Portal:Complex Systems Digital Campus/IKM

From Wikiversity
Jump to navigation Jump to search



CSDCUniTwinLogo.jpg
.
.
the repository for Open Questions, Challenges and Ressources of the
.




iKM
Integrated Knowledge Map


Challenge[edit]

Description[edit]

The rapid evolution of economy and society in the age of globalization requires a rapid evolution of competences and thus necessitates life-long education, training, and re-training. At the same time education must become more personalized, as the competences needed to master an increasingly sophisticated technological world continually diversify. Any researcher, expert, or simply any learner during his/her life-long activities needs periodically reconfigure his/her own personalized Knowledge Map (pKM) by exploring part of the whole Integrated Knowledge Map (IKM).

It is conjectured that human concepts are obeying to a 'preferential attachment law': a concept that is already very well connected is attracting less connected concepts. Such conjecture can be tested on Wikipedia (with the drawing tool used in ref[1]). Furthermore, the multilingual pages for each concept tends to reproduce the same attachments to the other concepts. Thus the preferential attachment network between concepts tends to overpass the cultural frontiers (testable hypothesis).

Keywords[edit]

Knowledge Map, personalized Knowledge Map, preferential attachment, preferential attachment network, knowledge neutrality.

Board[edit]

  • Paul Bourgine (chair | email)
  • Jorge Louça (co-chair)
  • Pierre Collet (co-chair)

Bibliography[edit]

References: [1] [2]

Sub-challenges[edit]

Encyclopedia & knowledge repositories[edit]

Description[edit]

The following corpus will be used and regularly updated:

- the multi-linguist Wikipedia as an encyclopedia,

- the whole set of abstracts of the Web of Science for covering all the scientific fields,

- the whole set of theoretical articles of ArXiv,

- the largest possible set of available video: Youtube, Dailymotion, TED conferences, conference replays, MOOCs, slides, etc ..).

Milestone: 3 months

Members[edit]

  • Jorge Louça (chair)
  • David Chavalarias (chair)
  • Michaël Kohlhase (co-chair)
  • Julien Baudry (co-chair)

The Integrated Knowledge Map and its dynamics[edit]

Description[edit]

The law of preferential attachment is de facto a 'gravity law', i.e. the underlying geometry is a Newtonian one or, equivalently, an Einsteinian one, i.e. an hyperbolic one [2]. In the AI point of view, instead to postulate the existence of hidden state (like with Hidden Markov Chain), it is an hypothesis about the hidden metric space behind the attachment network of concepts, e.g. the attachment network in Wikipedia. Reference [2] is proposing optimal statistics method for building the map of hyperbolic networks, here the Integrated Knowledge Map (IKM). The IKM is public. Furthermore:

- the representation of any hyperbolic map can be done from the central concepts to more and more peripheric layers: indeed, an hyperbolic geometry can develop from the center an infinite series of layers in a finite disk. A great representation will be similar to Prezi. Another very useful representation is simpler an horizontal Fancy Tree like the Mac Finder. It is even expected that the concept network is scale free: in such case, the layer representation is like an 'onion'.

- all the references attached to a concept can be available, enlarging those from Wikipedia by using all the corpus of the previous sub-challenge above

- in an hyperbolic space, the geodesics are giving the shortest path between two concepts. They are attracted by the center of the map. For this reason, the triangle between two concepts are thin.

- a great advantage of optimal statistics [2] is their 'accumulator property': the introduction of new concepts or of new links between concepts is incremental. And the dynamics of the concepts will be observable, especially their movement toward the center (thus more opportunities if learning them) or toward the periphery). Such movement is significative for the social interest attributed to each concept in the IKM

Such sub-challenge is both a tentative to integrate human knowledge but also to give to anyone the opportunity to learn what wanted to be learned at any time. It is the same aim as Wikipedia but it can be more efficient as described in the next sub-challenge.

Members[edit]

  • Paul Bourgine (chair)
  • Jorge Louça (co-chair)
  • Pierre Collet (co-chair)

Milestone: 6 months Ref [2]

Personalized Knowledge Map (pKM) and its dynamics[edit]

Description[edit]

The personalized Knowledge Map is just a subgraph of the IKM above. It is representing the state of knowledge of each individual.

Let suppose that someone want to study a new concept, e.g. a concept that have a social interest because its movement toward the center of IKM. This concept is at some hyperbolic distance of the concepts she already know. The study can start from the closest one (the advice can be also more open: starting from concepts already known with increasing distance). The geodesic between the starting known and the unknown target is with unknown concept. As seen above, the geodesic is attracted toward the center of the IKM and go up from layers to layers until a 'top concept'; then it goes down from layers to layers toward the target. The learner can choose in which order to learn the intermediate concepts of the geodesic, e.g. according her preference to learn bottom up or top down.

When someone is studying a new concept and choosing an educational resource to study it, it is grading the educational resource, e.g. with the scale: 'very difficult', 'difficult', 'satisfying', 'simple', 'very simple'; or, simpler, 'success', 'failure'. With the Majority Judgment, a complete order of the difficulty of the educational resource for the same concept is obtained, publicly. Symmetrically, a complete order of the strength of the learners is obtained anonymously. (symmetry has to be checked ??). Furthermore, for the same difficulty and the same concept, a resource can be judged on a scale 'very attractive to poorly attractive (again a preferential judgment). Thus advices can be provided to the learner taking into account her strength and the difficulty of the educational resource as well as its attractivity: that gives to the learner her quickest way to learn the new concept.

Members[edit]

  • Pierre Collet (chair)
  • Jorge Louça, (co-chair)
  • Jeffrey Johnson, (co-chair)

milestones: Majority Judgment

Majority Judgment

Personalized Knowledge Map and new curriculum[edit]

Description[edit]

As said in the introduction, a new curriculum can be necessary for a learner if she need new competences toward a new job in a rapidly changing world.

A new curriculum can be defined as a sub-graph with its top concept in the IKM and with a limited number of layers for avoiding too precise concepts that have to be acquired with further experience.

In case of a MOOC, the series of course can follow the sub-concepts of the top-concept.

If the learner preference is to follow its own learning trajectory, starting from the concepts it is knowing, the learning trajectory is similar to the previous sub-challenge. Learning first the top concept and its sub-concepts if unknown is recommended.

At the end of learning the new curriculum, it is highly recommended to participate to an exam session.

Members[edit]

  • Jorge Louça (chair)
  • Pierre Collet, (co-chair)
  • Jeffrey Johnson, (co-chair)

Final examen of a curriculum[edit]

Description[edit]

Exam session are happening when there is sufficiently candidates because it is proposed that the checking will be provided by the candidates themselves after some preparatory exam session. Some preliminary experiments show that the final grading is very similar to a grading by professors. Such experiments will continue.

Any final exam, whatever its organizational mode, is on the responsibility of one or more Universities. It plays the role of 'Proof of Authority'.

Members[edit]

  • Jeffrey Johnson, (co-chair)
  • Jorge Louça (chair)
  • Pierre Collet, (co-chair)

The knowledge map of the baby[edit]

Description[edit]

The learning plasticity of the baby is maximal. The baby can learn music in the pre-birth stage, the language of signs at six months, reading at 12 months, several human languages in the first three years, playing with numbers, the names of plants and animals and manifesting a lot of curiosity for everything.

Observing the progresses of each baby is crucial for its parents. It is especially important to detect as quick as possible some problems in the learning stages, e.g. autism. The early stimulation of the baby is in any case a must.

Members[edit]

  • Salma Mesmoudi, (chair)
  • Pierre Collet, (co-chair)

The knowledge map 3-6 years[edit]

Description[edit]

The learning plasticity at this age remains maximal. All the baby domains remains important. It can be added 'artistic' and 'scientific' activities. For scientific activities, little experiences as proposed by the French Association 'Main à la pâte' can be a must: after the experience, there is a discussion between the children on its result; if there is a divergence on the conclusions, a debate on the divergences is proposed.

Observing the progresses of each baby is crucial for its parents. It is especially important to detect as quick as possible some problems in the learning stages, e.g. autism. The early stimulation of the baby is in any case a must.

Public education at this age is certainly the best mean to avoid large further differences in the education of young people.

Members[edit]

  • Salma Mesmoudi, (chair)
  • Pierre Collet, (co-chair)

see European project Pollen

Data and/or tools to be shared[edit]

Return to the Wikiversity Portal of the Complex Systems Digital Campus

Return to the Complex Systems Digital Campus main website

  1. Deciphering the global organization of clustering in real complex networks: Pol Colomer-de-Simón, M. Ángeles Serrano, Mariano G. Beiró, J. Ignacio Alvarez-Hamelin & Marián Boguñá, Scientific Reports volume 3, Article number: 2517 (2013),DOI: 10.1038/srep02517 https://www.nature.com/articles/srep02517
  2. Krioukov, D., Papadopoulos, F., Kitsak, M., Vahdat, A. & Bogun˜a´,M. Hyperbolicgeometry of complex networks. Phys Rev E 82, 036106 (2010).