy resources can generate. On the communications side, energy consumption in mobile broadband networks is comparable to data centers.
The APROPOS project proposes adaptive Approximate Computing to optimize energy-accuracy trade-offs. Luckily, in many parts of the global data acquisition, transfer, computation, and storage systems, there exists the possibility to trade off accuracy to either less power or less time consumed – or both. For example, numerous sensors are measuring noisy or inexact inputs; the algorithms processing the acquired signals can be stochastic; the applications using the data may be satisfied with an “acceptable” accuracy instead of exact and absolutely correct results; the system may be resilient against occasional errors; and a coarse classification or finding the most probable matches may be enough for a data mining system. By introducing a new dimension, accuracy, to the design optimization, the energy efficiency can be improved by 10x-50x.

The main goal is to train the spearheads of the future generation to cope with the technologies, methodologies, and tools for successfully applying Approximate Computing to power and energy saving. The training, in this first-ever ITN addressing approximate computing, is largely done by researching energy-accuracy trade-offs on the circuit, architecture, software, and system-level solutions, bringing together world-leading experts from European organizations. In addition, we will provide network-wide and local training on the substance and the complementary skills needed in industrial and academic work life.

The main APROPOS project objectives are:
• [O1] To gain research skills needed to obtain an internationally recognized PhD and to become recognized scientists in their field;
• [O2] To gain an entrepreneurial and innovation-oriented attitude through exposure to SME and spin-off partners in and around the network;
• [O3] To gain experience in applying their research both in the academic and non-academic sectors by secondments and research visits;
• [O4] To gain transferrable skills such as entrepreneurship, scientific publishing, grant applications, career paths, CV writing, project management, managing IP, technology transfer, research ethics, and communication skills;
• [O5] To provide and appreciate a working atmosphere that cherishes diversity, gender balance, multi-cultural skills and versatility to work in a global environment;
• [O6] To acquire an open approach to research, through the open publication of research papers, data and software and a “reproducible research” mindset;
• [O7] To promote excellence and prepare them for applying to prestigious European and national fellowships;
• [O8] To be aware of the pros and cons of academic vs. industrial careers and learn from the experiences of their supervisory team.
During this period, APROPOS found and recruited 15 ESRs, most of whom have completed the first year. Indeed, the first year is expected to be the introductory period when ESRs mainly focus on the initial literature review and formalizing their research directions. The APROPOS ESRs have published 8 papers: 1 journal paper and 7 conference papers (ones with DOI have already been added to the Zenodo repository; a few more will be counted as soon as the main publisher informs us about their DOI). There are also 12 submitted/accepted papers. This way, the project outputs could be considered promising toward the project's overall goals.

The overall objectives were summarised as follows:
• [O1] Supervisory teams established and maintained; All ESRs enrolled and progressing in their studies.
• [O2] All planned training events were organized with industrial partners.
• [O3] Some intersectoral secondments have already started.
• [O4] All training events were complemented with transferrable skills training.
• [O5] All the efforts were put to provide and appreciate a working atmosphere that cherishes diversity, gender balance, multi-cultural skills, and versatility to work in a global environment.
• [O6] All results are published in Open Access. All ESRs have active ORCID accounts.
• [O7] Training related to fundraising were organized.

Besides, the ESRs have executed their research work per defined Work Packages boosting the state-of-the-art in the field of Approximate Computing.
Besides the training objectives, the ESRs have developed their first algorithms, methods, and have tested them.

Some of the applied research findings are listed below (marked with the involved ESR number):

• Provided a report on an overview of the principal methodologies explored in the current state-of-the-art in ML and DL for Approximate Computing [1,2,3];
• Analysed existing fault-tolerant approximate computing techniques [2];
• Reported on available benchmarking, software, and evaluation tools (also industrial) for Approximate Computing [5-9];
• Studied the opportunities related to dynamic and distributed operation of Approximate computing (including computational offloading to the network infrastructure) [4,10-15];
• Identified the opportunities of Approximate Computing use for recourse-constrained devices [4,10-15].

Regarding impact through social media, the APROPOS Twitter channel has 179 followers currently. Facebook, ResearchGate, and LinkedIn activities fit the corresponding venues. APROPOS currently has 2 press releases (in English and Finnish). Project representatives organized 4 outreach events in schools and other universities. All essential updates are available in the Blog. The progress is monitored by the project manager and handled by the APROPOS communications committee, which includes 2 ESRs.
APROPOS project logo

Verwandte Dokumente