Session Program

 

  • 12 July 2017
  • 05:00PM - 07:00PM
  • Room: Auditorium
  • Chairs: Yusuke Nojima and Sushmita Mitra
Abstract - The generalized Kolmogorov goodness-of-fit test for interval-valued data is proposed. Two versions of the test are considered -- each corresponding to a different view on the outcomes of the experiment, i.e. either the epistemic or ontic one. It is shown that each view on interval-valued data yield different approaches to data analysis and statistical inference.
Abstract - Activity recognition using smartphone provides a ubiquitous and unobtrusive way for people to realize health monitor and ambient assisted living. Since human activities has characteristics of high complexity and diversity, the accurate identification of activity greatly depends on the appropriate features extracted from limited smartphone signals and the efficiency of pattern recognition approaches. An unsupervised classification scheme based on the wavelet packet transform (WPT) and the half-cosine fuzzy clustering (HFC) is proposed in this paper for the automatic feature extraction and recognition of human activities on smartphone. The wavelet packet coefficient features combined with statistic features describe the sensor signals comprehensively. The novel half-cosine initialization eliminates the sensitivity of the fuzzy clustering to initial center distribution. Experiment results of public datasets reveal that the WP-based hybrid features are more suitable for human activity recognition than statistic features. The performance of proposed half- cosine fuzzy clustering is superior than those of FCM, HAC and K-means for the activity recognition on smartphone.
Abstract - Domain adaptation is a transfer learning approach that has been widely studied in the last decade. However, existing works still have two limitations: 1) the feature spaces of the domains are homogeneous, and 2) the target domain has at least a few labeled instances. Both limitations significantly restrict the domain adaptation approach when knowledge is transferred across domains, especially in the current era of big data. To address both issues, this paper proposes a novel fuzzy-based heterogeneous unsupervised domain adaptation approach. This approach maps the feature spaces of the source and target domains onto the same latent space constructed by fuzzy features. In the new feature space, the label spaces of two domains are maintained to reduce the probability of negative transfer occurring. The proposed approach delivers superior performance over current benchmarks, and the heterogeneous unsupervised domain adaptation (HeUDA) method provides a promising means of giving a learning system the associative ability to judge unknown things using related knowledge.
Abstract - Non-singleton fuzzification is used to model uncertain (e.g. noisy) inputs within fuzzy logic systems. In the standard approach, assuming the fuzzification type is known, the observed [noisy] input is usually considered to be the core of the input fuzzy set, usually being the centre of its membership function. This paper proposes a new fuzzification method (not type), in which the core of an input fuzzy set is not necessarily located at the observed input, rather it is dynamically adjusted based on statistical methods. Using the weighted moving average, a few past samples are aggregated to roughly estimate where the input fuzzy set should be located. While the added complexity is not huge, applying this method to the well-known Mackey-Glass and Lorenz time-series prediction problems, show significant error reduction when the input is corrupted by different noise levels.
Abstract - In a relational database context, fuzzy quantified queries have been long recognized for their ability to express different types of imprecise and flexible information needs. In this paper, we introduce the notion of fuzzy quantified statements in a (fuzzy) RDF database context. We show how these statements can be defined and implemented in FURQL, which is a fuzzy extension of the SPARQL query language that we previously proposed. Then, we present some experimental results that show the feasibility of this approach.
Abstract - In this paper, we propose to present a novel numerical information fusion method at the feature level. This method is based on the possibility theory using the Beta function and the type1 fuzzy theory. In this method, we proceed to estimate the possibility distribution of the numerical features by using the Beta function. So, we constitute the Beta-possbilistic knowledge base. Then, this Beta-possibilistic knowledge base will be the input of the fuzzy numerical features fusion method. We have evaluated this numerical information method on 15 benchmark databases. Also, we have compared this method with our numerical information fusion method based on the Beta-possibilistic reasoning and with four another numerical information fusion methods in order to determine the best one.