xtractis® Communications

Pangborn

The 8^th edition of the Pangborn Sensory Science Symposium will take place at Florence (Italy) from July 26 to July 30, 2009. “The Pangborn meeting has continued to grow and is recognized as the most important scientific symposium for the disciplines of sensory and consumer science. The main task of the organising team is to ensure for delegates a high quality scientific programme. The opening reception, plenary and concurrent sessions as well as poster presentations will be planned to deliver an exciting scientific programme representing the best of sensory research today.”

 

intellitech's contributions:

• Identification and prediction of robust non linear relationships between recipes, instrumental measurements and sensory evaluation of cake texture and flavour with xtractis® fuzzy models. Application to the discovery of optimal recipes. (8^th Pangborn, Florence, Italy, 2009).
  Download the poster.
• Why Cross Validation Methods are particularly relevant to Robust Modeling? Interests in Sensory Science (7^th Pangborn, Minneapolis, MN, USA, 2007).
  Download the poster.
• xtractis® fuzzy predictive models for virtual testing and prototyping. Sensory evaluation of cheese spreadability (6^th Pangborn, Harrogate, North Yorkshire, UK, 2005).
  Download the poster.

Sensometrics

Sensometrics is the scientific area that applies mathematical and statistical methods to problems from sensory and consumer science. The 9th Sensometrics Conference took place at St. Catharines (Canada) from July 20 to July 23, 2008.

intellitech's contributions:

• Is your accurate model actually robust? Regulation and validation methods by xtractis® (8^th Sensometrics, Ås, NORWAY, 2006).
  Download the oral presentation
• Fuzzy System to explain and predict the influence of storage conditions on the sensory profile of a salad dressing (8^th Sensometrics, Ås, NORWAY, 2006).
  Download the oral presentation.
• Variable selection by Support Vector Machines for fuzzy non linear modelling (8^th Sensometrics, Ås, NORWAY, 2006).
  Download the poster.
• Contribution of fuzzy mathematics to the predictive modelling of virtual testers/experts. Applications to the subjective evaluation of Champagne Wines (7^th Sensometrics, Davis, CA, USA, 2004)
  Download the oral presentation.
• Contribution of fuzzy mathematics to the predictive modelling of virtual testers/experts. Applications to the subjective evaluation of tomatoes (7^th Sensometrics, Davis, CA, USA, 2004).
  Download the poster.

AgroStat

The 10th edition of the European Symposium on Statistical Methods for the Food Industry took place at Louvain-la-Neuve (Belgium) from January 23 to January 25, 2008: "These days offer an ideal opportunity to statisticians, engineers, and users of statistical methods in food industries to meet and exchange experiences around stimulating themes like sensometrics, chemometrics, risk analysis and process control".

intellitech's contributions:

• Contribution of xtractis® methodology to the automatic extraction of robust fuzzy models. Application to the prediction of consumer liking and sensory evaluation, and to the optimization of product formulation (10^th AgroStat, Louvain-la-Neuve, BELGIUM, 2008).
  Download the paper
  Download the oral presentation

White Papers

Fuzzy logic and fuzzy inference systems: introduction and properties

Fuzzy mathematics, also known generically as fuzzy theory, include a number of theories which are generalizations or extensions of their classic equivalents: thus fuzzy sets theory is an extension of set theory, fuzzy logic is an extension of binary logic, fuzzy quantities theory is an extension of number and interval theory, possibility theory extends probability theory and more generally fuzzy measure theory extends measure theory. All these theories offer formally rigorous concepts, techniques and methods for collecting, representing and analysing fuzzy data. The specificity of fuzzy data is that it is imprecise, uncertain and subjective and these three main fuzzy characteristics often co-exist. A whole range of other synonyms fall under this notion of fuzziness, such as knowledge which is poorly specified, poorly described, imperfect, vague, qualitative, linguistic, partial, incomplete or approximate.

This conceptual leap makes mathematics more useful and greatly increases potential applications, because these notions provide a closer approximation to the real world. Fuzzy mathematics also offers the possibility to select and apply ad hoc operators depending on both the problem to be resolved and the personality of the decision-maker (optimistic, pessimistic or compromising behaviour…).

Because fuzzy logic is nuanced and gradual, it more closely approaches human logic allowing the measure of possibility to become an accurate replacement for the measure of probability, when the available information is sparse and/or of poor quality. This is particularly the case in sensory assessment panels, where only a small number of individuals act as sensors, who are very often imprecise. Formally, fuzzy theory defines an interface between qualitative/symbolic and quantitative/numeric concepts. From a practical point of view, it offers a natural approach to the resolution of multidimensional and complex problems characterized by a strong interaction of the components involved, where Human is both a sensor and a decision-maker/actuator.

• Introduction to fuzzy logic.
  Download the white paper
• Introduction à la logique floue.
  Télécharger le document

xtractis® non linear modelling approach

Fuzzy inference systems allow to easily and intuitively model any decision making process, whether it represents a physical measurement, a mathematical computation or a human evaluation. The decision making process is modelled as a deterministic relationship between inputs (available knowledge about the situation) and an output (the decision to be taken), implicitly expressed by linguistic rules.

The classical fuzzy modelling method derived from Artificial Intelligence uses available knowledge about the decision making process. The fuzzy rules are built thanks to a linguistic expression of this knowledge. However, in many situations (subjective evaluation, high complexity of the decision process), it is not possible to a priori define the linguistic rules that explain the process.

In those cases, the xtractis® approach proposes to automatically extract the linguistic rules explaining the process through automatic learning. This learning is performed on a database of several decision cases corresponding to different situations. This approach is similar to neural network training with a learning base, with the advantages of the fuzzy model paradigm over neural networks.

Databases used by xtractis® are divided into three different types: objective data (O) - recipes, analysis results or physico-chemical measurements, demographic or financial data … -, subjective data (S) - consumer liking - and subjective objectivised data (SO) given by human experts - sensory panelists - that represents the most objective evaluation of subjective attributes.

The classical issues arising in any learning process (overfitting, noisy data, limited amount of learning points) require implementing methods to analyze the quality of the learning base, supervise the learning process, and estimate the generalization capacity of the model to new unknown points.

• xtractis® non linear modelling approach.
  Download the white paper
• Approche xtractis® en modélisation non linéaire.
  Télécharger le document

Academic publications

Yusr Amamou's doctoral thesis: Trace analysis and perceptive exploration strategies modelling via a minimalist coupling sensory-motor device

The objective of our research is to analyze the subject’s interactions traces using the Tactos device, which comprises a stylus, a graphic tablet and a tactile stimulators connected to a computer equipped with an application software. The principle of this interface is to activate the stimulators when the cursor receptor field controlled by stylus crosses the virtual form that the subjects must perceive. The goal of our analysis is to define descriptors which categorize and differentiate the trajectories produced by the subjects, and to use these descriptors in order to model the various spontaneously elaborated perceptive strategies. To operate this analysis, we had resort to several mathematical tool descriptions, in particular the Fourier, and to various modelling methods.
The results, by the means of the models tested and validated, make it possible to differentiate indeed a certain number of strategies employed by the human subjects. These models could be integrated into the interface; thereafter the users can have a return on the nature of their explorations, which would contribute to the training and the appropriation of the interface.

• Thèse de Doctorat de Yusr Amamou : Analyse de traces et modélisation des stratégies d'exploration perceptive via un dispositif de couplage sensori-moteur minimaliste (UTC, Compiègne, FRANCE, 2008)
  Télécharger le document

Florian Kuhn's doctoral thesis: start-up high-tech performance prediction , quantitative approach from the Fuzzy Mathematics theory

La création et le développement d'entreprises start-up de haute technologie sont de plus en plus reconnues comme un facteur significatif d'innovation et de transfert des technologies. Cependant, les start-ups rencontrent une multitude de problèmes particuliers au cours de leur développement (comme, par exemple, une forte incertitude des facteurs clés de succès) pour lesquels une évaluation globale de l’entreprise et une analyse de ses forces et de ses faiblesses dans les différents domaines apportent un soutien très utile. Toutefois, les méthodes d’évaluation existantes ne tiennent pas compte de la particularité de la situation de l’entrepreneur, ou ne sont pas destinées à l’entrepreneur. Basé sur une analyse des facteurs clés de succès et la logique floue, le présent article propose une méthode pour combler ces lacunes. Ainsi est-il développé le concept d’un outil qui permet une vue globale et complète de l’entreprise, qui prend en compte les incertitudes et interdépendances des facteurs pertinents, et qui donne un retour instructif à l’entrepreneur.

• Thèse de doctorat de Florian Kuhn : Prédiction de performance des start-ups de haute technologie : approche quantitative à partir de la Théorie des Mathématiques Floues (UTC, Compiègne, FRANCE, 2006)
  Accessible par la bibliothèque de l'UTC