[bull-ia] Post-doctoral position in deep learning and semantics for smart factories

** Post-doctoral research topic
Deep learning and semantics for Smart Factories **

*Working environment:*

The Post-doc will work within a large and joint project between the
LITIS lab (Rouen, France) and the GREYC lab (Caen, France) supported by
the Normandy region. The successful candidate will be hosted in the
LITIS lab in Rouen.


A major characteristic of Industry 4.0 is the fusion of Internet
technologies and factories, giving birth to the so-called Smart
Factories. The tools, machines, workstations and operators are
interconnected, facilitating process traceability, adaptive and flexible
control of the production equipment and real-time response to uncertain

Cyber-Physical Systems (CPS) are central to Smart Factories and are
entitled to be part of intelligent machines, storage systems and
production facilities able to exchange information with autonomy and
intelligence. Such systems should be able to decide and trigger actions,
and control each other independently [1].

In a broad sense, CPSs are defined as the systems which offer
integration of computation, networking, and physical processes [2–5], or
in other words, as the systems where physical and software components
are deeply intertwined, each operating on different spatial and temporal
scales, exhibiting multiple and distinct behavioral modalities, and
interacting with each other in  ways that change with context [6].

Some of the defining characteristics of CPS include a) cyber capability
in every physical component, b) high-degree of automation, c) networking
at multiple scales, d) integration at multiple temporal and spatial
scales and e) reorganizing/reconfiguring dynamics [7].

Manufacturing companies are engaged in implementing these new
technologies, in order to enhance their manufacturing lines in a
intelligent way. Briefly, these companies need to analyse the flow of
data measured by the sensors in the CPSs and to understand their
meaning, locally and collectively, for intelligent decision making.

However, these companies, very often, also need to deal with the
existence of legacy systems and monolithic solutions that in the
best-case scenario would only provide limited interconnectivity by
providing pretty basic data logs in sometimes-exotic formats [1].

This is why we intend to explore the joint use of deductive and
inductive approaches in this context with the goal of providing
solutions to the companies willing to implement these Smart Factories.

The shared aim of semantic technologies and deep learning is to create
intelligent tools that simulate human skills such as reasoning,
validation and prediction. Both areas have had a significant impact on
the fields of analysis and representation of data and of knowledge.

Our hypothesis is that the joint use of semantics [8] and of deep
learning [9] will improve the quality of the interpretation of huge
amounts of data, will make sense of them and will allow to gain new
insights. Therefore, we will be interested in exploring how inductive
and deductive methods can play together to uphold data retrieval, reuse,
and integration.

On the one hand, top-down engineered ontologies and logical inferencing
[10] can play a key role in providing the vocabularies for querying
data, while deep learning and data mining can take profit of these
linked data to improve learning and the knowledge discovery. In
particular, we will be interested in investigating ontology-based
approaches to deep clustering or classification, and in exploring deep
learning Linked Data.

On the other hand, it will also be interesting to explore then how to
learn higher-level features, expressed as logical axioms from data
(coming from sensors, measurements of characteristics of the final
products, machine statistics) and then use these higher level features
for non-trivial deductive inferences (to predict eventual losses or
faults and to propose recommendations to address them).

Beyond the two points below, we will also be interested in using deep
learning for ontology learning, matching or alignment.


[1] Carlos Toro, Iñigo Barandiaran, Jorge Posada, “A Perspective on
Knowledge Based and Intelligent Systems Implementation in Industrie
4.0”, Procedia Computer Science, Volume 60, 2015, Pages 362-370, ISSN
1877-0509, http://dx.doi.org/10.1016/j.procs.2015.08.143.

[2] M. Conti et al., “Looking ahead in pervasive computing: challenges
and opportunities in the era of cyber-physical convergence,” Pervasive
and Mobile Computing, 2011.

[3] L. Sha, S. Gopalakrishnan, X. Liu, and Q. Wang, “Cyber-physical
systems: A new frontier,” Machine Learning in Cyber Trust, pp. 3–13, 2009.

[4] I. Horath and B. Gerritsen, “Cyber-physical systems: Concepts,
technologies and implementation principles,” in Tools and Methods of
Competitive Engineering Symposium (TMCE), 2012, pp. 19–36.

[5] E. Lee, “Computing needs time,” Communications of the ACM, vol.
52,no. 5, pp. 70–79, 2009.

[6] NSF, “Cyber physical systems nsf10515,”
http://www.nsf.gov/pubs/2010/nsf10515/nsf10515.htm, 2013.

[7] L. Miclea et al., “About dependability in cyber-physical systems,”
in EWDTS, 2011, pp. 17–21.

[8 C. Zanni-Merk. KREM: A Generic Knowledge-Based Framework for Problem
Solving in Engineering – Proposal and Case Studies, 7th International
Joint Conference on Knowledge Discovery, Knowledge Engineering and
Knowledge Management, Lisbonne, Portugal, pages 381-388, INSTICC (Eds.),
Science and Technology Publications, Lda, November 2015,

[9]  Ian Goodfellow and Yoshua Bengio and Aaron Courville, Deep
Learning, MIT Press, 2016
[10] Mariano Fernandez-Lopez and Oscar Corcho. 2010. Ontological
Engineering: With Examples from the Areas of Knowledge Management,
E-Commerce and the Semantic Web. First Edition. Springer Publishing
Company, Incorporated.

*The candidate*

Candidates will be required to have a PhD in computer science with
relevant skills in semantic technologies, deep learning and data mining
(not all of which require expertise, but experience in some of them will
be appreciated).


The employment contract will be for 12 months, starting in February /
March  2018, with a gross monthly salary of around €2500.  The
post-doctoral fellow will work in the LITIS offices at the Madrillet
campus in Saint-Etienne du Rouvray (Seine-Maritime).

*Application instructions*

The application consists of a motivation letter, CV (with a detailed
list of publications and links to e.g implementations on Github), names
and contact details of two references and any useful document.
Applications should be submitted before *December 10th* via electronic
mail to *the contacts below.*


Cecilia Zanni-Merk (cecilia.zanni-merk@insa-rouen.fr)
Bruno Cremilleux (bruno.cremilleux@unicaen.fr)