Extrinsic Representations Research Articles

Interactive Learning of Intrinsic and Extrinsic Properties for All-Day Semantic Segmentation.

Scene appearance changes drastically throughout the day. Existing semantic segmentation methods mainly focus on well-lit daytime scenarios and are not well designed to cope with such great appearance changes. Naively using domain adaption does not solve this problem because it usually learns a fixed mapping between the source and target domain and thus have limited generalization capability on all-day scenarios (i. e., from dawn to night). In this paper, in contrast to existing methods, we tackle this challenge from the perspective of image formulation itself, where the image appearance is determined by both intrinsic (e. g., semantic category, structure) and extrinsic (e. g., lighting) properties. To this end, we propose a novel intrinsic-extrinsic interactive learning strategy. The key idea is to interact between intrinsic and extrinsic representations during the learning process under spatial-wise guidance. In this way, the intrinsic representation becomes more stable and, at the same time, the extrinsic representation gets better at depicting the changes. Consequently, the refined image representation is more robust to generate pixel-wise predictions for all-day scenarios. To achieve this, we propose an All-in-One Segmentation Network (AO-SegNet) in an end-to-end manner. Large scale experiments are conducted on three real datasets (Mapillary, BDD100K and ACDC) and our proposed synthetic All-day CityScapes dataset. The proposed AO-SegNet shows a significant performance gain against the state-of-the-art under a variety of CNN and ViT backbones on all the datasets.

Detail-Preserving Shape Unfolding.

Canonical extrinsic representations for non-rigid shapes with different poses are preferable in many computer graphics applications, such as shape correspondence and retrieval. The main reason for this is that they give a pose invariant signature for those jobs, which significantly decreases the difficulty caused by various poses. Existing methods based on multidimentional scaling (MDS) always result in significant geometric distortions. In this paper, we present a novel shape unfolding algorithm, which deforms any given 3D shape into a canonical pose that is invariant to non-rigid transformations. The proposed method can effectively preserve the local structure of a given 3D model with the regularization of local rigid transform energy based on the shape deformation technique, and largely reduce geometric distortion. Our algorithm is quite simple and only needs to solve two linear systems during alternate iteration processes. The computational efficiency of our method can be improved with parallel computation and the robustness is guaranteed with a cascade strategy. Experimental results demonstrate the enhanced efficacy of our algorithm compared with the state-of-the-art methods on 3D shape unfolding.

Research on Software Runtime Credibility Based on Cluster Analysis

Monitoring and analyzing the reliability of software runtime behavior in an open, dynamic, and complex network environment is very important for modern distributed software. The characteristics of the extrinsic representation in the execution of distributed software are collected, and the data is executed based on the history of the software, and the wide square cube cluster is constructed, and the reliability is analyzed, and the fault tolerance processing and the fault diagnosis of the succeeding system are analyzed. Performance optimization is facilitated. Simulation experiments show that the proposed method has a unique advantage in comparison with other methods in reliability analysis of software behavior.

Ricci flow on surfaces of revolution: an extrinsic view

A Ricci flow (M, g(t)) on an n-dimensional Riemannian manifold M is an intrinsic geometric flow. A family of smoothly embedded submanifolds \((S(t), g_E)\) of a fixed Euclidean space \(E = \mathbb {R}^{n+k}\) is called an extrinsic representation in \(\mathbb {R}^{n+k}\) of (M, g(t)) if there exists a smooth one-parameter family of isometries \((S(t), g_E) \rightarrow (M, g(t))\). When does such a representation exist? We formulate a new way of framing this question for Ricci flows on surfaces of revolution immersed in \(\mathbb {R}^3\). This framework allows us to construct extrinsic representations for the Ricci flow initialized by any compact surface of revolution immersed in \(\mathbb {R}^3\). In particular, we exhibit the first explicit extrinsic representations in \(\mathbb {R}^4\) of the Ricci flows initialized by toroidal surfaces of revolution immersed in \(\mathbb {R}^3\).

Generalization via superposition: combined effects of mixed reference frame representations for explicit and implicit learning in a visuomotor adaptation task.

Studies on generalization of learned visuomotor perturbations have generally focused on whether learning is coded in extrinsic or intrinsic reference frames. This dichotomy, however, is challenged by recent findings showing that learning is represented in a mixed reference frame. Overlooked in this framework is how learning appears to consist of multiple processes, such as explicit reaiming and implicit motor adaptation. Therefore, the proposed mixed representation may simply reflect the superposition of explicit and implicit generalization functions, each represented in different reference frames. Here we characterized the individual generalization functions of explicit and implicit learning in relative isolation to determine whether their combination could predict the overall generalization function when both processes are in operation. We modified the form of feedback in a visuomotor rotation task in an attempt to isolate explicit and implicit learning and tested generalization across new limb postures to dissociate the extrinsic/intrinsic representations. We found that the amplitude of explicit generalization was reduced with postural change and was only marginally shifted, resembling an extrinsic representation. In contrast, implicit generalization maintained its amplitude but was significantly shifted, resembling a mixed representation. A linear combination of individual explicit and implicit generalization functions accounted for nearly 85% of the variance associated with the generalization function in a typical visuomotor rotation task, where both processes are in operation. This suggests that each form of learning results from a mixed representation with distinct extrinsic and intrinsic contributions and the combination of these features shapes the generalization pattern observed at novel limb postures. NEW & NOTEWORTHY Generalization following learning in visuomotor adaptation tasks can reflect how the brain represents what it learns. In this study, we isolated explicit and implicit forms of learning and showed that they are derived from a mixed reference frame representation with distinct extrinsic and intrinsic contributions. Furthermore, we showed that the overall generalization pattern at novel workspaces is due to the superposition of independent generalization effects developed by explicit and implicit learning processes.

Beautiful Physics: Re-vision of Aesthetic Features of Science Through the Literature Review

This study intends to reveal the aesthetic features of physics through the literature review and to give some implications for teaching the beauty of physics. Among the papers published in the internationally prominent journals, indexed in SCI, SSCI, and A&HCI, this study selects a total of 88 articles dealing with aesthetics of physics, and categorizes the aesthetics of physics into intrinsic and extrinsic representation of the aesthetics. As a result, simplicity, symmetry, harmony, and unity, centering on mathematical formalism, are found as the intrinsic aspects of aesthetics, whereas elegance, wonder and sublimity are mentioned as the extrinsic aspects of the aesthetics, which were evoked from the intrinsic aesthetics above. In terms of ontological and epistemological nature of aesthetics, modern physics does not give a clear answer about the simplicity of nature due to the indeterminacy, and the aesthetic judgment is linked to both intrinsic nature of an object and extrinsic sense of human mind, and finally aesthetic evaluation is determined by empirical worth of aesthetics and shared meaning of aesthetic attitude. Thus, this study gives some implications for the use of visual image in physics and teaching aesthetics of physics.

Musculoskeletal geometry accounts for apparent extrinsic representation of paw position in dorsal spinocerebellar tract.

Proprioception, the sense of limb position and motion, arises from individual muscle receptors. An important question is how and where in the neuroaxis our high level "extrinsic" sense of limb movement originates. In the 1990s, a series of papers detailed the properties of neurons in the dorsal spinocerebellar tract (DSCT) of the cat. Despite their direct projections from sensory receptors, it appeared that half of these neurons had consistent, high-level tuning to paw position rather than to joint angles (or muscle lengths). These results suggested that many DSCT neurons compute paw position from lower level sensory information. We examined the contribution of musculoskeletal geometry to this apparent extrinsic representation by simulating a three-joint hindlimb with mono- and biarticular muscles, each providing a muscle spindlelike signal, modulated by the muscle length. We simulated neurons driven by randomly weighted combinations of these signals and moved the paw to different positions under two joint-covariance conditions similar to the original experiments. Our results paralleled those experiments in a number of respects: 1) Many neurons were tuned to paw position relative to the hip under both conditions. 2) The distribution of tuning was strongly bimodal, with most neurons driven by whole-leg flexion or extension. 3) The change in tuning between conditions clustered around zero (median absolute change ~20°). These results indicate that, at least for these constraint conditions, extrinsic-like representation can be achieved simply through musculoskeletal geometry and convergent muscle length inputs. Consequently, they suggest a reinterpretation of the earlier results may be required.NEW & NOTEWORTHY A classic experiment concluding that many dorsal spinocerebellar tract neurons encode paw position rather than joint angles has been cited by many studies as evidence for high-level computation occurring within a single synapse of the sensors. However, our study provides evidence that such a computation is not required to explain the results. Using simulation, we replicated many of the original results with purely random connectivity, suggesting that a reinterpretation of the classic experiment is needed.

A shape deformation algorithm for constrained multidimensional scaling

We present a new Euclidean embedding technique based on volumetric shape registration. Extrinsic representation of the intrinsic geometry of a shape is preferable in various computer graphics applications as it poses only a small degrees of freedom to deal with during processing. A popular Euclidean embedding approach to achieve such a representation is multidimensional scaling (MDS), which, however, distorts the original geometric details drastically. Our method introduces a constraint on the original MDS formulation in order to preserve the initial geometric details while the input shape is pulled towards its MDS pose using the perfectly accurate bijection in between. The regularizer of this registration framework is chosen in such a way that the system supports large deformations yet remains fast. Consequently, we produce a detail-preserving MDS pose in 90s for a 53K-vertex high-resolution mesh on a modest computer. We can also add pairwise point constraints on the deforming shape without any additional cost. Detail-preserving MDS is superior for non-rigid shape retrieval and useful for shape segmentation, as demonstrated.

The duality of temporal encoding - the intrinsic and extrinsic representation of time.

While time is well acknowledged for having a fundamental part in our perception, questions on how it is represented are still matters of great debate. One of the main issues in question is whether time is represented intrinsically at the neural level, or is it represented within dedicated brain regions. We used an fMRI block design to test if we can impose covert encoding of temporal features of faces and natural scenes stimuli within category selective neural populations by exposing subjects to four types of temporal variance, ranging from 0% up to 50% variance. We found a gradual increase in neural activation associated with the gradual increase in temporal variance within category selective areas. A second level analysis showed the same pattern of activations within known brain regions associated with time representation, such as the Cerebellum, the Caudate, and the Thalamus. We concluded that temporal features are integral to perception and are simultaneously represented within category selective regions and globally within dedicated regions. Our second conclusion, drown from our covert procedure, is that time encoding, at its basic level, is an automated process that does not require attention allocated toward the temporal features nor does it require dedicated resources.

The coordinate system for force control.

The primary objective of this study was to establish the coordinate frame for force control by observing how parameters of force that are not explicitly specified by a motor task vary across the workspace. We asked subjects to apply a force of a specific magnitude with their hand. Subjects could complete the task by applying forces in any direction of their choice in the transverse plane. They were tested with the arm in seven different configurations. To estimate whether contact forces are represented in extrinsic or intrinsic coordinates, we applied the parallel transport method of differential geometry to the net joint torques applied during the task. This approach allowed us to compare the force variability observed at different arm configurations with the force variability that would be expected if the control system were applying an invariant pattern of joint torques at the tested configurations. The results indicate that for the majority of the subjects, the predominant pattern was consistent with an invariant representation in joint coordinates. However, two out of eleven subjects also demonstrated a preference for extrinsic representation. These findings suggest that the central nervous system can represent contact forces in both coordinate frames, with a prevalence toward intrinsic representations.

Population Coding and the Labeling Problem: Extrinsic Versus Intrinsic Representations

Current population coding methods, including weighted averaging and Bayesian estimation, are based on extrinsic representations. These require that neurons be labeled with response parameters, such as tuning curve peaks or noise distributions, which are tied to some external, world-based metric scale. Firing rates alone, without this external labeling, are insufficient to represent a variable. However, the extrinsic approach does not explain how such neural labeling is implemented. A radically different and perhaps more physiological approach is based on intrinsic representations, which have access only to firing rates. Because neurons are unlabeled, intrinsic coding represents relative, rather than absolute, values of a variable. We show that intrinsic coding has representational advantages, including invariance, categorization, and discrimination, and in certain situations it may also recover absolute stimulus values.

생리적 내재반응 및 얼굴표정 간 확률 관계 모델 기반의 감정인식 시스템에 관한 연구

The current vision-based approaches for emotion recognition, such as facial expression analysis, have many technical limitations in real circumstances, and are not suitable for applications that use them solely in practical environments. In this paper, we propose an approach for emotion recognition by combining extrinsic representations and intrinsic activities among the natural responses of humans which are given specific imuli for inducing emotional states. The intrinsic activities can be used to compensate the uncertainty of extrinsic representations of emotional states. This combination is done by using PRMs (Probabilistic Relational Models) which are extent version of bayesian networks and are learned by greedy-search algorithms and expectation-maximization algorithms. Previous research of facial expression-related extrinsic emotion features and physiological signal-based intrinsic emotion features are combined into the attributes of the PRMs in the emotion recognition domain. The maximum likelihood estimation with the given dependency structure and estimated parameter set is used to classify the label of the target emotional states.

Motor memory is encoded as a gain-field combination of intrinsic and extrinsic action representations.

Actions can be planned in either an intrinsic (body-based) reference frame or an extrinsic (world-based) frame, and understanding how the internal representations associated with these frames contribute to the learning of motor actions is a key issue in motor control. We studied the internal representation of this learning in human subjects by analyzing generalization patterns across an array of different movement directions and workspaces after training a visuomotor rotation in a single movement direction in one workspace. This provided a dense sampling of the generalization function across intrinsic and extrinsic reference frames, which allowed us to dissociate intrinsic and extrinsic representations and determine the manner in which they contributed to the motor memory for a trained action. A first experiment showed that the generalization pattern reflected a memory that was intermediate between intrinsic and extrinsic representations. A second experiment showed that this intermediate representation could not arise from separate intrinsic and extrinsic learning. Instead, we find that the representation of learning is based on a gain-field combination of local representations in intrinsic and extrinsic coordinates. This gain-field representation generalizes between actions by effectively computing similarity based on the (Mahalanobis) distance across intrinsic and extrinsic coordinates and is in line with neural recordings showing mixed intrinsic-extrinsic representations in motor and parietal cortices.

Cortical involvement in the recruitment of wrist muscles.

In executing a voluntary movement, one is faced with the problem of translating a specification of the movement in task space (e.g., a visual goal) into a muscle-recruitment pattern. Among many brain regions, the primary motor cortex (MI) plays a prominent role in the specification of movements. In what coordinate frame MI represents movement has been a topic of considerable debate. In a two-dimensional wrist step-tracking experiment, Kakei et al. described some MI cells as encoding movement in a muscle-coordinate frame and other cells as encoding movement in an extrinsic-coordinate frame. This result was interpreted as evidence for a cascade of transformations within MI from an extrinsic representation of movement to a muscle-like representation. However, we present a model that demonstrates that, given a realistic extrinsic-like representation of movement, a simple linear network is capable of representing the transformation from an extrinsic space to the muscle-recruitment patterns implementing the movements on which Kakei et al. focused. This suggests that cells exhibiting extrinsic-like qualities can be involved in the direct recruitment of spinal motor neurons. These results call into question models that presume a serial cascade of transformations terminating with MI pyramidal tract neurons that vary their activation exclusively with muscle activity. Further analysis of the model shows that the correlation between the activity of an MI neuron and a muscle does not predict the strength of the connection between the MI neuron and muscle. This result cautions against the use of correlation methods as a measure of cellular connectivity.

Learned dynamics of reaching movements generalize from dominant to nondominant arm.

Accurate performance of reaching movements depends on adaptable neural circuitry that learns to predict forces and compensate for limb dynamics. In earlier experiments, we quantified generalization from training at one arm position to another position. The generalization patterns suggested that neural elements learning to predict forces coded a limb's state in an intrinsic, muscle-like coordinate system. Here, we test the sensitivity of these elements to the other arm by quantifying inter-arm generalization. We considered two possible coordinate systems: an intrinsic (joint) representation should generalize with mirror symmetry reflecting the joint's symmetry and an extrinsic representation should preserve the task's structure in extrinsic coordinates. Both coordinate systems of generalization were compared with a naïve control group. We tested transfer in right-handed subjects both from dominant to nondominant arm (D-->ND) and vice versa (ND-->D). This led to a 2 x 3 experimental design matrix: transfer direction (D-->ND/ND-->D) by coordinate system (extrinsic, intrinsic, control). Generalization occurred only from dominant to nondominant arm and only in extrinsic coordinates. To assess the dependence of generalization on callosal inter-hemispheric communication, we tested commissurotomy patient JW. JW showed generalization from dominant to nondominant arm in extrinsic coordinates. The results suggest that when the dominant right arm is used in learning dynamics, the information could be represented in the left hemisphere with neural elements tuned to both the right arm and the left arm. In contrast, learning with the nondominant arm seems to rely on the elements in the nondominant hemisphere tuned only to movements of that arm.