Action Representation Research Articles

Generic measure preserving transformations and the closed groups they generate

We show that, for a generic measure preserving transformation T, the closed group generated by T is not isomorphic to the topological group \(L^0(\lambda , {{\mathbb {T}}})\) of all Lebesgue measurable functions from [0, 1] to \({\mathbb {T}}\) (taken with pointwise multiplication and the topology of convergence in measure). This result answers a question of Glasner and Weiss. The main step in the proof consists of showing that Koopman representations of ergodic boolean actions of \(L^0(\lambda , {\mathbb T})\) possess a non-trivial spectral property not shared by all unitary representations of \(L^0(\lambda , {{\mathbb {T}}})\). The main tool underlying our arguments is a theorem on the form of unitary representations of \(L^0(\lambda , {{\mathbb {T}}})\) from our earlier work.

Ising-like and Fibonacci anyons from KZ-equations

In this work we present solutions to Knizhnik-Zamolodchikov (KZ) equations corresponding to conformal block wavefunctions of non-Abelian Ising-like and Fibonacci Anyons. We solve these equations around regular singular points in configuration space in terms of hypergeometric functions and derive explicit monodromy representations of the braid group action. This confirms the correct non-Abelian statistics of the solutions. One novelty of our approach is that we explicitly keep track of spin basis states and identify conformal blocks uniquely with such states at relevant points in moduli space.

Jean M. Breny and Shannon L. McMorrow. (2021). Photovoice for Social Justice: Visual Representation in Action. Thousand Oaks, CA: SAGE. VitalSource, 101 pages. (ISBN 978-1-5443-5551-1)

Jean M. Breny and Shannon L. McMorrow. (2021). <i>Photovoice for Social Justice: Visual Representation in Action</i>. Thousand Oaks, CA: SAGE. VitalSource, 101 pages. (ISBN 978-1-5443-5551-1)

Offline reinforcement learning with representations for actions

Prevailing offline reinforcement learning (RL) methods limit the policy within the area supported by the offline dataset to avoid the distributional shift problem. But potential high-reward actions, which are out of the distribution of the dataset, are neglected in these methods. To address such issue, we propose a new method, which generalizes from the offline dataset to out-of-distribution (OOD) actions. Specifically, we design a novel action embedding model to help infer the effect of actions. As a result, our value function reaches a better generalization over the action space, and further alleviate the distributional shift caused by overestimation of OOD actions. Theoretically, we give an information-theoretic explanation on the improvement of the value function’s generalization over the action space. Experiments on D4RL demonstrate that our model improves the performance compared to previous offline RL methods, especially when the experience in the offline dataset is good. We conduct further study and validate that the value function’s generalization on OOD actions is improved, which reinforces the effectiveness of our proposed action embedding model.

Force and electromyography reflections of sensory action-effect weighting during pinching

Ideomotor theories suggest that different action-effects are not equally important in goal-directed actions, and that task-relevant information are weighted stronger during the representation of actions. This stronger weighting of task-relevant action-effects might also enable to utilize them as retrieval cues of the corresponding motor patterns. The aim of the present study was to investigate how the consistent presence or absence of a sound action-effect influenced the retrieval of the motor components of a simple, everyday action (pinching) as reflected by the pattern of force application and surface electromyogram (sEMG) recorded from the abductor pollicis brevis (APB) and first dorsal interosseous (FDI). Participants applied pairs of pinch impulses to a force sensitive resistor (FSR). The presence or absence of a sound action-effect and the between-action interval (BAI, 2 or 4s) were manipulated blockwise, whereas the target force level (low or high) was randomly cued from trial to trial. When actions resulted in a sound, force and sEMG activity were reduced. This effect was more pronounced for low target force level trials, which is compatible with a stronger weighting of the sound action-effect when the intensity of the tactile and proprioceptive action-effects is low. Surprisingly, the FDI activity was more variable within actions pairs in the 2s BAI conditions, which suggests that action pairs separated by the longer time interval might have been represented differently from those separated by the shorter interval.

Machine Vision-Based Human Action Recognition Using Spatio-Temporal Motion Features (STMF) with Difference Intensity Distance Group Pattern (DIDGP)

In recent years, human action recognition is modeled as a spatial-temporal video volume. Such aspects have recently expanded greatly due to their explosively evolving real-world uses, such as visual surveillance, autonomous driving, and entertainment. Specifically, the spatio-temporal interest points (STIPs) approach has been widely and efficiently used in action representation for recognition. In this work, a novel approach based on the STIPs is proposed for action descriptors i.e., Two Dimensional-Difference Intensity Distance Group Pattern (2D-DIDGP) and Three Dimensional-Difference Intensity Distance Group Pattern (3D-DIDGP) for representing and recognizing the human actions in video sequences. Initially, this approach captures the local motion in a video that is invariant to size and shape changes. This approach extends further to build unique and discriminative feature description methods to enhance the action recognition rate. The transformation methods, such as DCT (Discrete cosine transform), DWT (Discrete wavelet transforms), and hybrid DWT+DCT, are utilized. The proposed approach is validated on the UT-Interaction dataset that has been extensively studied by past researchers. Then, the classification methods, such as Support Vector Machines (SVM) and Random Forest (RF) classifiers, are exploited. From the observed results, it is perceived that the proposed descriptors especially the DIDGP based descriptor yield promising results on action recognition. Notably, the 3D-DIDGP outperforms the state-of-the-art algorithm predominantly.

Unsupervised skeleton-based action representation learning via relation consistency pursuit

Unsupervised skeleton-based action representation learning via relation consistency pursuit

Acute changes in oxytocin predict behavioral responses to foundation training in horses

Ensuring horse welfare is a central aim in equestrian activities. Training is an important context for welfare, as horses form long-lasting representations of people and actions at a young age. However, only a few studies have addressed horses’ emotional responses during early training with humans. In this study, we followed N = 19 young horses, including naïve yearlings and more experienced two- to three-year-olds, through five foundation training sessions over nine months. Our goal was to combine physiological and behavioral measures to assess emotional responses to early foundation training. Specifically, we measured salivary oxytocin (sOXT) in N = 100 samples and salivary cortisol (sCORT) in N = 96 samples before and after training sessions. We also recoded behavioral responses during training. Changes in sOXT during training predicted individual variation in behavioral responses: Horses who showed more affiliative human-directed behaviors during training had increases in sOXT, while horses who showed more behavioral indicators of discomfort during training had decreases in sOXT. Salivary cortisol was not related to individual behavioral responses, but experienced horses had lower sCORT concentrations both before and after training, and all horses showed decreases in sCORT and in behaviors indicative of fear or discomfort as training progressed. In addition, sCORT increased during longer training sessions, consistent with the established role of cortisol in responding to physical stressors. We conclude that individual variation in positive or negative behavioral responses to foundation training corresponds with acute changes in sOXT concentrations in young horses, suggesting that sOXT may be useful as a non-invasive indicator of emotional responses in young horses.

When proxy-driven learning is no better than random: The consequences of representational incompleteness.

Machine learning is widely used for personalisation, that is, to tune systems with the aim of adapting their behaviour to the responses of humans. This tuning relies on quantified features that capture the human actions, and also on objective functions—that is, proxies – that are intended to represent desirable outcomes. However, a learning system’s representation of the world can be incomplete or insufficiently rich, for example if users’ decisions are based on properties of which the system is unaware. Moreover, the incompleteness of proxies can be argued to be an intrinsic property of computational systems, as they are based on literal representations of human actions rather than on the human actions themselves; this problem is distinct from the usual aspects of bias that are examined in machine learning literature. We use mathematical analysis and simulations of a reinforcement-learning case study to demonstrate that incompleteness of representation can, first, lead to learning that is no better than random; and second, means that the learning system can be inherently unaware that it is failing. This result has implications for the limits and applications of machine learning systems in human domains.

Show Me What You've B/Seen: A Brief History of Depiction.

Already at a relatively early stage, modern sign language linguistics focused on the representation of (actions, locations, and motions of) referents (1) through the use of the body and its different articulators and (2) through the use of particular handshapes (in combination with an orientation, location, and/or movement). Early terminology for (1) includes role playing, role shifting, and role taking and for (2) classifier constructions/predicates and verbs of motion and location. More recently, however, new terms, including enactment and constructed action for (1) and depicting signs for (2) have been introduced. This article provides a brief overview of the history of enactment and depiction in the sign linguistic literature but mainly focuses on issues related to terminology (and terminology shifts). First, we consider the relation between role shifting and constructed action. We question the idea that these terms can be used interchangeably and rather suggest that they capture different, but related functions. Subsequently, we zoom in on the conceptualization of depicting signs, indicating verbs, pointing signs and fully lexical signs and the relation between these signs and the method of depicting. Where earlier research often associates depicting with the use of specific types of structures, we promote the idea that depicting is a semiotic diverse practice. In doing so, we show that the conceptualization of the different sign types and the terms that are used to refer to these phenomena do not accurately capture the way these signs are used in actual signed discourse and propose a reconceptualization of the different sign types in the lexico-grammar of Flemish Sign Language (VGT) as composite signs that can describe, depict and indicate meaning in various ways. In this way, this article illustrates (1) the risks that may come with the execution of terminology shifts and (2) the importance of making a clear distinction between form and function, i.e., we show that it is important to be careful with assuming a (too) exclusive relation between a certain function and one or more particular forms.

The Link Between Action Verb Processing and Action Observation: A Developmental Study.

In this work we aimed to assess the typical link in human development between action observation and language. For this, we studied, in 68 children aged 5-11years of age, how action verbs can prime action representation. While children 7-8years of age benefited from a congruent action verb prime when they had to judge an image representing an action, this effect was not present in 5-6-year-olds. Thus, the link between language and action observation changes during development at about age 7. We discussed these findings in consideration of current theories proposed to account for the action-language link.

Colloquy with Juan José García-Noblejas on understanding communication from the perspective of Aristotle's Poetics

In this interview we will explore García-Noblejas’ thought on public communication, which is deeply marked by his interpretation of Aristotle’s Poetics and influenced by some contemporary thinkers. For García-Noblejas, communication is a service and a form of practical knowledge with five dimensions: ethics, politics, rhetoric, aesthetics, and poetics. Poetics is the most important of these, since its symbolic perspective of the representation of human actions allows us to deeply understand the narrative and dramatic texts typical of journalism, advertising, and propaganda, as well as written and audiovisual fiction. Poetics is the art of creating “possible worlds” in which the reader/spectator recognizes himself (or herself) and with which he dialogues starting from his own identity. García-Noblejas defends public communication that is based on the gift of self and looks to community fruitfulness. He also discusses some frequent pathologies in the media. Finally, he proposes his theory of the Second Navigation to discern the meaning of a text and explains why it is necessary to give a transcendent meaning to the Aristotelian concept of catharsis.

Social representation of military profession and military actions

Annotation: The research of social representation allow as to shape the social reality of a variety of phenomena, values, objects, etc., and everything around us through the collectiveindividualistic view of society members. At present time, the topic of war has a primary actuality worldwide taking into consideration military actions that take place in Ukraine. The research was started before the military actions initiated by Russian Federation on Ukrainian territory, and continued after – February-April 2022. We would like to mention that the given situation marked as well the results of our research, taking into account the fact that our main topic is military field. In this order of ideas, we researched the social representation of military profession and social representation of military actions. The subjects involved in research are employers with active military status.

Inferior: The Challenges of Gender Parity in the Artificial Intelligence Ecosystem-A Case for Canada.

Artificial Intelligence (AI) systems are gaining momentum in complementing and/or replacing performing tasks typically done with the aid of human ability. AI systems, inherently human creations, are, however, beset by, wittingly or unwittingly, so-called male chauvinism, despite all the advancements made in the progress of civilization to make inroads for women's equitable participation in the labor force, particularly in relation to the digital economy, and more importantly, AI. In regards to the Canadian context, this perspective has examined the evidence to find research highlighting gender representation in the Canadian AI ecosystem. We found a lack of studies on women and their contribution to AI-related activities. Canadian women's participation in their country's AI sector therefore should go beyond mere instruments such as the Montreal Declaration for a Responsible Development of AI, and disjointed interests. On a more general level, the paucity in a paradigm shift toward AI-female friendly policies from design phase to implementation omits the female voice for adequate representation for action. Advocating for Canadian women in the AI sector requires a voice of unison best achieved through parliamentary action. This perspective is thus issuing a clarion call to attaining gender fairness and equity, global principles under the United Nations (UN) Sustainable Development Goals, to which the Government of Canada is committed.

Invaders, Attackers and Destroyers: Trespassing-related Terms and Representations in Nigerian Newspaper Headlines

Discourses on herding have focussed on the "exact" representations of the social actions of itinerant herders who clash with farmers while grazing on supposed cattle routes. Media coverage on the herdsmen-farmers conflict has deployed ideologically laden terms to represent herding as trespassing on farmlands and herders as foreigners and trespassers. Using van Leeuwen's Representation of Social Actions and Actors model and Martin and White's Appraisal Framework, this paper examines how different trespassing-related terms (i.e. invade, attack and destroy) were deployed in the Nigerian newspaper headlines to represent herders and their activities with a view to discussing the kinds of representations that were constructed of the nomads through the texts. Findings revealed that using transactive role allocations, nominalization, descriptivation, identification, aggregation and attitudinal lexicalization, these social actors were evaluated negatively as intruders, raiders, and destroyers. The negative othering underscores the general perception and suspicious treatment of nomads in their host farming communities.

1970년대 중반 한국화단의 평면 양상: 반복 행위가 표상된 드로잉을 중심으로

This study discusses the “new flatness” that stood out in the Korean art scene in the mid-1970s and affirms the individuality of flat works by showing that these works cannot be generalized from the perspective of formalism. New flatness was avant-garde on the international stage that emerged in the wake of the Protests of 1968. It is explained as a project that sought to subvert the painting medium through pictorial elements in the context of post-structuralism. The Korean art community accepted this contemporary avant-garde through the 8th Biennale de Paris (1973). Its influence can be confirmed in Ufan Lee’s From the Line, Seobo Park’s Ecriture, and Lee Kun-Yong’s The Method of Drawing. The dominant narrative at the time emphasized the similarity of these works for their representation of repetitive actions. This paper argues against this narrative of formal similarity and will substantiate the individuality of flat works.

The problem of integrated quality assurance and safety of vehicle operation

Problem. The article analyzes the theoretical and methodological foundations of product quality management. The problem of the quality and safety of the operation of vehicles is considered from three angles. The first is directly the problem of the quality and reliability of vehicles as products. Therefore, a quantitative assessment of the quality of a wheeled vehicle is an urgent task and will be discussed in the article. The second aspect is the human-machine interaction in the "man - car - road - environment" system and directly the quality of the human performer driving the car. This aspect depends immediately on the ergonomic qualities of the car. The third aspect concerns the quality of the society and the issues that affect the quality of a quality vehicle designer. Goal. The main purpose of the research is to substantiate the methodological foundations of vehicle quality management as a product. The practical purpose of this work is to develop a methodology for a comprehensive quantitative assessment of the quality, reliability and safety of wheeled vehicles. Methodology. The paper considers groups of product quality characteristics. The possibility of a comprehensive qualitative assessment of the quality of vehicles is determined. The relevance of such an assessment is due to the wide variety of vehicles, the improvement of their technical characteristics against the backdrop of a slower development of the characteristics of the environment in which they are used. Another important circumstance is slow development of methods for assessing the quality of vehicles, which makes it difficult to choose the best option in accordance with the goals of their operation. Originality. The paper shows an extensive range of vehicle properties. A technique for a comprehensive assessment of the quality of vehicles is proposed. An assumption is made about the representation of quality indicators in a multidimensional space, in which each dimension corresponds to a certain group of properties. Practical value. To improve the objectivity of a comprehensive assessment, a number of units of property characteristics are described that have a metric dimension and are based on the physical and material representation of actions. It is concluded that it is necessary to develop and refine the methodology for a comprehensive quantitative assessment of quality in the area under research.

Contrastive Learning from Extremely Augmented Skeleton Sequences for Self-Supervised Action Recognition

In recent years, self-supervised representation learning for skeleton-based action recognition has been developed with the advance of contrastive learning methods. The existing contrastive learning methods use normal augmentations to construct similar positive samples, which limits the ability to explore novel movement patterns. In this paper, to make better use of the movement patterns introduced by extreme augmentations, a Contrastive Learning framework utilizing Abundant Information Mining for self-supervised action Representation (AimCLR) is proposed. First, the extreme augmentations and the Energy-based Attention-guided Drop Module (EADM) are proposed to obtain diverse positive samples, which bring novel movement patterns to improve the universality of the learned representations. Second, since directly using extreme augmentations may not be able to boost the performance due to the drastic changes in original identity, the Dual Distributional Divergence Minimization Loss (D3M Loss) is proposed to minimize the distribution divergence in a more gentle way. Third, the Nearest Neighbors Mining (NNM) is proposed to further expand positive samples to make the abundant information mining process more reasonable. Exhaustive experiments on NTU RGB+D 60, PKU-MMD, NTU RGB+D 120 datasets have verified that our AimCLR can significantly perform favorably against state-of-the-art methods under a variety of evaluation protocols with observed higher quality action representations. Our code is available at https://github.com/Levigty/AimCLR.

Target Languages (vs. Inductive Biases) for Learning to Act and Plan

Recent breakthroughs in AI have shown the remarkable power of deep learning and deep reinforcement learning. These developments, however, have been tied to specific tasks, and progress in out-of-distribution generalization has been limited. While it is assumed that these limitations can be overcome by incorporating suitable inductive biases, the notion of inductive biases itself is often left vague and does not provide meaningful guidance. In the paper, I articulate a different learning approach where representations do not emerge from biases in a neural architecture but are learned over a given target language with a known semantics. The basic ideas are implicit in mainstream AI where representations have been encoded in languages ranging from fragments of first-order logic to probabilistic structural causal models. The challenge is to learn from data the representations that have traditionally been crafted by hand. Generalization is then a result of the semantics of the language. The goals of this paper are to make these ideas explicit, to place them in a broader context where the design of the target language is crucial, and to illustrate them in the context of learning to act and plan. For this, after a general discussion, I consider learning representations of actions, general policies, and subgoals ("intrinsic rewards"). In these cases, learning is formulated as a combinatorial problem but nothing prevents the use of deep learning techniques instead. Indeed, learning representations over languages with a known semantics provides an account of what is to be learned, while learning representations with neural nets provides a complementary account of how representations can be learned. The challenge and the opportunity is to bring the two together.

Action-Aware Embedding Enhancement for Image-Text Retrieval

Image-text retrieval plays a central role in bridging vision and language, which aims to reduce the semantic discrepancy between images and texts. Most of existing works rely on refined words and objects representation through the data-oriented method to capture the word-object cooccurrence. Such approaches are prone to ignore the asymmetric action relation between images and texts, that is, the text has explicit action representation (i.e., verb phrase) while the image only contains implicit action information. In this paper, we propose Action-aware Memory-Enhanced embedding (AME) method for image-text retrieval, which aims to emphasize the action information when mapping the images and texts into a shared embedding space. Specifically, we integrate action prediction along with an action-aware memory bank to enrich the image and text features with action-similar text features. The effectiveness of our proposed AME method is verified by comprehensive experimental results on two benchmark datasets.