Rich Representation Research Articles

Assessing the Impact of Geopolitical Risk on Longevity Bond Pricing: Insights from Bayesian Multivariate Regression

AbstractThis paper investigates the multivariate pricing of coupon longevity bonds (CLBs) using the Fama–French–Lee–Carter (FF–LC) five-vector model in the framework of Bayesian integrated nested Laplace approximation (INLA) in the presence of geopolitical risk (GPR). The variance-covariance and correlation matrices are utilized to capture the interdependence between factors. We prove the generalization of multivariate Bayesian INLA with the basic probability assignment which is utilized as a posterior uncertainty belief associated with the GPR uncertainty category (a rich representation of GPR uncertainty) that is an element of the frame of discernment in the CLB posterior estimation. INLA Bayesian principal component analysis (INLA-BPCA) is applied to the model prediction parameters generating a multivariate normally distributed posterior. The deviance information criterion (DIC) assesses optimal factor selection. The results show that the BPCA posterior gains a feature that allows for a balance between the goodness-of-fit and complexity in hierarchical model selection by incorporating the retained principal components (or the effective number of parameters) from the DIC formula. Furthermore, it is also evident in our results, that the DIC outperforms the Bayesian information criterion (BIC), and Watanabe–Akaike information criterion (WAIC). The DIC is more suitable for Bayesian-based parametric models with high complexity. Lastly, the INLA-BPCA-DIC is applied to select the best longevity factors that yield a low longevity price of risk for insurers and practitioners, to attenuate the risks associated with investing in CLBs in the presence of geopolitical uncertainty shocks.

Language Model Crossover: Variation through Few-Shot Prompting

This article pursues the insight that language models naturally enable an intelligent variation operator similar in spirit to evolutionary crossover. In particular, language models of sufficient scale demonstrate in-context learning, i.e., they can learn from associations between a small number of input patterns to generate outputs incorporating such associations (also called few-shot prompting). This ability can be leveraged to form a simple but powerful variation operator, i.e., to prompt a language model with a few text-based genotypes (such as code, plain-text sentences, or equations), and to parse its corresponding output as those genotypes’ offspring. The promise of such language model crossover (which is simple to implement and can leverage many different open source language models) is that it enables a simple mechanism to evolve semantically rich text representations (with few domain-specific tweaks), and naturally benefits from current progress in language models. Experiments in this article highlight the versatility of language-model crossover, through evolving binary bit-strings, sentences, equations, text-to-image prompts, and Python code. The conclusion is that language model crossover is a flexible and effective method for evolving genomes representable as text.

One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems

Sequential recommendation systems aim to predict users’ next likely interaction based on their history. However, these systems face data sparsity and cold-start problems. Utilizing data from other domains, known as multi-domain methods, is useful for alleviating these problems. However, traditional multi-domain methods rely on meaningless ID-based item representation, which makes it difficult to align items with similar meanings from different domains, yielding sup-optimal knowledge transfer. This paper introduces LLM-Rec , a framework that utilizes pre-trained large language models (LLMs) for domain-agnostic recommendation. Specifically, we mix user's behaviors from multiple domains and concatenate item titles into a sentence, then use LLMs for generating user and item representations. By mixing behaviors across different domains, we can exploit the knowledge encoded in LLMs to bridge the semantic across over multi-domain behaviors, thus obtaining semantically rich representations and improving performance in all domains. Furthermore, we explore the underlying reasons why LLMs are effective and investigate whether LLMs can understand the semantic correlations as the recommendation model, and if advanced techniques like scaling laws in NLP also work in recommendations. We conduct extensive experiments with LLMs ranging from 40M to 6.7B to answer the above questions and to verify the effectiveness of LLM-Rec in multi-domain recommendation \({}^{\ddagger}\) .

SC-VAE: Sparse coding-based variational autoencoder with learned ISTA

Learning rich data representations from unlabeled data is a key challenge towards applying deep learning algorithms in downstream tasks. Several variants of variational autoencoders (VAEs) have been proposed to learn compact data representations by encoding high-dimensional data in a lower dimensional space. Two main classes of VAEs methods may be distinguished depending on the characteristics of the meta-priors that are enforced in the representation learning step. The first class of methods derives a continuous encoding by assuming a static prior distribution in the latent space. The second class of methods learns instead a discrete latent representation using vector quantization (VQ) along with a codebook. However, both classes of methods suffer from certain challenges, which may lead to suboptimal image reconstruction results. The first class suffers from posterior collapse, whereas the second class suffers from codebook collapse. To address these challenges, we introduce a new VAE variant, termed sparse coding-based VAE with learned ISTA (SC-VAE), which integrates sparse coding within variational autoencoder framework. The proposed method learns sparse data representations that consist of a linear combination of a small number of predetermined orthogonal atoms. The sparse coding problem is solved using a learnable version of the iterative shrinkage thresholding algorithm (ISTA). Experiments on two image datasets demonstrate that our model achieves improved image reconstruction results compared to state-of-the-art methods. Moreover, we demonstrate that the use of learned sparse code vectors allows us to perform downstream tasks like image generation and unsupervised image segmentation through clustering image patches. The code is available at https://github.com/sotiraslab/SC-VAE.

Inter- and Intrahemispheric Sources of Vestibular Signals to V1.

Head movements are sensed by the vestibular organs. Unlike classical senses, signals from vestibular organs are not conveyed to a dedicated cortical area but are broadcast throughout the cortex. Surprisingly, the routes taken by vestibular signals to reach the cortex are still largely uncharted. Here we show that the primary visual cortex (V1) receives real-time head movement signals - direction, velocity, and acceleration - from the ipsilateral pulvinar and contralateral visual cortex. The ipsilateral pulvinar provides the main head movement signal, with a bias toward contraversive movements (e.g. clockwise movements in left V1). Conversely, the contralateral visual cortex provides head movement signals during ipsiversive movements. Crucially, head movement variables encoded in V1 are already encoded in the pulvinar, suggesting that those variables are computed subcortically. Thus, the convergence of inter- and intrahemispheric signals endows V1 with a rich representation of the animal's head movements.

Deep reinforcement learning-based spatio-temporal graph neural network for solving job shop scheduling problem

The job shop scheduling problem (JSSP) is a well-known NP-hard combinatorial optimization problem that focuses on assigning tasks to limited resources while adhering to certain constraints. Currently, deep reinforcement learning (DRL)-based solutions are being widely used to solve the JSSP by defining the problem structure on disjunctive graphs. Some of the proposed approaches attempt to leverage the structural information of the JSSP to capture the dynamics of the environment without considering the time dependency within the JSSP. However, learning graph representations only from the structural relationship of nodes results in a weak and incomplete representation of these graphs which does not provide an expressive representation of the dynamics in the environment. In this study, unlike existing frameworks, we defined the JSSP as a dynamic graph to explicitly consider the time-varying aspect of the JSSP environment. To this end, we propose a novel DRL framework that captures both the spatial and temporal attributes of the JSSP to construct rich and complete graph representations. Our DRL framework introduces a novel attentive graph isomorphism network (Attentive-GIN)-based spatial block to learn the structural relationship and a temporal block to capture the time dependency. Additionally, we designed a gated fusion block that selectively combines the learned representations from the two blocks. We trained the model using the proximal policy optimization algorithm of reinforcement learning. Experimental results show that our trained model exhibits significant performance enhancement compared to heuristic dispatching rules and learning-based solutions for both randomly generated datasets and public benchmarks.

MaskDGNets: Masked-attention guided dynamic graph aggregation network for event extraction.

Considering that the traditional deep learning event extraction method ignores the correlation between word features and sequence information, it cannot fully explore the hidden associations between events and events and between events and primary attributes. To solve these problems, we developed a new framework for event extraction called the masked attention-guided dynamic graph aggregation network. On the one hand, to obtain effective word representation and sequence representation, an interaction and complementary relationship are established between word vectors and character vectors. At the same time, a squeeze layer is introduced in the bidirectional independent recurrent unit to model the sentence sequence from both positive and negative directions while retaining the local spatial details to the maximum extent and establishing practical long-term dependencies and rich global context representations. On the other hand, the designed masked attention mechanism can effectively balance the word vector features and sequence semantics and refine these features. The designed dynamic graph aggregation module establishes effective connections between events and events, and between events and essential attributes, strengthens the interactivity and association between them, and realizes feature transfer and aggregation on graph nodes in the neighborhood through dynamic strategies to improve the performance of event extraction. We designed a reconstructed weighted loss function to supervise and adjust each module individually to ensure the optimal feature representation. Finally, the proposed MaskDGNets framework is evaluated on two baseline datasets, DuEE and CCKS2020. It demonstrates its robustness and event extraction performance, with F1 of 81.443% and 87.382%, respectively.

MST-m6A: A Novel Multi-Scale Transformer-based Framework for Accurate Prediction of m6A Modification Sites Across Diverse Cellular Contexts

N6-methyladenosine (m6A) modification, a prevalent epigenetic mark in eukaryotic cells, is crucial in regulating gene expression and RNA metabolism. Accurately identifying m6A modification sites is essential for understanding their functions within biological processes and the intricate mechanisms that regulate them. Recent advances in high-throughput sequencing technologies have enabled the generation of extensive datasets characterizing m6A modification sites at single-nucleotide resolution, leading to the development of computational methods for identifying m6A RNA modification sites. However, most current methods focus on specific cell lines, limiting their generalizability and practical application across diverse biological contexts. To address the limitation, we propose MST-m6A, a novel approach for identifying m6A modification sites with higher accuracy across various cell lines and tissues. MST-m6A utilizes a multi-scale transformer-based architecture, employing dual k-mer tokenization to capture rich feature representations and global contextual information from RNA sequences at multiple levels of granularity. These representations are then effectively combined using a channel fusion mechanism and further processed by a convolutional neural network to enhance prediction accuracy. Rigorous validation demonstrates that MST-m6A significantly outperforms conventional machine learning models, deep learning models, and state-of-the-art predictors. We anticipate that the high precision and cross-cell-type adaptability of MST-m6A will provide valuable insights into m6A biology and facilitate advancements in related fields. The proposed approach is available at https://github.com/cbbl-skku-org/MST-m6A/ for prediction and reproducibility purposes.

EEG-Oriented Self-Supervised Learning With Triple Information Pathways Network.

Recently, deep learning-based electroencephalogram (EEG) analysis and decoding have attracted widespread attention for monitoring the clinical condition of users and identifying their intention/emotion. Nevertheless, the existing methods generally model EEG signals with limited viewpoints or restricted concerns about the characteristics of the EEG signals, and thus represent complex spectro-/spatiotemporal patterns and suffer from high variability. In this work, we propose the novel EEG-oriented self-supervised learning methods and a novel deep architecture to learn rich representation, including information about the diverse spectral characteristics of neural oscillations, the spatial properties of electrode sensor distribution, and the temporal patterns of both the global and local viewpoints. Along with the proposed self-supervision strategies and deep architectures, we devise a feature normalization strategy to resolve the intra-/inter-subject variability problem. We demonstrate the validity of our proposed deep learning framework on the four publicly available datasets by conducting comparisons with the state of the art baselines. It is also noteworthy that we exploit the same network architecture for the four different EEG paradigms and outperform the comparison methods, thereby meeting the challenge of the task-dependent network architecture engineering in EEG-based applications.

ChartLine: Automatic Detection and Tracing of Curves in Scientific Line Charts Using Spatial-Sequence Feature Pyramid Network.

Line charts are prevalent in scientific documents and commercial data visualization, serving as essential tools for conveying data trends. Automatic detection and tracing of line paths in these charts is crucial for downstream tasks such as data extraction, chart quality assessment, plagiarism detection, and visual question answering. However, line graphs present unique challenges due to their complex backgrounds and diverse curve styles, including solid, dashed, and dotted lines. Existing curve detection algorithms struggle to address these challenges effectively. In this paper, we propose ChartLine, a novel network designed for detecting and tracing curves in line graphs. Our approach integrates a Spatial-Sequence Attention Feature Pyramid Network (SSA-FPN) in both the encoder and decoder to capture rich hierarchical representations of curve structures and boundary features. The model incorporates a Spatial-Sequence Fusion (SSF) module and a Channel Multi-Head Attention (CMA) module to enhance intra-class consistency and inter-class distinction. We evaluate ChartLine on four line chart datasets and compare its performance against state-of-the-art curve detection, edge detection, and semantic segmentation methods. Extensive experiments demonstrate that our method significantly outperforms existing algorithms, achieving an F-measure of 94% on a synthetic dataset.

Improving information theory of context-aware phrase embeddings in HR domain

The object of the study is context-aware phrase representations. The growing need to automate candidate recruitment and job recommendation processes has paved the way for the utilization of text embeddings. These embeddings involve translating the semantic essence of text into a continuous, high-dimensional vector space. By learning context-aware rich and meaningful representations of phrases within the human resource domain, the efficacy of similarity searches and matching procedures is enhanced, which contributes to a more streamlined and effective recruitment process. However, existing approaches do not take into account the context when modeling phrases. This necessitates the improvement of information technology analysis in this area. In this paper, it is proposed to mark the beginning and end of phrases in the text using special tokens. This made it possible to reduce the requirements for computing power by calculating all phrase representations present in the text simultaneously. The effectiveness of the improvement was tested on a new dataset to compare and evaluate the models in the task of modeling phrases in the field of human resources management. The proposed approach to modeling phrase representations with regard to context in the field of human resources management leads to an improvement in computational efficiency by up to 50 % and an increase in accuracy by up to 10 %. The architecture of the machine learning model for creating context-aware phrase representations is developed, which is characterized by the presence of blocks for taking into account phrase boundaries. Experiments and comparisons with existing approaches have confirmed the effectiveness of the proposed solution. In practice, the proposed information analysis technology can be used to automate the process of identifying and normalizing candidates' skills in online recruiting

AlphaFold Meets De Novo Drug Design: Leveraging Structural Protein Information in Multitarget Molecular Generative Models.

Recent advancements in deep learning and generative models have significantly expanded the applications of virtual screening for drug-like compounds. Here, we introduce a multitarget transformer model, PCMol, that leverages the latent protein embeddings derived from AlphaFold2 as a means of conditioning a de novo generative model on different targets. Incorporating rich protein representations allows the model to capture their structural relationships, enabling the chemical space interpolation of active compounds and target-side generalization to new proteins based on embedding similarities. In this work, we benchmark against other existing target-conditioned transformer models to illustrate the validity of using AlphaFold protein representations over raw amino acid sequences. We show that low-dimensional projections of these protein embeddings cluster appropriately based on target families and that model performance declines when these representations are intentionally corrupted. We also show that the PCMol model generates diverse, potentially active molecules for a wide array of proteins, including those with sparse ligand bioactivity data. The generated compounds display higher similarity known active ligands of held-out targets and have comparable molecular docking scores while maintaining novelty. Additionally, we demonstrate the important role of data augmentation in bolstering the performance of generative models in low-data regimes. Software package and AlphaFold protein embeddings are freely available at https://github.com/CDDLeiden/PCMol.

Self-Supervised Image Aesthetic Assessment Based on Transformer

Visual aesthetics has always been an important area of computational vision, and researchers have continued exploring it. To further improve the performance of the image aesthetic evaluation task, we introduce a Transformer into the image aesthetic evaluation task. This paper pioneers a novel self-supervised image aesthetic evaluation model founded upon Transformers. Meanwhile, we expand the pretext task to capture rich visual representations, adding a branch for inpainting the masked images in parallel with the tasks related to aesthetic quality degradation operations. Our model’s refinement employs the innovative uncertainty weighting method, seamlessly amalgamating three distinct losses into a unified objective. On the AVA dataset, our approach surpasses the efficacy of prevailing self-supervised image aesthetic assessment methods. Remarkably, we attain results approaching those of supervised methods, even while operating with a limited dataset. On the AADB dataset, our approach improves the aesthetic binary classification accuracy by roughly 16% compared to other self-supervised image aesthetic assessment methods and improves the prediction of aesthetic attributes.

JobFormer: Skill-Aware Job Recommendation with Semantic-Enhanced Transformer

Job recommendation aims to provide potential talents with suitable job descriptions (JDs) consistent with their career trajectory, which plays an essential role in proactive talent recruitment. In real-world management scenarios, the available JD-user records always consist of JDs, user profiles, and click data, in which the user profiles are typically summarized as the user's skill distribution for privacy reasons. Although existing sophisticated recommendation methods can be directly employed, effective recommendation still has challenges considering the information deficit of JD itself and the natural heterogeneous gap between JD and user profile. To address these challenges, we proposed a novel skill-aware recommendation model based on the designed semantic-enhanced Transformer to parse JDs and complete personalized job recommendation. Specifically, we first model the relative items of each JD and then adopt an encoder with the local-global attention mechanism to better mine the intra-job and inter-job dependencies from JD tuples. Moreover, we adopt a two-stage learning strategy for skill-aware recommendation, in which we utilize the skill distribution to guide JD representation learning in the recall stage, and then combine the user profiles for final prediction in the ranking stage. Consequently, we can embed rich contextual semantic representations for learning JDs, while skill-aware recommendation provides effective JD-user joint representation for click-through rate (CTR) prediction. To validate the superior performance of our method for job recommendation, we present a thorough empirical analysis of large-scale real-world and public datasets to demonstrate its effectiveness and interpretability.

Studio as a Catalyst for Incremental and Ambitious Teacher Learning

This article explores how the practice-focused Studio professional learning (PL) model can provide incremental and ambitious teacher learning opportunities. We argue that when the model’s structures and practices are grounded in ambitious and equitable teaching, they catalyze incremental teacher learning. Studio, like lesson study, supports teachers in considering the entailments of lessons, focusing on the live shared enactment to strengthen teaching and learning through collaborative analysis and reflection. To build our argument, we drew from two Studio projects that had shared structures of cycles of learning and routines, as well as shared practices of using rich representations and collective interpretations of teaching. While both projects’ structures and practices take up ambitious and equitable teaching, they use different routines and attend to different features of equitable teaching. Building on a history of PL models, such as lesson study, which use similar structures and practices as powerful catalysts of teacher learning, we argue that Studio’s structures and practices can catalyze teachers’ incremental learning of ambitious and equitable teaching. We discuss the implications for future research based on this argument and for those leading PL.

Evolutionary-Scale Enzymology Enables Biochemical Constant Prediction Across a Multi-Peaked Catalytic Landscape.

Quantitatively mapping enzyme sequence-catalysis landscapes remains a critical challenge in understanding enzyme function, evolution, and design. Here, we expand an emerging microfluidic platform to measure catalytic constants- k cat and K M -for hundreds of diverse naturally occurring sequences and mutants of the model enzyme Adenylate Kinase (ADK). This enables us to dissect the sequence-catalysis landscape's topology, navigability, and mechanistic underpinnings, revealing distinct catalytic peaks organized by structural motifs. These results challenge long-standing hypotheses in enzyme adaptation, demonstrating that thermophilic enzymes are not slower than their mesophilic counterparts. Combining the rich representations of protein sequences provided by deep-learning models with our custom high-throughput kinetic data yields semi-supervised models that significantly outperform existing models at predicting catalytic parameters of naturally occurring ADK sequences. Our work demonstrates a promising strategy for dissecting sequence-catalysis landscapes across enzymatic evolution and building family-specific models capable of accurately predicting catalytic constants, opening new avenues for enzyme engineering and functional prediction.

Problem of Surface Waves on Water in Higher School Laboratory Workshop

Traditionally, in lecture courses on General Physics, the example of waves propagating on the water surface is used for the qualitative description of wave phenomena. At the same time, in corresponding laboratory courses, there are almost no quantitative tasks on the same topic, the purpose of which would be to measure the surface wave characteristics. The reason for this difference must be associated with the complexity of the physical mechanism of generation and propagation of waves on the water surface. In this work, a laboratory problem has been developed to measure the group velocity of circular waves traveling on the water surface. It uses a standard laboratory water tank equipped with a surface wave generator. The proposed experiment offers an opportunity for quantitative measurement of surface wave characteristics, a dispute of only qualitative observations of wave phenomena commonly found in current curricula, due to the introduction of two improvements. First, to produce the water waves’ rich in-contrast representations on a screen, the optimal generator amplitude is specially set for each experiment; and second, to measure the group wavelength, the device magnification is preliminary calculated based on the laws of geometric optics. The tests carried out have shown that the relative accuracy of these measurements is quite acceptable: within a few percent. The introduction of the problem of surface waves on water into the General Physics laboratory workshop for engineering specialties of technical universities will contribute to improving the quality of mastering the subject by students.

Region-aware image-based human action retrieval with transformers

Human action understanding is a fundamental and challenging task in computer vision. Although there exists tremendous research on this area, most works focus on action recognition, while action retrieval has received less attention. In this paper, we focus on the neglected but important task of image-based action retrieval which aims to find images that depict the same action as a query image. We establish benchmarks for this task and set up important baseline methods for fair comparison. We present a Transformer-based model that learns rich action representations from three aspects: the anchored person, contextual regions, and the global image. A fusion transformer is designed to model the relationships among different features and effectively fuse them into an action representation. Experiments on both the Stanford-40 and PASCAL VOC 2012 Action datasets show that the proposed method significantly outperforms previous approaches for image-based action retrieval.

Intentional looks facilitate faster responding in observers

Humans construct rich representations of other people’s mental states. Here we investigated how intentionality in eye gaze affected perception and responses to gaze. Observers viewed videos of human gazers looking left or right. Unbeknownst to the observers, the gazers could either choose where to look (self-chosen gaze) or were explicitly instructed where to look (computer-instructed gaze). In Experiment 1, observers reported the direction of the gazer’s upcoming look before the eye movement was initiated. Faster responses were found for self-chosen relative to computer-instructed gaze. In Experiments 2 and 3, observers responded by reporting the location of a peripheral target that appeared at the gazed-at or not gazed-at location. Faster responses were found for gazed-at relative to not gazed-at targets and at longer cue-target intervals for self-chosen relative to computer-instructed gaze. The examination of the eye movement kinematics indicated that self-chosen gaze shifts were marked by a larger magnitude of motion within the eye region prior to the eye movement occurring relative to computer-instructed ones. Thus, perceived intentionality in eye gaze facilitates responses in observers with the information about mental states communicated via subtle properties of eye motion.

Efficient Neural Decoding Based on Multimodal Training.

Neural decoding methods are often limited by the performance of brain encoders, which map complex brain signals into a latent representation space of perception information. These brain encoders are constrained by the limited amount of paired brain and stimuli data available for training, making it challenging to learn rich neural representations. To address this limitation, we present a novel multimodal training approach using paired image and functional magnetic resonance imaging (fMRI) data to establish a brain masked autoencoder that learns the interactions between images and brain activities. Subsequently, we employ a diffusion model conditioned on brain data to decode realistic images. Our method achieves high-quality decoding results in semantic contents and low-level visual attributes, outperforming previous methods both qualitatively and quantitatively, while maintaining computational efficiency. Additionally, our method is applied to decode artificial patterns across region of interests (ROIs) to explore their functional properties. We not only validate existing knowledge concerning ROIs but also unveil new insights, such as the synergy between early visual cortex and higher-level scene ROIs, as well as the competition within the higher-level scene ROIs. These findings provide valuable insights for future directions in the field of neural decoding.