Abstract: We report two studies validating a new standardized set of filmed emotion expressions, the Amsterdam Dynamic Facial Expression Set (ADFES). The ADFES is distinct from existing datasets in that it includes a face-forward version and two different head-turning versions (faces turning toward and away from viewers), North-European as well as Mediterranean models (male and female), and nine discrete emotions (joy, anger, fear, sadness, surprise, disgust, contempt, pride, and embarrassment). Study 1 showed that the ADFES received excellent recognition scores. Recognition was affected by social categorization of the model: displays of North-European models were better recognized by Dutch participants, suggesting an ingroup advantage. Head-turning did not affect recognition accuracy. Study 2 showed that participants more strongly perceived themselves to be the cause of the other's emotion when the model's face turned toward the respondents. The ADFES provides new avenues for research on emotion expression and is available for researchers upon request.

Abstract: Text categorization is widely used when organizing documents in a digital form. Due to the increasing number of documents in digital form, automated text categorization has become more promising in the last ten years. A major problem of text categorization is its large number of features. Most of those are irrelevant noise that can mislead the classifier. Therefore, feature selection is often used in text categorization to reduce the dimensionality of the feature space and to improve performance. In this study, two-stage feature selection and feature extraction is used to improve the performance of text categorization. In the first stage, each term within the document is ranked depending on their importance for classification using the information gain (IG) method. In the second stage, genetic algorithm (GA) and principal component analysis (PCA) feature selection and feature extraction methods are applied separately to the terms which are ranked in decreasing order of importance, and a dimension reduction is carried out. Thereby, during text categorization, terms of less importance are ignored, and feature selection and extraction methods are applied to the terms of highest importance; thus, the computational time and complexity of categorization is reduced. To evaluate the effectiveness of dimension reduction methods on our purposed model, experiments are conducted using the k-nearest neighbour (KNN) and C4.5 decision tree algorithm on Reuters-21,578 and Classic3 datasets collection for text categorization. The experimental results show that the proposed model is able to achieve high categorization effectiveness as measured by precision, recall and F-measure.

Abstract: During the 1990s and early 2000s, cognitive neuroscience investigations of human category learning focused on the primary goal of showing that humans have multiple category-learning systems and on the secondary goals of identifying key qualitative properties of each system and of roughly mapping out the neural networks that mediate each system. Many researchers now accept the strength of the evidence supporting multiple systems, and as a result, during the past few years, work has begun on the second generation of research questions-that is, on questions that begin with the assumption that humans have multiple category-learning systems. This article reviews much of this second generation of research. Topics covered include (1) How do the various systems interact? (2) Are there different neural systems for categorization and category representation? (3) How does automaticity develop in each system? and (4) Exactly how does each system learn?

Abstract: Portable high-quality sports cameras (e.g. head or helmet mounted) built for recording dynamic first-person video footage are becoming a common item among many sports enthusiasts. We address the novel task of discovering first-person action categories (which we call ego-actions) which can be useful for such tasks as video indexing and retrieval. In order to learn ego-action categories, we investigate the use of motion-based histograms and unsupervised learning algorithms to quickly cluster video content. Our approach assumes a completely unsupervised scenario, where labeled training videos are not available, videos are not pre-segmented and the number of ego-action categories are unknown. In our proposed framework we show that a stacked Dirichlet process mixture model can be used to automatically learn a motion histogram codebook and the set of ego-action categories. We quantitatively evaluate our approach on both in-house and public YouTube videos and demonstrate robust ego-action categorization across several sports genres. Comparative analysis shows that our approach outperforms other state-of-the-art topic models with respect to both classification accuracy and computational speed. Preliminary results indicate that on average, the categorical content of a 10 minute video sequence can be indexed in under 5 seconds.

Abstract: Most theories of categorization emphasize how continuous perceptual information is mapped to categories. However, equally important are the informational assumptions of a model, the type of information subserving this mapping. This is crucial in speech perception where the signal is variable and context dependent. This study assessed the informational assumptions of several models of speech categorization, in particular, the number of cues that are the basis of categorization and whether these cues represent the input veridically or have undergone compensation. We collected a corpus of 2,880 fricative productions (Jongman, Wayland, & Wong, 2000) spanning many talker and vowel contexts and measured 24 cues for each. A subset was also presented to listeners in an 8AFC phoneme categorization task. We then trained a common classification model based on logistic regression to categorize the fricative from the cue values and manipulated the information in the training set to contrast (a) models based on a small number of invariant cues, (b) models using all cues without compensation, and (c) models in which cues underwent compensation for contextual factors. Compensation was modeled by computing cues relative to expectations (C-CuRE), a new approach to compensation that preserves fine-grained detail in the signal. Only the compensation model achieved a similar accuracy to listeners and showed the same effects of context. Thus, even simple categorization metrics can overcome the variability in speech when sufficient information is available and compensation schemes like C-CuRE are employed.

Abstract: Subordinate-level categorization typically rests on establishing salient distinctions between part-level characteristics of objects, in contrast to basic-level categorization, where the presence or absence of parts is determinative. We develop an approach for subordinate categorization in vision, focusing on an avian domain due to the fine-grained structure of the category taxonomy for this domain. We explore a pose-normalized appearance model based on a volumetric poselet scheme. The variation in shape and appearance properties of these parts across a taxonomy provides the cues needed for subordinate categorization. Training pose detectors requires a relatively large amount of training data per category when done from scratch; using a subordinate-level approach, we exploit a pose classifier trained at the basic-level, and extract part appearance and shape information to build subordinate-level models. Our model associates the underlying image pattern parameters used for detection with corresponding volumetric part location, scale and orientation parameters. These parameters implicitly define a mapping from the image pixels into a pose-normalized appearance space, removing view and pose dependencies, facilitating fine-grained categorization from relatively few training examples.

Abstract: It is commonly believed that race is perceived through another's facial features, such as skin color. In the present research, we demonstrate that cues to social status that often surround a face systematically change the perception of its race. Participants categorized the race of faces that varied along White–Black morph continua and that were presented with high-status or low-status attire. Low-status attire increased the likelihood of categorization as Black, whereas high-status attire increased the likelihood of categorization as White; and this influence grew stronger as race became more ambiguous (Experiment 1). When faces with high-status attire were categorized as Black or faces with low-status attire were categorized as White, participants' hand movements nevertheless revealed a simultaneous attraction to select the other race-category response (stereotypically tied to the status cue) before arriving at a final categorization. Further, this attraction effect grew as race became more ambiguous (Experiment 2). Computational simulations then demonstrated that these effects may be accounted for by a neurally plausible person categorization system, in which contextual cues come to trigger stereotypes that in turn influence race perception. Together, the findings show how stereotypes interact with physical cues to shape person categorization, and suggest that social and contextual factors guide the perception of race.

Abstract: Individuals who qualify equally for membership in two racial groups provide a rare window into social categorization and perception. In 5 experiments, we tested the extent to which a rule of hypodescent, whereby biracial individuals are assigned the status of their socially subordinate parent group, would govern perceptions of Asian–White and Black–White targets. In Experiment 1, in spite of posing explicit questions concerning Asian–White and Black–White targets, hypodescent was observed in both cases and more strongly in Black–White social categorization. Experiments 2A and 2B used a speeded response task and again revealed evidence of hypodescent in both cases, as well as a stronger effect in the Black–White target condition. In Experiments 3A and 3B, social perception was studied with a face-morphing task. Participants required a face to be lower in proportion minority to be perceived as minority than in proportion White to be perceived as White. Again, the threshold for being perceived as White was higher for Black–White than for Asian–White targets. An independent categorization task in Experiment 3B further confirmed the rule of hypodescent and variation in it that reflected the current racial hierarchy in the United States. These results documenting biases in the social categorization and perception of biracials have implications for resistance to change in the American racial hierarchy.

Abstract: The categories that social targets belong to are often activated automatically. Most studies investigating social categorization have used visual stimuli or verbal labels, whereas ethnolinguistic identity theory posits that language is an essential dimension of ethnic identity. Language should therefore be used for social categorization. In 2 experiments, using the "Who Said What?" paradigm, the authors investigated social categorization by using accents (auditory stimuli) and looks (visual stimuli) to indicate ethnicity, either separately or in combination. Given either looks or accents only, the authors demonstrated that ethnic categorization can be based on accents, and the authors found a similar degree of ethnic categorization by accents and looks. When ethnic cues of looks and accents were combined by creating cross categories, there was a clear predominance of accents as meaningful cues for categorization, as shown in the respective parameters of a multinomial model. The present findings are discussed with regard to the generalizability of findings using one channel of presentation (e.g., visual) and the asymmetry found with different presentation channels for the category ethnicity.

Abstract: Recently, image categorization has been an active research topic due to the urgent need to retrieve and browse digital images via semantic keywords. This paper formulates image categorization as a multi-label classification problem using recent advances in matrix completion. Under this setting, classification of testing data is posed as a problem of completing unknown label entries on a data matrix that concatenates training and testing features with training labels. We propose two convex algorithms for matrix completion based on a Rank Minimization criterion specifically tailored to visual data, and prove its convergence properties. A major advantage of our approach w.r.t. standard discriminative classification methods for image categorization is its robustness to outliers, background noise and partial occlusions both in the feature and label space. Experimental validation on several datasets shows how our method outperforms state-of-the-art algorithms, while effectively capturing semantic concepts of classes.

Abstract: A B S T R A C T The purpose of this study was to examine the hypothesis that helping preschoolers learn words through categorization may enhance their ability to retain words and their conceptual properties, acting as a bootstrap for self-learning. We examined this hypothesis by investigating the effects of the World of Words instructional program, a supplemental intervention for children in preschool designed to teach word knowledge and conceptual development through taxonomic categorization and embedded multimedia. Participants in the study included 3- and 4-year-old children from 28 Head Start classrooms in 12 schools, randomly assigned to treatment and control groups. Children were assessed on word knowledge, expressive language, conceptual knowledge, and categories and properties of concepts in a yearlong intervention. Results indicated that children receiving the WOW treatment consistently outperformed their control counterparts; further, treatment children were able to use categories to identify the meaning of novel words. Gains in word and categorical knowledge were sustained six months later for those children who remained in Head Start. These results suggest that a program targeted to learning words within taxonomic categories may act as a bootstrap for self-learning and inference generation.

Abstract: A persistent mystery in language acquisition is how speakers are able to learn seemingly arbitrary distributional restrictions. This article investigates one such case: the fact that speakers resist using certain adjectives prenominally (e.g. ?? the asleep man ). Experiment 1 indicates that speakers tentatively generalize or categorize the distributional restriction beyond their previous experience. Experiment 2 demonstrates that speakers are sensitive to statistical preemption—that is, speakers learn not to use a formulation if an alternative formulation with the same function is consistently witnessed. Moreover, they are able to generalize the restriction to apply to other members of the category as well. Finally, experiment 3 finds evidence that speakers discount a pseudopreemptive context, rationally ignoring it as uninformative.

Abstract: A thematic relation is a temporal, spatial, causal, or functional relation between things that perform complementary roles in the same scenario or event. For example, cows and milk are related by a production theme, and sails and anchors are related via a boating theme. Thematic relations are distinct from mere associations, scripts, and ad hoc categories. They also contrast and complement taxonomic (categorical) relations such as “fruits” and “furniture.” Thematic relations and taxonomic relations arise from distinct processes, as evidenced by numerous neuropsychological and behavioral dissociations. Thematic relations may be apprehended uncontrollably and rapidly according to how frequently and recently they have been encountered. They exert profound effects on many core cognitive processes, including similarity, categorization, memory, language, inference, and analogy, and they exhibit robust processing differences across individuals and cultures. In sum, without such thematic thinking, models of cognition will remain categorically limited.

Abstract: 1. Introduction Emmanuel M. Pothos and Andy J. Wills 2. The generalized context model: an exemplar model of classification Robert M. Nosofsky 3. Prototype models of categorization: basic formulation, predictions, and limitations John Paul Minda and J. David Smith 4. COVIS F. Gregory Ashby, Erick J. Paul and W. Todd Maddox 5. Semantics without categorization Timothy T. Rogers and James L. McClelland 6. Models of attentional learning John K. Kruschke 7. An elemental model of associative learning and memory Evan Livesey and Ian McLaren 8. Nonparametric Bayesian models of categorization Thomas L. Griffiths, Adam N. Sanborn, Kevin R. Canini, Daniel J. Navarro and Joshua B. Tenenbaum 9. The simplicity model of unsupervised categorization Emmanuel M. Pothos, Nick Chater and Peter Hines 10. Adaptive clustering models of categorization John V. McDonnell and Todd M. Gureckis 11. COBWEB models of categorization and probabilistic concept formation Wayne Iba and Pat Langley 12. The knowledge and resonance (KRES) model of category learning Harlan D. Harris and Bob Rehder 13. The contribution (and drawbacks) of models to the study of concepts Gregory L. Murphy 14. Formal models of categorization: insights from cognitive neuroscience Lukas Strnad, Stefano Anzellotti and Alfonso Caramazza 15. Comments on models and categorization theories: the razor's edge Douglas Medin.

Abstract: Visual learning admits different levels of complexity, from the formation of a simple associative link between a visual stimulus and its outcome, to more sophisticated performances, such as object categorization or rules learning, that allow flexible responses beyond simple forms of learning. Not surprisingly, higher-order forms of visual learning have been studied primarily in vertebrates with larger brains, while simple visual learning has been the focus in animals with small brains such as insects. This dichotomy has recently changed as studies on visual learning in social insects have shown that these animals can master extremely sophisticated tasks. Here we review a spectrum of visual learning forms in social insects, from color and pattern learning, visual attention, and top-down image recognition, to interindividual recognition, conditional discrimination, category learning, and rule extraction. We analyze the necessity and sufficiency of simple associations to account for complex visual learning in Hymenoptera and discuss possible neural mechanisms underlying these visual performances.

Abstract: The visual recognition problem is central to computer vision research. From robotics to information retrieval, many desired applications demand the ability to identify and localize categories, places, and objects. This tutorial overviews computer vision algorithms for visual object recognition and image classification. We introduce primary representations and learning approaches, with an emphasis on recent advances in the field. The target audience consists of researchers or students working in AI, robotics, or vision who would like to understand what methods and representations are available for these problems. This lecture summarizes what is and isn't possible to do reliably today, and overviews key concepts that could be employed in systems requiring visual categorization. Table of Contents: Introduction / Overview: Recognition of Specific Objects / Local Features: Detection and Description / Matching Local Features / Geometric Verification of Matched Features / Example Systems: Specific-Object Recognition / Overview: Recognition of Generic Object Categories / Representations for Object Categories / Generic Object Detection: Finding and Scoring Candidates / Learning Generic Object Category Models / Example Systems: Generic Object Recognition / Other Considerations and Current Challenges / Conclusions

Abstract: Automated scene interpretation has benefited from advances in machine learning, and restricted tasks, such as face detection, have been solved with sufficient accuracy for restricted settings. However, the performance of machines in providing rich semantic descriptions of natural scenes from digital images remains highly limited and hugely inferior to that of humans. Here we quantify this “semantic gap” in a particular setting: We compare the efficiency of human and machine learning in assigning an image to one of two categories determined by the spatial arrangement of constituent parts. The images are not real, but the category-defining rules reflect the compositional structure of real images and the type of “reasoning” that appears to be necessary for semantic parsing. Experiments demonstrate that human subjects grasp the separating principles from a handful of examples, whereas the error rates of computer programs fluctuate wildly and remain far behind that of humans even after exposure to thousands of examples. These observations lend support to current trends in computer vision such as integrating machine learning with parts-based modeling.

Abstract: Model description Conceptual overview According to the generalized context model (GCM) (Nosofsky, 1986), people represent categories by storing individual exemplars (or examples) in memory, and classify objects based on their similarity to these stored exemplars. For example, the model assumes that people represent the category of ‘birds’ by storing in memory the vast collection of different sparrows, robins, eagles, ostriches (and so forth) that they have experienced. If an object is sufficiently similar to some of these bird exemplars, then the person would tend to classify the object as a ‘bird’. This exemplar view of categorization contrasts dramatically with major alternative approaches that assume that people form abstract summary representations of categories, such as rules or idealized prototypes. The standard version of the GCM adopts a multidimensional scaling (MDS) approach to modelling similarity relations among exemplars (Shepard, 1958, 1987). In this approach, exemplars are represented as points in a multidimensional psychological space. Similarity between exemplars is a decreasing function of their distance in the space. In many applications, a first step in the modelling is to conduct similarity-scaling studies to derive MDS solutions for the exemplars and to discover their locations in the multidimensional similarity space (Nosofsky, 1992b). A crucial assumption in the modelling, however, is that similarity is not an invariant relation, but a highly context-dependent one. To take an example from Medin and Schaffer (1978), humans and mannequins may be judged as highly similar in a context that emphasizes structural appearance, but would be judged as highly dissimilar in a context that emphasizes vitality.

Abstract: Acetylcholine (ACh) is a neurotransmitter acting via muscarinic and nicotinic receptors that is implicated in several cognitive functions and impairments, such as Alzheimer’s disease. It is believed to especially affect the acquisition of new information, which is particularly important when behavior needs to be adapted to new situations and to novel sensory events. Categorization, the process of assigning stimuli to a category, is a cognitive function that also involves information acquisition. The role of ACh on categorization has not been previously studied. We have examined the effects of scopolamine, an antagonist of muscarinic ACh receptors, on visual categorization in macaque monkeys using familiar and novel stimuli. When the peripheral effects of scopolamine on the parasympathetic nervous system were controlled for, categorization performance was disrupted following systemic injections of scopolamine. This impairment was observed only when the stimuli that needed to be categorized had not been seen before. In other words, the monkeys were not impaired by the central action of scopolamine in categorizing a set of familiar stimuli (stimuli which they had categorized successfully in previous sessions). Categorization performance also deteriorated as the stimulus became less salient by an increase in the level of visual noise. However, scopolamine did not cause additional performance disruptions for difficult categorization judgments at lower coherence levels. Scopolamine, therefore, specifically affects the assignment of new exemplars to established cognitive categories, presumably by impairing the processing of novel information. Since we did not find an effect of scopolamine in the categorization of familiar stimuli, scopolamine had no significant central action on other cognitive functions such as perception, attention, memory, or executive control within the context of our categorization task.

Abstract: One of the most fascinating issues in neuroscience is how the brain makes decisions Recent evidence points to the parietal cortex as an important locus for certain kinds of decisions Because parietal neurons are also involved in movements, it has been proposed that decisions are encoded in an intentional, action-based framework based on the movements used to report decisions An alternative or complementary view is that decisions represent more abstract information not linked to movements per se Parallel experiments on categorization suggest that parietal neurons can indeed represent abstract categorical outcomes that are not linked to movements This could provide a unified or complementary view of how the brain decides and categorizes

Abstract: A ranked set of users may be calculated from an expertise categorization for each user and a person's trust in the users for specific categories. The ranked set of users may be used for presenting search results, recommendations, social marketing, or other uses. A person's reputation may be determined through various online activities. A person's trust in another person may be related to their proximity and activity in one or more social networks.

Abstract: Humans are able to categorize complex natural scenes very rapidly and effortlessly, which has led to an assumption that such ultra-rapid categorization is driven by feedforward activation of ventral brain areas. However, recent accounts of visual perception stress the role of recurrent interactions that start rapidly after the activation of V1. To study whether or not recurrent processes play a causal role in categorization, we applied fMRI-guided transcranial magnetic stimulation on early visual cortex (V1/V2) and lateral occipital cortex (LO) while the participants categorized natural images as containing animals or not. The results showed that V1/V2 contributed to categorization speed and to subjective perception during a long activity period before and after the contribution of LO had started. This pattern of results suggests that recurrent interactions in visual cortex between areas along the ventral stream and striate cortex play a causal role in categorization and perception of natural scenes.

Abstract: Four experiments (E1-E2-E3-E4) investigated whether different acquisition modalities lead to the emergence of differences typically found between concrete and abstract words, as argued by the Words As Tools (WAT) proposal. To mimic the acquisition of concrete and abstract concepts, participants either manipulated novel objects or observed groups of objects interacting in novel ways (training1). In TEST 1 participants decided whether two elements belonged to the same category. Later they read the category labels (training2); labels could be accompanied by an explanation of their meaning. Then participants observed previously seen exemplars and other elements, and were asked which of them could be named with a given label (TEST2). Across the experiments, it was more difficult to form abstract than concrete categories (TEST 1); even when adding labels, abstract words remained more difficult than concrete words (TEST 2). TEST3 differed across the experiments. In E1 participants performed a feature production task. Crucially, the associations produced with the novel words reflected the pattern evoked by existing concrete and abstract words, as the first evoked more perceptual properties. In E2-E3-E4, TEST3 consisted of a color verification task with manual/verbal (keyboard-microphone) responses. Results showed the microphone use to have an advantage over keyboard use for abstract words, especially in the explanation condition. This supports WAT: due to their acquisition modality, concrete words evoke more manual information; abstract words elicit more verbal information. This advantage was not present when linguistic information contrasted with perceptual one. Implications for theories and computational models of language grounding are discussed.

Abstract: In this article we describe an object perception system for autonomous robots performing everyday manipulation tasks in kitchen environments. The perception system gains its strengths by exploiting that the robots are to perform the same kinds of tasks with the same objects over and over again. It does so by learning the object representations necessary for the recognition and reconstruction in the context of pick-and-place tasks. The system employs a library of specialized perception routines that solve different, well-defined perceptual sub-tasks and can be combined into composite perceptual activities including the construction of an object model database, multimodal object classification, and object model reconstruction for grasping. We evaluate the effectiveness of our methods, and give examples of application scenarios using our personal robotic assistants acting in a human living environment.

Abstract: Visual categorization is important to manage large collections of digital images and video, where textual metadata is often incomplete or simply unavailable. The bag-of-words model has become the most powerful method for visual categorization of images and video. Despite its high accuracy, a severe drawback of this model is its high computational cost. As the trend to increase computational power in newer CPU and GPU architectures is to increase their level of parallelism, exploiting this parallelism becomes an important direction to handle the computational cost of the bag-of-words approach. When optimizing a system based on the bag-of-words approach, the goal is to minimize the time it takes to process batches of images. this paper, we analyze the bag-of-words model for visual categorization in terms of computational cost and identify two major bottlenecks: the quantization step and the classification step. We address these two bottlenecks by proposing two efficient algorithms for quantization and classification by exploiting the GPU hardware and the CUDA parallel programming model. The algorithms are designed to (1) keep categorization accuracy intact, (2) decompose the problem, and (3) give the same numerical results. In the experiments on large scale datasets, it is shown that, by using a parallel implementation on the Geforce GTX260 GPU, classifying unseen images is 4.8 times faster than a quad-core CPU version on the Core i7 920, while giving the exact same numerical results. In addition, we show how the algorithms can be generalized to other applications, such as text retrieval and video retrieval. Moreover, when the obtained speedup is used to process extra video frames in a video retrieval benchmark, the accuracy of visual categorization is improved by 29%.

Abstract: Working memory is crucial for many higher-level cognitive functions, ranging from mental arithmetic to reasoning and problem solving. Likewise, the ability to learn and categorize novel concepts forms an indispensable part of human cognition. However, very little is known about the relationship between working memory and categorization, and modeling in category learning has thus far been largely uninformed by knowledge about people's memory processes. This article reports a large study (N = 113) that related people's working memory capacity (WMC) to their category-learning performance using the 6 problem types of Shepard, Hovland, and Jenkins (1961). Structural equation modeling revealed a strong relationship between WMC and category learning, with a single latent variable accommodating performance on all 6 problems. A model of categorization (the Attention Learning COVEring map, ALCOVE; Kruschke, 1992) was fit to the individual data and a single latent variable was sufficient to capture the variation among associative learning parameters across all problems. The data and modeling suggest that working memory mediates category learning across a broad range of tasks.