Research, development and innovation
Research & Innovation, a driving force for innovation
Research and development are some of the essential tools for innovation, provided you invest in innovation. Innovation is essential for ensuring the growth of a business.
Innovation is a major element in the development strategy. In rapidly evolving environments, it is necessary to adapt and stand out in order to remain competitive and attractive. Having a culture of innovation helps attract and retain talent.
Two R & D projects
Octopeek promotes innovation by supporting R & D through projects led by PhD students. This helps reinforce the company’s technological leadership.
Octopeek currently supports two PhD students: Julien Hay and Moncef Mouffok. They are under the guidance of:
- Mahmoud Zakaria, Deep Learning Expert, PhD in Telecom and IT
- Ouassim Ait Elhara, PhD in IT (specializing in artificial intelligence)
- Mahdi Hannoun, Big Data expert, PhD in IT (specializing in distributed computing)
Learning by incremental transfer from heterogeneous data
(Thesis of Moncef Mouffok performed in partnership with Paris-Saclay University in the IBISC laboratory)
To train a model through supervised learning, it is necessary to have a large amount of labeled training data. However, the
labeling of data requires manual preparation which is time consuming.
The objective of Moncef Mouffok’s thesis is to implement expertise that will allow methods and techniques to be acquired that avoid and replace all the manual work in labeling data.
One of these methods is to exploit and transfer knowledge learned from models on large labeled databases to a specific database.
This new method could save time in classification, regression or ranking projects (recommendation system) in areas where there is very little or no labeled data.
At Octopeek, the first use of this method will make it possible to predict whether someone is leaving or retired, for example, as part of a data enrichment project.
Enrichment of profiles and recommendation of articles based on semantic analysis
(Thesis of Julien Hay performed in partnership with CentraleSupelec in the LRI laboratory))
Internet users share and comment on a significant amount of content on social networks. This dynamic data provides new clues that can be exploited by the emerging methods of automatic natural language processing (NLP).
Julien Hay puts forward models that extract points of interest. To achieve this, these models exploit what is written and read by the users. This dynamic data, carrying many semantic clues, allow both the prediction of preferences and the offering of content of interest to users, with ever increasing precision. The goal is to go beyond existing models that are based only on:
- Static data (occupation, age…)
- Collaborative treatments (recommendation through collaborative filtering)
In an environment of increasing infobesity, these models must allow users time savings in:
- Their choice of reading (recommendation of news articles)
- Their choice of products (service, insurance…).