Aquatic macroinvertebrates were used to validate the classifications used. We pooled the Surbers samples obtained in each stream to use in statistical analysis. At the laboratory, individuals were sorted, counted and identified to taxonomic level of genus. Diptera and Lepidoptera were excluded from analyses because they were identified mostly to family level. The analyses were performed taking into account importance of the taxonomic level genus, family and data type presence-absence, abundance.
We used a distance-based Multivariate Analysis of Variance db-MANOVA to assess if macroinvertebrate communities differed among stream classes of the classifications adopted in this study. The significance values were obtained by We also evaluated whether macroinvertebrate abundance and richness differed among stream classes of stream classifications using a permutation test with We also evaluated whether environmental variables differed between classifications through a db-MANOVA using Euclidean distance for the classifications 2 to 5 except for the classification based on environment variables.
The variables were normalized to have the same measurement scale. The value of is obtained by taking the average of the mean similarities within each group W i , where i is any group within the classification Figure 2. CS values close to 1 indicate a strong classification Hawkins et al.
We used the mean similarity dendrogram approach Van Sickle, to represent the values observed for and the similarity values for each stream class W i. If the classifications are strong, the branches of dendrograms, the W i values, are relatively long compared to values. The MRPP test the differences between two or more groups defined a priori.
We use The mean similarity dendrograms were obtained from a plot with the result of the meandist function for vegan package in R environment R Development Core Team, The observed values of genera richness were similar among groups of classifications except for the stream order classification, which was higher in class 2 than class 1.
For families, differences among groups were observed between the PNRH hydroregions and ecoregions classifications. For the ecoregions, family richness was higher in the Laguna dos Patos ecoregion and similar in the other two Table 3. For abundance data, differences were observed in the stream order classification, the higher values also observed in class 2.
The communities differed between the classes of stream classifications used in this study, independent of the taxonomic resolution and the data type used. The exception was family abundance data that were similar among stream classes defined by stream order Table 4. In general, the CS values were numerically low for the five classifications used and ranging from 0. The taxonomic resolution and data type were important to the result observed, and the CS values for the genus and family depended on the classification system used Table 5.
For taxonomic resolution the macroinvertebrates identified to genera had CS value higher than family data for the stream classification schemes, except for abundance on hydro-PNRH approach. The stream classification obtained by grouping of environmental variables was better than the other classifications used. The mean dendrograms for abundance show arms, i. W i values, relatively longer than presence-absence data.
This indicates that groups based on abundance data are more similar internally than those observed for the presence-absence data Figures 2 and 3. The relatively short arms observed in some classifications and values lower than the global mean between groups similarity , show that these groups are more heterogeneous and, thus, contribute to the low CS values observed.
In relation to taxonomic resolution, arms of the dendrograms obtained with genus data tended to be longer than those obtained with family data. The classifications used in this study were able to capture both the environmental and aquatic macroinvertebrates community variation. Except for the classifications resulting from the grouping of environmental variables and stream orders, the other classifications, the two hydroregions and the ecoregion are different arrangements of the watersheds that drain into the Uruguay River and the Atlantic Ocean.
Thus, we could expect similar results. Our results also show that the taxonomic resolution and data type were important when evaluating the change in macroinvertebrates community and classifications strength CS. The genus level obtained higher values than family when we observed the classification scheme. Added to this, the data type presence-absence and abundance showed slightly similar results among classifications schemes, the abundance data were higher to presence-absence results.
The exception was hydroregion by PNRH, which showed an inversion when we observed the same taxonomic level. The classification strength values observed for the family in general were smaller than those observed for genus when observing the same stream classification Van Sickle and Hughes, ; Sandin and Johnson, ; Houghton, Previous studies have shown that when CS values are not similar between taxonomic levels, the finer level had higher CS value Hawkins et al. Houghton argues that it is more likely that two types of rivers contain the same taxon when we use a coarser taxonomic resolution family when compared to the use of genus and species.
These slightly similar values can explain the weak performance of the hydroregions for family level data. Higher taxonomic resolutions, such as genus and species, tend to be more sensitive to subtle changes in the environment than family Resh and Unzicker, ; Cranston, Studies have shown that hydroregions or other geographical classification schemes usually present lower CS values than classifications based on local environmental variables Van Sickle and Hughes, ; Waite et al.
Van Sickle and Hughes evaluated the performance of environmental classifications for fish and amphibian communities and found that the CS values for ecoregions were higher than hydroregions. They also observed that classifications based on stream order obtained values similar to those obtained with hydroregions. Waite et al. For the genus data, ecoregions obtained 1. Except for ecoregion, values for the genus level were higher than those seen for family level. On the other hand, Houghton showed that CS values obtained between classifications based on watersheds and ecological provinces were different, and the latter showed higher values when comparisons were done for the same spatial scale between the classifications.
The differences between the classification approaches observed in this study are explained primarily by environmental differences among stream classes within each classification. Yang Y, Loog M.
A benchmark and comparison of active learning for logistic regression. Zanaty E. Egypt Inform J. Lossless pruned Naive Bayes for big data classifications. Big Data Res. Tian Q, Chen S. A novel ordinal learning strategy: ordinal nearest-centroid projection. Knowl Based Syst. Mortality prediction system for heart failure with orthogonal relief and dynamic radius means.
Int J Med Inform. Improving random forests by neighborhood projection for effective text classification. Inf Syst. Privacy-preserving ridge regression on distributed data.
Zhu X, Ghahramani Z. Learning from labeled and unlabeled data with label propagation; Yu J, Kim SB. Consensus rate-based label propagation for semi-supervised classification. MathSciNet Google Scholar. Download references. DN carried out the analysis and implementation as well as testing of the software simulations.
TR and PA coordinated and discussed the design of the study and content of this work. All authors read and approved the final manuscript. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. You can also search for this author in PubMed Google Scholar. Correspondence to Maria Teresa Restivo. Reprints and Permissions. Nogueira, D. Comparing classification techniques for identification of grasped objects.
BioMed Eng OnLine 18, 21 Download citation. Received : 10 December Accepted : 26 February Published : 07 March Anyone you share the following link with will be able to read this content:. Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative. Skip to main content. Search all BMC articles Search. Download PDF. Abstract Background This work presents a comparison and selection of different machine learning classification techniques applied in the identification of objects using data collected by an instrumented glove during a grasp process.
Methods The adopted method uses the data from a commercial instrumented glove. Results As a result of this work, three classification techniques presented similar accuracies for the classification of objects. Conclusions Classification techniques were used in two classifier structures, one based on a single model and the other on a cascade model. Background Virtual environments VE have been received significant attention when prepared for diagnoses and treatments, for example, of motor and speech e-rehabilitation in patients [ 1 ].
Table 1 Examples of studies based on the use of technology applied in aphasia Full size table. System architecture. Full size image. CORe implementation The CORe implementation, using the Unity engine, presents a set of features including: Integration of different health monitoring devices: allows patients to use devices in order to carry out e-rehabilitation exercises represented on a virtual environment, promoting e-rehabilitation exercises at home; Local and remote storage of data collected by health monitoring devices during activities for later analysis and reproduction; Real-time and remote view of e-rehabilitation activities: allows a single therapist to connect with many users in a virtual lobby; Gamification in e-rehabilitation: offers game-like activities that take advantage of engagement and motivation for matching the task demands with appropriate feedback and interactive elements; Multiplatform: the software environment supports Windows, Android and also WebGL making the stored data remotely accessible anytime, everywhere and for everyone CORe has not been developed to run on Unix OS.
Database The database is running on a main server for storing the data from the three users and their e-rehabilitation activities. Diagram of the communication between the e-rehabilitation software application and the database. Data collection The aphasia problem can be helped with the development of an application for the identification of grasped objects to be used either as an evaluation or e-training tool.
Set of objects. Table 2 Division of shape group Full size table. Classification techniques ML techniques were used to identify objects, manipulated by an instrumented glove.
Methods, results and discussion The multiclass classification techniques were tested in two scenarios and with two classifier structures, using three models. Table 3 Training and test time of model M0, and accuracy of the first classifier structure Full size table. Table 4 Training time of models M1 and M2 and test time and accuracy of the second classifier structure Full size table. First structure for object classification. Second classifier structure of objects classification using M1 and M2 models.
Table 5 Testing times and accuracies—both scenarios Full size table. Conclusion This paper explores and compares different classification techniques for identification of grasped objects. As future work, the following aspects were identified: To increase the set of objects to be identified since up to now eight different objects divided into four type of shapes were considered ; To include the use of other instrumented gloves with higher number of sensors for comparison studies 5DT Glove used has only five sensors ; To extend the work for clinical trials following our contacts with experts in aphasia area as the expected follow up for an engineering laboratory prototype development process.
References 1. Google Scholar 2. Google Scholar 3. Google Scholar 4. This result may be viewed as part of a more general enterprise to understand the richness and flexibility of conceptual representations and how they are learned. There are many other tasks that require categories, and for these tasks, too, what is learned is likely to be related to how we use them. Not only do we need to broaden the investigation to include other category-learning tasks, we also need to consider that some of the flexibility of conceptual representations may arise from the multiple types of interactions with category members during learning.
To make the task easier during learning, only three answers appeared at the same time, counterbalanced across participants. The confidence is presented for all responses because the proportions of correct responses are low and different in the two conditions. This difference, however, was not replicated in Experiment 2, likely due to the addition of catch trials see the Exp.
It is possible that the change from a real-world background to a white background for test could hurt memory for exemplars, but we felt it critical to examine what participants had learned about the specific birds, not just the often distinctive backgrounds.
We did consider using the white-background pictures for study as well, but decided it was more important to make the study pictures look as if they were of real-world items photographed in their contexts.
Anderson, J. The adaptive nature of human categorization. Psychological Review, 98 , — Article Google Scholar. Anderson, A. A further investigation of category learning by inference. Chin-Parker, S. The effect of category learning on sensitivity to within-category correlations.
Diagnosticity and prototypicality in category learning: A comparison of inference learning and classification learning. Journal of Experimental Psychology. Learning, Memory, and Cognition, 30 , — PubMed Article Google Scholar. Erickson, J. Inference and classification learning of abstract coherent categories. Learning, Memory, and Cognition, 31 , 86— Gelman, S. Psychological Science, 10 , — Categories and induction in young children. Cognition, 23 , — Jacoby, L. Retrieval practice enhances natural category learning: Effects on accuracy and on metacognition at the levels of items and categories.
Google Scholar. Johansen, M. Category representation for classification and feature inference. Learning, Memory, and Cognition, 31 , — Kruschke, J. Psychological Review, 99 , 22— Markman, A. Category use and category learning. Psychological Bulletin, , — Murphy, G. The big book of concepts. Nosofsky, R. Choice, similarity, and the context theory of classification. Learning, Memory, and Cognition, 10 , — Paivio, A.
Why are pictures easier to recall than words? Psychonomic Science, 11 , 1—2. Rehder, B. Feature inference learning and eyetracking. Journal of Memory and Language, 60 , — Eyetracking and selective attention in category learning. Cognitive Psychology, 51 , 1— Abstract coherent categories.
Learning, Memory, and Cognition, 27 , — Rosch, E. Family resemblances: Studies in the internal structure of categories. Cognitive Psychology, 7 , — Sakamoto, Y. Learning and retention through predictive inference and classification. Journal of Experimental Psychology: Applied. Shepard, R. Recognition memory for words, sentences, and pictures. Journal of Verbal Learning and Verbal Behavior, 6 , — Yamauchi, T. Category learning by inference and classification.
Journal of Memory and Language, 39 , — Download references. Department of Psychology, University of Illinois, E. Daniel St. You can also search for this author in PubMed Google Scholar. Correspondence to Erin L. We thank Chris Wahlheim and Larry Jacoby for generously providing the bird pictures from their Bird Learning project, which were used in both experiments; the Ross lab meeting for their valuable feedback; and Robert Molitor for collecting and organizing the data.
We also thank the anonymous reviewers for their comments on an earlier draft. Reprints and Permissions. Jones, E. Classification versus inference learning contrasted with real-world categories.
Mem Cogn 39, — Download citation. Published : 15 December Issue Date : July Anyone you share the following link with will be able to read this content:. Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative. Skip to main content. Search SpringerLink Search. Download PDF. Abstract Categories are learned and used in a variety of ways, but the research focus has been on classification learning.
Classification versus inference learning: Paradigms and results To better understand how different ways of learning affect what knowledge is acquired, some recent studies have compared classification learning to another major means of category learning, inference, where a learner predicts a missing feature of a classified item e.
Table 1 Family resemblance category structure Full size table. Classification versus inference: Explanations If performance differences between classification and inference learning are due to inherent task differences, it suggests that different category-learning tasks may lead to differences in category knowledge. The present experiments The goal of the present experiments is to contrast these explanations of classification—inference differences.
Within each of theseclassifications, materials are often further organized into groups based on their chemical composition or certain physical or mechanical properties.
Answer: When objects are classified, they are simply put into a group with other similar objects. The classification systems used in biology are based on the similarities and differences in organisms. Without classification systems, scientists would have to talk about individuals and not groups.
0コメント