March 8, 2018

FUZZ-IEEE Tutorials


Times 1 2 3 4 5 6 7 8 9 10 11
8.00 – 10.00 Session 1 CEC1_01 CEC2-01 CEC3_01 CEC4_01 FUZZ5_01 (Part 1) FUZZ6_01 (Part 1) HYB7_01 Part 1 IJCNN1_01 (Part 1) IJCNN1_02 (Part 1) IJCNN3_02 (Part 1) IJCNN4_01 (Part 1)
10.00-10.15 Coffee
10:15 – 12.15 Session 2 CEC1-02 CEC2-02 CEC3_02 CEC4_02 FUZZ5_01 (Part 2) FUZZ6_01 (Part 2) HYB7_01 (part 2) IJCNN1_01 (Part 2) IJCNN1_02 (Part 2) IJCNN3_02 (Part 2) IJCNN4_01 (Part 2)
12.15- 13:00 Lunch
13:00 – 15:00 Session 3 CEC1-03 CEC2_03 (Part 1) CEC3_03 CEC4_03 (Part 1) FUZZ5_02 FUZZ6_02 (Part 1) HYB7_02 IJCNN1_03 IJCNN1_02 (Part 3) IJCNN3_01 IJCNN4_02
15:00 – 15.15 Coffee
15:15 – 17:15 Session 4 CEC1-04 CEC2_03 (Part 2) CEC3_04 CEC4_03 (Part 2) FUZZ5_03 FUZZ6_02 (Part 2) HYB7_03 IJCNN1_04 IJCNN1_02 (Part 4) IJCNN3_03 IJCNN4_03 (Part 1)
17:15 – 19:15 Session 5 CEC1-05 CEC2-04 CEC3_05 CEC4_05 FUZZ5_04 IJCNN3_04 HYB7_04 IJCNN1_05 IJCNN2_01 IJCNN3_05 IJCNN4_03 (Part 2)


FUZZ5_01 Type-2 Fuzzy Sets and Systems
FUZZ5_02 Fuzzy Systems in Medicine and Healthcare
FUZZ5_03 Support Fuzzy Machines: From Kernels on Fuzzy Sets to Machine Learning Applications
FUZZ5_04 Fuzzy Sets, Computer Science and (Fuzzy) Algorithms
FUZZ6_01 Fuzzy Logic and Machine Learning
FUZZ6_02 Uncertainty Modeling in Learning from Big Data

Title: Type-2 Fuzzy Sets and Systems (FUZZ5_01)

Organized by Christian Wagner, Jon Garibaldi, and Josie McCullochDescription:

General type-2 fuzzy sets and systems are paradigms which enable fine-grained capturing, modelling and reasoning with uncertain information. While recent years have seen increasing numbers of applications from control to intelligent agents and environmental management, the perceived complexity of general type-2 fuzzy sets and
systems still makes their adoption a daunting and not time-effective proposition to the majority of researchers.
This tutorial is designed to give researchers a practical introduction to general type-2 fuzzy sets and systems. Over three hours, the modular tutorial will address three main aspects of using and working with general type-2 fuzzy sets and systems:

1. Introduction to General Type-2 Fuzzy Sets and Systems
The first component of the tutorial will provide attendees with a concise and practice-led overview of general type-2 fuzzy sets and systems, reviewing the motivation behind their definition, their structure in relation to type-1 and interval type-2 fuzzy sets and systems, as well as a set of recent applications.

2. Designing General Type-2 Fuzzy Sets and Systems
In the second part of the tutorial, two distinct aspects will be discussed. First, attendees will be given a practical introduction to designing their own general type-2 fuzzy system. Using the online browser-based toolkit JuzzyOnline, participants will be guided in the design of a general type-2 fuzzy system, relating their own design to the design of type- 1 fuzzy systems at each stage.
Second, the design of general type-2 fuzzy sets will be discussed through a presentation of a key set of recently introduced processes to create general type-2 fuzzy sets from data.

3. Coding General Type-2 Fuzzy Sets and Systems
The final part of the tutorial will focus on the programmatic implementation and use of general type-2 fuzzy sets and systems. Currently available software tools and toolkit for general type-2 fuzzy sets and system applications will be briefly reviewed, highlighting usage areas from inference to the computation of measures such as similarity and distance. Finally, interested participants will be supported in the development of a simple general type-2 fuzzy system based on the freely available Juzzy, Python and/or R based general type-2 APIs.

Timing: The overall time of the tutorial will be three hours, with an approximately even
split over all three tutorial components listed above

Pre-requisites: Basic knowledge of type-1 fuzzy sets and systems is the only pre-
requisite for attendees to be able to benefit from this tutorial.

Title: Fuzzy Systems in Medicine and Healthcare (FUZZ5_02)

Organized by Uzay Kaymak and J. M. SousaDescription:

Improving access and delivery of health care is a very actual topic in many countries. In order to meet the challenges posed by growing medical costs, aging population and limited resources, new technologies and new methods are being developed. Fuzzy systems play an important role in this context due to their ability to model nonlinear
system behavior, to deal with non‐probabilistic uncertainty and to describe model behavior in natural language, making communication with decision makers easier. This tutorial discusses how fuzzy set theory can be used to learn interpretable and transparent models that can be used to support and improve decision making in the healthcare. Cases will be shown regarding (clinical) decision support by using fuzzy set theory, process analysis with fuzzy sets, linguistic summarization of medical data and fuzzy systems‐based support of healthy aging and wellbeing. All of this material will be preceded by a general discussion of the challenges for data‐driven improvement of
healthcare processes.
The tutorial will be given in the form of a lecture and consists of two parts. In the first part, a general introduction will be provided regarding the challenges for data‐driven approaches to improving healthcare processes. Afterwards, fuzzy set theory‐based solutions that deal with these challenges will be presented in a broad overview. The overview includes fuzzy modeling techniques, interpretability‐focused fuzzy models, advanced fuzzy models combining fuzziness with probabilistic uncertainty, modeling of linguistic information by using fuzzy sets and flexible information fusion by using fuzzy set theory. In the second part, cases from the industry, in which applications of fuzzy set theory for decision support have played a central role, will be presented. The cases will cover predictive fuzzy modeling for medical decision support, linguistic summarization of medical data and healthcare protocol analysis by using fuzzy sets.

The tutorial will complete with an outlook of promising research directions for fuzzy sets in the healthcare domain. The tutorial will take 90 minutes and will be presented by two lecturers in an interactive way.

Intended audience

The tutorial is of interest for researchers, practitioners and graduate‐level students (PhD or advanced Master students) working in the fields of soft computing and computational intelligence, who are interested in fuzzy set applications in medicine and healthcare. The attendants are expected to have a basic knowledge of fuzzy sets and
fuzzy systems.

Title: Support Fuzzy Machines: From Kernels on Fuzzy Sets to Machine Learning Applications (FUZZ5_03)

Organized by Jorge Guevara, Roberto Hirata Jr, Stéphane CanuDescription:

Most of the time machine learning algorithms consider real-valued attributes as the primal object of interest. However, there are many applications where the data contains set-valued attributes, for example, astronomers usually work with clusters of celestial objects, meteorologists need to perform data science on interval-valued data. Further, imprecise, uncertainty and noise measurements can be modeled by set-valued attributes. Nowadays, the support distribution machines (also called support measure machines) is a widely used technique for solving machine learning tasks on that kind of data. Such approach basically estimates a kernel function between a pair of set-valued attributes, i.e., between the empirical distributions of the points in each set. Thus, the resulting kernel matrix can be used in a kernel machine to solve a particular task. However, there are some common situations where it is more parsimonious to use fuzzy sets for modeling set-valued attributes than empirical distributions, for example, interval-valued datasets, datasets corrupted by noise, datasets with imprecise measurements or datasets with qualitative data. Fuzzy sets can be defined in a simpler way for those examples by using either prior expert knowledge or data-driven approaches.
The aim of this tutorial is to introduce a new class of kernel machines: the support fuzzy machines for solving practical machine learning tasks on datasets containing fuzzy-set valued attributes. Those learning machines learn a functional representation of the fuzzy set-valued data in a Reproducing Kernel Hilbert Space of functions, and in a similar way that support vector machines rely on some support vectors for characterizing the solution, the support fuzzy machines rely on some fuzzy sets. The core idea of this machines is to use positive definite kernels functions on fuzzy sets for defining a gram matrix that can be used for classification, anomaly detection or regression tasks. Such kernels define a similarity measure of fuzzy sets as inner products on Hilbert spaces by using the kernel trick.

Intended audience

Machine learning and data science practitioners, and researchers interested in unconventional data analytics.

Title: Fuzzy Sets, Computer Science and (Fuzzy) Algorithms (FUZZ5_04)

Organized by Rudolf SeisingDescription:

50 years ago, in 1968, Lotfi A. Zadeh published the paper “Fuzzy Algorithms” in the journal Information and Control. In these years, he was very active in the debate on the education in Computer science that emerged in that time as a new scientific discipline from the field of electrical engineering. As chairman of the Department of Electrical Engineering he was responsible for bringing about the “Berkeley solution”, which ultimately led to the formation of a department for computer science in the College of Letters and Science and a programme in computer science in the College of Engineering within his Department.

In “Computer Science as a Discipline”, which appeared also in 1968, he brought into focus that Computer Science (CS) “cuts across the boundaries of many established fields” and that the parts of CS differ from one another “in degrees of emphasis”. Here he linked his reflection on CS education with fuzzy sets: “Specifically, let us regard
computer science as a name for a fuzzy set of subjects and attempt to concretize its meaning by associating with various subjects their respective degrees of containment (ranging from 0 to 1) in the fuzzy set of computer science. For example, a subject such as «programming languages» which plays a central role in computer science will have a degree of containment equal to unity. On the other hand, a peripheral subject such as «mathematical logic» will have a degree of containment of, say, 0.6.” In a “Containment Table for Computer Science” he arranged the most relevant “subjects in question and their degrees of containment in computer science”. For example, he gave “Theory of Algorithms” the degree of containment of 0.9 in computer science.

Algorithms depend upon precision. An algorithm must be completely unambiguous and error-free in order to result in a solution. The path to a solution amounts to a series of commands, which must be executed in succession. Algorithms formulated mathematically or in a programming language are based on set theory. Each constant and variable is precisely defined, every function and procedure has a definition set and a value set. Each command builds upon them. Successfully running a series of commands requires that each result (output) of the execution of a command lies in the definition range of the following command, that it is, in other words, an element of the
input set for the series. Not even the smallest inaccuracies may occur when defining these coordinated definition and value ranges.

However, in “Fuzzy Algorithms” Zadeh fuzzified the commands. Later in an interview, he told me: “I began to see that in real life situations people think certain things. They thought like algorithms but not precisely defined algorithms.” In the paper he wrote. “All people function according to fuzzy algorithms in their daily life […] they use recipes for cooking, consult the instruction manual to fix a TV, follow prescriptions to treat illnesses or heed the appropriate guidance to park a car. Even though activities like this are not normally called algorithms: “For our point of view, however, they may be regarded as very crude forms of fuzzy algorithms”.
Algorithms are very old concepts in mathematics. The may be oldest algorithms that we know is the Euclidean one to compute the greatest common divisor to two numbers.

This appeared in his Book VII of his Elements in the 3rd century before Christ. The “Algorithm” is named after the Persian scholar in the House of Wisdom in Baghdad, Muḥammad ibn Mūsā al-Khwārizmī (82o CE), who wrote a treatise in the Arabic language, which was translated into Latin in the 12th century under the title “Algoritmi de numero Indorum”. This title means "Algoritmi on the numbers of the Indians", where "Algoritmi" was the translator's Latinization of Al-Khwarizmi's name.
Today, algorithms are important in Machine Learning. This history started with the aim to analyze learning behavior with statistical work on biological classification, verbal learning and concept learning as “experiments in induction” in the fields of Statistics and Artificial Intelligence. Algorithms for matching and prediction in data sets established in the 1960s in the new field “Data Analysis”.

Intended audience

The intended audience will be scientists in CI with interests in the historical and philosophical fundamentals of Computer Science, Artificial and Computational Intelligence.

Title: Fuzzy Logic and Machine Learning (FUZZ6_01)

Organized by Hamid R. TizhooshDescription:

In this tutorial, I will talk about the state of the art of fuzzy algorithms in machine learning. In the first part, we will review fuzzy algorithms when they are applied on typical machine-learning tasks such as search, classification, approximation and learning. In the second part, the relationship between fuzzy methods and other machine-learning approaches are reviewed whereas hybrid schemes will be in foreground. In both parts, relevant literature will be reviewed. Matlab/Python examples will be executed/displayed to demonstrate the effect of major methods for relevant applications such as data mining, signal processing, image analysis, and big data.
Links to online resources will be included in the material which also contains the source codes and the presentation slides.

The tutorial runs for 3 hours and will cover:
1. Brief History of Fuzzy Logic
2. Brief History of Machine Learning
3. Fuzzy Algorithms for Search, Classification, Approximation and Learning
4. Fuzzy Algorithms and Other Machine-Learning Methods
5. Applications: Data Mining, signal processing, image analysis, and big data
6. Matlab/Python examples

At the end of each section/topic, multiple choice questions will be asked via Kahoot online platform such that the audience can participate. This interactive platform with anonymous statistics has received a lot of positive feedback for effective lectures and tutorials.

Title: Uncertainty Modeling in Learning from Big Data (FUZZ6_02)

Organized by Xizhao Wang


The tutorial will contain the following content.

1. Introduction to fuzziness and uncertainty. Uncertainty is a natural phenomenon in machine learning, which can be embedded in the entire process of data preprocessing, learning and reasoning. For example, the training samples
are usually imprecise, Incomplete or noisy, the classification boundaries of samples may be fuzzy, and the knowledge used for learning the target concept may be rough. Uncertainty can be used for selecting extended attributes and informative samples in decision tree inductive learning and active learning respectively. If the uncertainty can be effectively modeled and handled during the process of processing and implementation, machine learning algorithms
will be more flexible and more efficient. This part will focus on the uncertainty definition and relationships/differences among different uncertainties.

2. Big data and its 5V features. Big data refers to datasets that are so large that conventional database management and data analysis tools are insufficient to work with them. Big data, which was called massive data, has become a bigger-­‐than-­‐ever problem with the quick developments of data collection and storage technologies. Nowadays, many complex processes can generate big data, for example, there are a greater number of Earth Observing Satellites than ever before, collecting many terabytes of data per day. This part will concentrate on the big data challenges and the current handling strategies.

3. Modeling uncertainty in big data learning. This is to show that the representation, measure, and handling of uncertainty have a significant impact on the performance of learning algorithms. Usually the modeling/handling of
uncertainty is associated with the feature-­‐type and volume of dada. Recent research shows that making clear the change/adaptation of uncertainty with feature-­‐type and volume of data is a very difficult issue. This difficulty is significantly increasing if we deal with the big data. This part will talk mainly about the how to model the uncertainty and how the learning performance is improved through uncertainty handling.

4. The new challenges that uncertainty brings to big data learning. Big data has two more features, i.e., multimodality and changed-­‐uncertainty. The former means that the types of data can be very complex while the latter indicates that the modeling and measure of uncertainty for big data is significantly different from
that for normal sized data. This part will mainly address the critical issue, i.e., big data destroys the fundamental assumption of statistical learning, i.e., the assumption of samples independently identically distributed, and therefore, some new learning theory and methodology need to be re-­‐built.

5. Concluding remarks. This part will give an overview on learning with uncertainty from big data, summarizing the results acquired in recent years’ study on big data leaning problems. Some remarks on the concept of fuzzy-­‐learning which considers the learning as two categories according to the problem nature are finally presented.