Automatic scene recognition for low-resource devices using evolving classifiers

Andréu, Javier; Baruah, Rashmi Dutta; Angelov, Plamen

doi:10.1109/fuzzy.2011.6007720

Cited by 4 publications

(4 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is also possible that more complex, composite concepts are also formed such as "market place", "war", "hospital", "office", etc. The process of forming concepts based on clouds of data with similar properties is not considered in detail in this paper, but this can be automated as described in [15]. This process can be done with a cognitive feedback (the dotted line in Figure 1) or in some cases sub-consciously in a fully unsupervised manner when we associate some scene or music etc.…”

Section: The Concept Of the Proposed Methodsmentioning

confidence: 99%

Cybernetics of the Mind: Learning Individual's Perceptions Autonomously

Angelov

Iglesias

et al. 2017

IEEE Syst. Man Cybern. Mag.

Self Cite

View full text Add to dashboard Cite

In this paper, we describe an approach to computational modelling and autonomous learning of the perception of sensory inputs by individuals. A hierarchical process of summarization of raw data with heterogeneous nature is proposed. At the lower layer of the hierarchy, the raw data autonomously forms semantically meaningful concepts.Instead of clustering based on the visual or audio similarity, the concepts are being formed at the second layer of the hierarchy based on the observed physiological variables (PVs) like heartbeat, skin conductance, etc. and mapped to the emotional state of the individual. Wearable sensors were used in the experiments. Methodologically, we used the recently introduced Empirical Data Analytics (EDA) computational framework and the data partitioning method within EDA to form the data clouds (cluster-like formations with no pre-defined shape) autonomously and used AnYa type of IF-THEN fuzzy rule-based models to describe the mapping of the observable PVs to the emotional states. Multi-modal typicality distributions, which have properties like pdf and indicate the empirical likelihood distributions of the data, can be further derived without making prior restrictive assumptions traditionally required in statistical approaches. The quality of the classifier is evaluated by a confusion matrix and classification rate/precision. The experimental results are very encouraging: for this extremely complicated problem, we got 74.58% correct classification results without any pre-training.

show abstract

Section: The Concept Of the Proposed Methodsmentioning

confidence: 99%

Cybernetics of the Mind: Learning Individual's Perceptions Autonomously

Angelov

Iglesias

et al. 2017

IEEE Syst. Man Cybern. Mag.

Self Cite

View full text Add to dashboard Cite

show abstract

“…One of the problems with conventional (off-line pre-trained) classifiers is that they do not generalize well from training images to new scenes. Therefore, in [18] we propose to use simpl_eClass that updates its structure (rule-base) and adapts itself to each unseen image. Further, its suitability to low-resource devices can be attributed to its computational efficiency.…”

Section: On-line Scene Classification Using Fuzzy Rule-based Clasmentioning

confidence: 99%

“…Calculate the output of each rule (consequent part of (11)). Calculate the overall output using (17) and assign a class label to the sample using (18). Else calculate membership degree using (4), the rule firing strength using (3), and the weight of each rule (λ) using (17).…”

Section: Classificationmentioning

confidence: 99%

“…To determine the class label for a given sample the overall output ( ŷ ) is estimated by weighted sum of outputs of all the rules ( 17), and the index corresponding to the highest value element of ŷ is designated as the class label of the sample (18). (17) where ŷ is the overall estimated output, i τ is the firing strength of the i th rule and calculated using (3) and (4).…”

Section: B Simpl_eclass1mentioning

confidence: 99%

See 1 more Smart Citation

Simpl_eClass: Simplified potential-free evolving fuzzy rule-based classifiers

Baruah

Angelov

Andréu

2011

2011 IEEE International Conference on Systems, Man, and Cybernetics

Self Cite

View full text Add to dashboard Cite

This paper presents the sequel of evolving fuzzy rule-based classifier eClass, called here as simplified evolving classifier, simpl_eClass. Similarly to eClass, simpl_eClass comprises of two different classifiers, namely zero and first order (simpl_eClass0 and simpl_eClass1). The two classifiers differ from each other in terms of the consequent part of the fuzzy rules, and the classification strategy used. The design of simpl_eClass is based on the density increment principle introduced recently in so called simpl_eTS+ approach. The rule learning in simpl_eClass does not involve computation of potential values that allows it to attain computationally much less expensive model update phase compared to eClass. As compared to other FRB classifiers, it retains all the advantages of eClass, such as being on-line and evolving, having zero and first order. In comparison with other non-fuzzy classifiers it has the advantage of interpretability and transparency (especially zero order type). The goals of this paper are to demonstrate the applicability of simpl_eTS+ to classification task, and to empirically show that the simplification of eClass to simpl_eClass by using potential-free approach does not compromise the accuracy of the classifiers. In order to attain the goals, the classifiers are tested by performing several experiments using benchmark data sets. The simpl_eClass1 classifier is also applied to the real-life problem of on-line scene categorization for low-resource devices benefiting from its low computational cost. The results obtained from the experiments endorse that simpl_eClass achieves the accuracy of eClass while simplifying rule learning process.

show abstract