Exploiting Positional Information for Session-Based Recommendation

For present e-commerce platforms, it is important to accurately predict users’ preference for a timely next-item recommendation. To achieve this goal, session-based recommender systems are developed, which are based on a sequence of the most recent user-item interactions to avoid the influence raised from outdated historical records. Although a session can usually reflect a user’s current preference, a local shift of the user’s intention within the session may still exist. Specifically, the interactions that take place in the early positions within a session generally indicate the user’s initial intention, while later interactions are more likely to represent the latest intention. Such positional information has been rarely considered in existing methods, which restricts their ability to capture the significance of interactions at different positions. To thoroughly exploit the positional information within a session, a theoretical framework is developed in this paper to provide an in-depth analysis of the positional information. We formally define the properties of forward-awareness and backward-awareness to evaluate the ability of positional encoding schemes in capturing the initial and the latest intention. According to our analysis, existing positional encoding schemes are generally forward-aware only, which can hardly represent the dynamics of the intention in a session. To enhance the positional encoding scheme for the session-based recommendation, a dual positional encoding (DPE) is proposed to account for both forward-awareness and backward-awareness . Based on DPE, we propose a novel Positional Recommender (PosRec) model with a well-designed Position-aware Gated Graph Neural Network module to fully exploit the positional information for session-based recommendation tasks. Extensive experiments are conducted on two e-commerce benchmark datasets, Yoochoose and Diginetica and the experimental results show the superiority of the PosRec by comparing it with the state-of-the-art session-based recommender models.

Growth orientations, rather than heterogeneous growth rates, dominate jaw joint morphogenesis in the larval zebrafish

In early limb embryogenesis, synovial joints acquire specific shapes which determine joint motion and function. The process by which the opposing cartilaginous joint surfaces are moulded into reciprocal and interlocking shapes, called joint morphogenesis, is one of the least understood aspect of joint formation and the cell-level dynamics underlying it are yet to be unravelled. In this research, we quantified key cellular dynamics involved in growth and morphogenesis of the zebrafish jaw joint and synthesised them in a predictive computational simulation of joint development. Cells in larval zebrafish jaw joints labelled with cartilage markers were tracked over a forty-eight hour time window using confocal imaging. Changes in distance and angle between adjacent cell centroids resulting from cell rearrangement, volume expansion and extracellular matrix (ECM) deposition were measured and used to calculate the rate and direction of local tissue deformations. We observed spatially and temporally heterogeneous growth patterns with marked anisotropy over the developmental period assessed. There was notably elevated growth at the level of the retroarticular process of the Meckel's cartilage, a feature known to undergo pronounced shape changes during zebrafish development. Analysis of cell dynamics indicated a dominant role for cell volume expansion in growth, with minor influences from ECM volume increases and cell intercalation. Cell proliferation in the joint was minimal over the timeframe of interest. Synthesising the dynamic cell data into a finite element model of jaw joint development resulted in accurate shape predictions. Our biofidelic computational simulation demonstrated that zebrafish jaw joint growth can be reasonably approximated based on cell positional information over time, where cell positional information derives mainly from cell orientation and cell volume expansion. By modifying the input parameters of the simulation, we were able to assess the relative contributions of heterogeneous growth rates and of growth orientation. The use of uniform rather than heterogeneous growth rates only minorly impacted the shape predictions whereas isotropic growth fields resulted in altered shape predictions. The simulation results suggest that growth anisotropy is the dominant influence on joint growth and morphogenesis. This study addresses the gap of the cellular processes underlying joint morphogenesis, with implications for understanding the aetiology of developmental joint disorders such as developmental dysplasia of the hip and arthrogryposis.

Regenerating aggregates of hydra display unique cytoskeletal organisation that is absent in a regeneration-deficient strain

Hydra has the unique ability to regenerate from aggregates of dissociated single cells that lack positional information. We compared two strains of hydra, a strain of hydra that was capable of regenerating from aggregates and a strain of hydra that was deficient in this type of regeneration. We observed unique actin cytoskeletal arrangements that were present in the regenerates of regeneration-competent strain but not in the regeneration-deficient strain. Concomitantly, the regeneration-deficient strain failed to organise the extracellular cytoskeleton of laminin and collagen between ectodermal and endodermal epithelial cells. These interesting preliminary observations highlight the importance of the cytoskeletal organisation in regeneration of hydra and suggest that regeneration from the aggregates of dissociated cells through de novo patterning requires correct structural organisation of cytoskeletal elements.

Magnetic maps in animal navigation

In addition to providing animals with a source of directional or ‘compass’ information, Earth’s magnetic field also provides a potential source of positional or ‘map’ information that animals might exploit to assess location. In less than a generation, the idea that animals use Earth’s magnetic field as a kind of map has gone from a contentious hypothesis to a well-established tenet of animal navigation. Diverse animals ranging from lobsters to birds are now known to use magnetic positional information for a variety of purposes, including staying on track along migratory pathways, adjusting food intake at appropriate points in a migration, remaining within a suitable oceanic region, and navigating toward specific goals. Recent findings also indicate that sea turtles, salmon, and at least some birds imprint on the magnetic field of their natal area when young and use this information to facilitate return as adults, a process that may underlie long-distance natal homing (a.k.a. natal philopatry) in many species. Despite recent progress, much remains to be learned about the organization of magnetic maps, how they develop, and how animals use them in navigation.

How to make a multi-billion dollar thermometer

The United States’ Global Positioning System (GPS), and similar geolocation systems such as Galileo, GLONASS, and Beidou are used by people all over the globe. Modern receivers of these global navigation satellite systems can track multiple satellites from different constellations. Casual, non-technical users are probably aware that the positional information provided is typically accurate to within a few meters. We could expect physics students to infer that, because these systems rely on the travel time of radio signals, this implies time measurement accuracy on the scale of tens of nanoseconds. This feature has led to GPS-enabled Internet time servers providing stratum 1 accuracy for under $1000. In this paper, we will show that we can couple a GPS unit to a field programmable gate array (FPGA) to determine the temperature in a room. The more serious application of this GPS-FPGA pairing is to provide precise time-stamping of events, thereby synchronizing data collection between stations across a room or across the globe.

Axonal Projection Patterns of the Dorsal Interneuron Populations in the Embryonic Hindbrain

Unraveling the inner workings of neural circuits entails understanding the cellular origin and axonal pathfinding of various neuronal groups during development. In the embryonic hindbrain, different subtypes of dorsal interneurons (dINs) evolve along the dorsal-ventral (DV) axis of rhombomeres and are imperative for the assembly of central brainstem circuits. dINs are divided into two classes, class A and class B, each containing four neuronal subgroups (dA1-4 and dB1-4) that are born in well-defined DV positions. While all interneurons belonging to class A express the transcription factor Olig3 and become excitatory, all class B interneurons express the transcription factor Lbx1 but are diverse in their excitatory or inhibitory fate. Moreover, within every class, each interneuron subtype displays its own specification genes and axonal projection patterns which are required to govern the stage-by-stage assembly of their connectivity toward their target sites. Remarkably, despite the similar genetic landmark of each dINs subgroup along the anterior-posterior (AP) axis of the hindbrain, genetic fate maps of some dA/dB neuronal subtypes uncovered their contribution to different nuclei centers in relation to their rhombomeric origin. Thus, DV and AP positional information has to be orchestrated in each dA/dB subpopulation to form distinct neuronal circuits in the hindbrain. Over the span of several decades, different axonal routes have been well-documented to dynamically emerge and grow throughout the hindbrain DV and AP positions. Yet, the genetic link between these distinct axonal bundles and their neuronal origin is not fully clear. In this study, we reviewed the available data regarding the association between the specification of early-born dorsal interneuron subpopulations in the hindbrain and their axonal circuitry development and fate, as well as the present existing knowledge on molecular effectors underlying the process of axonal growth.

Incremental 2D Grid Map Generation from RGB-D Images

In this paper, we extend RGB-D SLAM to address the problem that sparse map-building RGB-D SLAM cannot directly generate maps for indoor navigation and propose a SLAM system for fast generation of indoor planar maps. The system uses RGBD images to generate positional information while converting the corresponding RGBD images into 2D planar lasers for 2D grid navigation map reconstruction of indoor scenes under the condition of limited computational resources, solving the problem that the sparse point cloud maps generated by RGB-D SLAM cannot be directly used for navigation. Meanwhile, the pose information provided by RGB-D SLAM and scan matching respectively is fused to obtain a more accurate and robust pose, which improves the accuracy of map building. Furthermore, we demonstrate the function of the proposed system on the ICL indoor dataset and evaluate the performance of different RGB-D SLAM. The method proposed in this paper can be generalized to RGB-D SLAM algorithms, and the accuracy of map building will be further improved with the development of RGB-D SLAM algorithms.

A Novel 4D Track-before-Detect Approach for Weak Targets Detection in Clutter Regions

A 4D TBD approach is developed here for closely weak extended target tracking and overcoming heterogeneous clutter background and various clutter regions. The 4D measurements in this work are the points containing three positional information in spatial space and corresponding timestamp. The proposed method is mainly designed to address two issues. The first one is the dilemma between the weak target detection and difficult computation originating from the high dimensions of measurement. The second issue is the suppression of inhomogeneous background clutter and various clutter regions. The extension experiment using synthetic data showcases that no false alarm track would be built in the clutter regions, and the detection rate of close targets exceeds 94%. The experiments using real 3D radar also prove that the method works well in tracking closely maneuvering extended targets even if a clutter region exists.

An Attention-based Text Detection and Recognition Method for Terminals of Current Transformer’s Secondary Circuit

Recognizing irregular text in real industrial scenes is a challenging task due to the background clutter, low resolutions or distortions. In this work, an attention-based text detection and recognition method for terminals of current transformer’s secondary circuit is proposed. It consists of three major components: pre-processing, text detection and text recognition. In text recognition module, a novel spatial temporal embedding is designed to better utilize the positional information. During training, the proposed framework only requires sequence-level annotations, instead of extra fine-grained character-level boxes or segmentation masks as in previous work. Despite its simplicity, the proposed method achieves good performance on the dataset collected in actual working scene.