FARIA Project: Adaptive Artificial Intelligence (AI) in Distributed Networks

Expertise Sought

  • Evolutionary theory expert: understanding the fundamental laws of adaptation
  • Mathematician: understanding of spatio-temporality and ergodicity
  • Computer scientists/data scientist: deep comprehension of AI adaptation to new conditions
  • Computer scientists/data scientist creating new AI learning models and embrace transfer and federated learning to enhance adaptation and optimization

Research Abstract

The aim of this project is to study “Nature of Adaptive Algorithms” based on the law of physics (Greek: φύσις, fysis, nature) to better understand human-nature interaction and an adoptive nature of process, where the Gaussian proposal distribution is updated along the process using the full information cumulated so far. An adaptive Metropolis algorithm has the correct ergodic properties. In mathematics ergodicity expresses the idea that a point of a moving system, either a dynamic system or stochastic process will eventually visit all parts of the space that the system moves in, in a uniform and random sense. A stronger concept than ergodicity is that of mixing, which aims to mathematically describe the common sense notions of mixing, such as mixing drinks or mixing cooking ingredients.

There is a wide spectrum of artificial intelligence algorithms which belong to different classes of approaches (conservative/modern), including fuzzy logic, expert systems, evolutionary strategies and genetic programming. (Russell & Norvig 1995). However, there are also differences between different paradigms in natural sciences and these differences mean that there are different paradigmatic underpinnings for modeling artificial intelligence algorithms that determine the key assumptions of information, ontology of organizing and ontology of modeling spacetime. One of the fastest-growing sub-fields of computational intelligence is evolutionary computation. In this field, there are many algorithms to solve optimization problems. These algorithms mimic and simulate the Darwinian theory of survival of the fittest in nature (Reynolds 1987). Such algorithms mostly mimic biological evolution in nature (Mirjalili et. al. 2019). In other words, they use genealogical information of population adaptation to the environment. This means that evolutionary algorithms (EA) search only a part of search space using heuristics information, but instead utilize adaptation of population.

Paradigm for modeling theory of evolution based on genealogical data produces evolutionary algorithms that are equipped with several random (stochastic) components, which select and combine solutions in each population. They model the theory of evolution based on genealogical data. This makes them unreliable in finding similar solutions in each run as opposed to deterministic algorithms. Deterministic algorithms (e.g. brute force search) find the same solution in every run, but suffer from slower speed and local solution stagnation when applied to large-scale problems (Mirjalili et. al. 2019). The project demonstrates potential innovations that will substantially advance in science at a global level.

Our theoretical goals are to a) WP 1. understand the fundamental laws of adaptation, b) WP 2. understanding of spatiotemporality and ergodicity c) WP 3. deep comprehension of adaptation to new conditions and d) WP 4. creating new AI learning models and embrace transfer and federated learning to enhance adaptation and optimization.

The main research question is, how AI algorithms adapt to new conditions? Data from nature will be collected by project pilots using off-the-shelf nature storage and transmission equipment. As a result of the project, we will have a better understanding of the interaction between man and nature and the effects of man on nature and its sustainable development.

The project is a new era of informatics and extremely significant for AI researchers (de Bézenac E, Pajot, A. & Gallinari, P. (2019). Modeling physical phenomena based on laws of physics is a paradigm shift towards nature-based AI. The project elaborates adaptive algorithms based on the laws of physics. Equality, transparency, trust, and security are identified as ethical factors that are essential, when developing self-organizing adaptive AI systems.

Keywords

Artificial intelligence algorithms, Adaptation, AI paradigms, Assumption of information, Modeling paradigms, Modeling spacetime

References

  1. de Bézenac E, Pajot, A. & Gallinari, P. (2019). Deep Learning for Physical Processes: Incorporating Prior Scientific Knowledge. Journal of Statistics.
  2. Mirjalili , S., Song Dong , J. Ali Safa, S. & Hossam, F. (2019). Genetic Algorithm: Theory, Literature Review, and Application in Image Reconstruction, Nature-Inspired Optimizers pp 69-85
  3. Russell, P. & Norvig, P. (1995). Artificial intelligence – a modern approach. Eglewood Cliffs, NJ.
  4. Reynolds, C. W. (1987, August). Flocks, herds and schools: A distributed behavioral model. In ACM SIGGRAPH computer graphics (Vol. 21, No. 4, pp. 25-34). ACM.

Biosketch of Principal Investigator

1. Personal details and the date of the CV

  • Laakkonen Mika-Petri
  • https://orcid.org/0000-0003-3230-7958
  • 12th January 2022 2.

2. Degrees

  • PhD (Information technology), 17th, January 2007, University of Lapland, Finland, contact details +35816341341
  • Licentiate of Education (Cognitive Science). 28th January 2002, University of Lapland
  • Master Degree of Education (Educational Sciences), 30th, May 1995, University of Lapland

3. Other education and expertise

  • Matriculation exam, 30th May 1989, Upper level secondary school, Oulainen, Finland, Matriculation exam, High School, 30th May 1988, Lexington, USA 

4. Language skills

  • Finnish, English (C2), Swedish (C2), French (C2), German (B2)

5. Current employment

  • January 1st 2020 – December 31st 2021, Research Director, University of Lapland, Rovaniemi, Finland

6. Previous work experience (recent 2018 - 2021)

  • September 1st, 2018 – December 31st, 2019, Visiting professor, Université de la Sorbonne, Pierre et Marie Curie Université, Paris, France
  • March 1st, 2019 – April 30th, 2019, Visiting associate professor Versailles Saint-Quentin-en- Yvelines University, Paris, France
  • July 1st – August 31st, 2018, Research Director, University of Lapland, Rovaniemi, Finland
  • January 1st to June 31st, 2018, Associate professor of Applied Information Technology, University of Lapland, Rovaniemi, Finland

7. Personal funding and grants (recent 2018 – 2021)

  • PI (Principal Investigator), March 22nd, 2019 – January 31st, 2022, French Ministry of Higher Education, Research and Innovation, Finnish Academy of Sciences Finnish Science Association, Maupertuis 2020 OPEN: France-Finland bilateral funding program for scientific cooperation, Fund € 3000
  • PI (Principal Investigator), July 1st 2021– December 31st 2021, Provincial Voluntary Development Funding (AKKE), Digital Aviation Fund € 77 809
  • PI (Principal Investigator), September 1st 2018 – December 31st, 2019 Design in Smart Mobility Business Services, Business Finland Ltd., Fund € 438 000
  • Member of the project preparation working group, September 1st 2018 – October 31st 2021, Arctic Smartness RDI - Excellence (ASR), ELY centre, Fund € 655 656
  • Project developer and researcher, September 1st 2018 – December 31st 2021, ERDF, Future Bio-Arctic Design - Organic Smart Textile (F.BAD), Fund € 496 581
  • PI (Principal Investigator), InterregEurope, September 1st 2018 – December 31st 2021, RegionArts: "Enhancing SME growth by the integration of Artist in ICT projects Fund € 1 776 120 • Project developer, September 1st 2018 – December 31st 2021, Prosoc (Arctic Societies Vocational Research Group), Fund € 450 000 €

8. Research output

  • Total number of publications (first or second writer): 33 (22 peer-reviewed). For the most important publications for the research project, see the included appendix. For a list of all research output, see my profile at https://orcid.org/0000-0003-3230-7958

9. Research supervision and leadership experience

  • Supervisor of over 150 theses
  • Two doctoral students started (2021)

10. Teaching merits

  • University Teacher since 1998 in artificial intelligence (deep learning), computer sciences, HCI (human-computer interaction) and future research methodology.
  • Keynote speaker in numerous international conferences at Berkeley University, Fudan University, University of Helsinki, Oxford University, Technical University of Arkhangelsk, University of Florence and Sorbonne University

11. Other key academic merits, scientific and societal impacts (2018 – 2022)

  • Currently (2022-) I am member of FCAI (Finnish Center for Artificial Intelligence) community, which brings together of top talents in academia, industry and the public sector to solve real- life problems using both exiting and novel AI. FCAI is one of the Academy of Finland’s Finnish flagships, hubs of top level-research and impact.
  • 2021 – Member of Finnish-American Research and innovation Acceletors FARIA strategic network. The main goal of the network is to carry out collaborative projects aimed at cutting- edge research and innovation together with American partners in artificial intelligence
  • 2019 - European Commission's Directorate-General for CONNECT (Communications Networks, Content and Technology), independent expert, the Horizon 2020 flagship Artificial Intelligence project, funded by the European Commission, looking particularly at predictive algorithms related to health technologies.
  • Editor-in Chief: Topic of Special Issue is: “Artificial Intelligence (AI) in contemporary society” (jufo 2)
  • 2019 Society for Futures Studies (2019) referee, Malaska Award, (Graduation Award),
  • 2018 National Audit Group of the Finnish Universities (2018), evaluator of the quality of research at the University of the Arts and IAMCR (International Association for Media and Communication Research) conference reviewer

List of ten most important publications (related to the topic of the project)

  1. Laakkonen, M., Kivivirta, V. & Mazari, A. (2022). Four modelling paradigms for artificial intelligence algorithms, Philosophy of Science (evaluation phase/jufo 3)
  2. Laakkonen, M. & Kivivirta, V. (2022). Elevators as media object: Manipulating information in time, New Media & Society: SAGE Journals (jufo 3)
  3. Laakkonen, M. 2021. Artificial Intelligence (AI): Hidden Rules of our Society, Artificial Intelligence in our contemporary society. Special Issue of Journal of Administrative Studies" (jufo 2)
  4. Laakkonen, M. (2021). Information, ethics, and the digital society (ed.) SoPhi (Publisher) (jufo 1)
  5. Kivivirta, V. Laakkonen, M-P., Myllykoski, J. & Rantakari A. (2021) Predictive algorithms in MNC: Whiteheadian process ontological view of prediction, 37th EGOS Colloquium 2021, University Amsterdam, The Netherlands (peer-reviewed)
  6. Kivivirta, V & Laakkonen, M-P. (2020). Thinking organizing with Whitehead in the age of predictive algorithms, Organizations, Artifacts, and Practices (OAP) Workshop 2020, University of California, Berkeley (peer-reviewed)
  7. Kotilainen, S., Laakkonen, M-P. & Okkonen, J. (2018). Children's trust in artificial intelligent applications: Needs for media literacy 3.0", to the Media Education Research Section for IAMCR 2018 - Oregon. June 20 – 24, 2018. (peer-reviewed)
  8. Laakkonen, M. (2018). Cognitive stages in rational thinking - towards human technology. Electronic Imaging & the Visual Arts, EVA (2018) Florence, conference, May 9 – 10th, 2018. (peer-reviewed)
  9. Keinänen, J & Laakkonen, M. (2012). Virtual Tutor: Designing a Teaching Model for a Programmed Guide, AACE Association for the Advancement of Computing in Education, November 6 – 8, 2012 (jufo 2)
  10. Laakkonen, M. 2003. The Future Relationship Between Virtual Reality and Human body, The Good, The Bad and The Irrelevant Conference, User and the Future of information and communication technologies, 2003 September 3rd – 5th, Helsinki, Finland (peer- reviewed/nominated for the best paper in conference)