Search

While achieving high prediction accuracy is a fundamental goal in machine learning, an equally important task is finding a small number of features with high explanatory power. One popular selection technique is permutation importance, which assesses a variable's impact by measuring the change in prediction error after permuting the variable. However, this can be problematic due to the need to create artificial data, a problem shared by other methods as well. Another problem is that variable selection methods can be limited by being model-specific. We introduce a new model-independent approach, Variable Priority (VarPro), which works by utilizing rules without the need to generate artificial data or evaluate prediction error. The method is relatively easy to use, requiring only the calculation of sample averages of simple statistics, and can be applied to many data settings, including regression, classification, and survival. We investigate the asymptotic properties of VarPro and show, among other things, that VarPro has a consistent filtering property for noise variables. Empirical studies using synthetic and real-world data show the method achieves a balanced performance and compares favorably to many state-of-the-art procedures currently used for variable selection.

Assigning orders to drivers under localized spatiotemporal context (micro-view order-dispatching) is a major task in Didi, as it influences ride-hailing service experience. Existing industrial solutions mainly follow a two-stage pattern that incorporate heuristic or learning-based algorithms with naive combinatorial methods, tackling the uncertainty of both sides' behaviors, including emerging timings, spatial relationships, and travel duration, etc. In this paper, we propose a one-stage end-to-end reinforcement learning based order-dispatching approach that solves behavior prediction and combinatorial optimization uniformly in a sequential decision-making manner. Specifically, we employ a two-layer Markov Decision Process framework to model this problem, and present \underline{D}eep \underline{D}ouble \underline{S}calable \underline{N}etwork (D2SN), an encoder-decoder structure network to generate order-driver assignments directly and stop assignments accordingly. Besides, by leveraging contextual dynamics, our approach can adapt to the behavioral patterns for better performance. Extensive experiments on Didi's real-world benchmarks justify that the proposed approach significantly outperforms competitive baselines in optimizing matching efficiency and user experience tasks. In addition, we evaluate the deployment outline and discuss the gains and experiences obtained during the deployment tests from the view of large-scale engineering implementation., Comment: 8 pages, 4 figures

This work presents a novel progressive image vectorization technique aimed at generating layered vectors that represent the original image from coarse to fine detail levels. Our approach introduces semantic simplification, which combines Score Distillation Sampling and semantic segmentation to iteratively simplify the input image. Subsequently, our method optimizes the vector layers for each of the progressively simplified images. Our method provides robust optimization, which avoids local minima and enables adjustable detail levels in the final output. The layered, compact vector representation enhances usability for further editing and modification. Comparative analysis with conventional vectorization methods demonstrates our technique's superiority in producing vectors with high visual fidelity, and more importantly, maintaining vector compactness and manageability. The project homepage is https://szuviz.github.io/layered_vectorization/.

Collage techniques are commonly used in visualization to organize a collection of geometric shapes, facilitating the representation of visual features holistically, as seen in word clouds or circular packing diagrams. Typically, packing methods rely on object-space optimization techniques, which often necessitate customizing the optimization process to suit the complexity of geometric primitives and the specific application requirements. In this paper, we introduce a versatile image-space collage technique designed to pack geometric elements into a given shape. Leveraging a differential renderer and image-space losses, our optimization process is highly efficient and can easily accommodate various loss functions. We demonstrate the diverse visual expressiveness of our approach across various visualization applications. The evaluation confirmed the benefits of our method in terms of both visual quality and time performance. The project page is https://szuviz.github.io/pixel-space-collage-technique/.

Information graphics are pivotal in effective information dissemination and storytelling. However, creating such graphics is extremely challenging for non-professionals, since the design process requires multifaceted skills and comprehensive knowledge. Thus, despite the many available authoring tools, a significant gap remains in enabling non-experts to produce compelling information graphics seamlessly, especially from scratch. Recent breakthroughs show that Large Language Models (LLMs), especially when tool-augmented, can autonomously engage with external tools, making them promising candidates for enabling innovative graphic design applications. In this work, we propose a LLM-centric interface with the agent GraphiMind for automatic generation, recommendation, and composition of information graphics design resources, based on user intent expressed through natural language. Our GraphiMind integrates a Textual Conversational Interface, powered by tool-augmented LLM, with a traditional Graphical Manipulation Interface, streamlining the entire design process from raw resource curation to composition and refinement. Extensive evaluations highlight our tool's proficiency in simplifying the design process, opening avenues for its use by non-professional users. Moreover, we spotlight the potential of LLMs in reshaping the domain of information graphics design, offering a blend of automation, versatility, and user-centric interactivity.

Renal pathology, as the gold standard of kidney disease diagnosis, requires doctors to analyze a series of tissue slices stained by H&E staining and special staining like Masson, PASM, and PAS, respectively. These special staining methods are costly, time-consuming, and hard to standardize for wide use especially in primary hospitals. Advances of supervised learning methods have enabled the virtually conversion of H&E images into special staining images, but achieving pixel-to-pixel alignment for training remains challenging. In contrast, unsupervised learning methods regarding different stains as different style transfer domains can utilize unpaired data, but they ignore the spatial inter-domain correlations and thus decrease the trustworthiness of structural details for diagnosis. In this paper, we propose a novel virtual staining framework AGMDT to translate images into other domains by avoiding pixel-level alignment and meanwhile utilizing the correlations among adjacent tissue slices. We first build a high-quality multi-domain renal histological dataset where each specimen case comprises a series of slices stained in various ways. Based on it, the proposed framework AGMDT discovers patch-level aligned pairs across the serial slices of multi-domains through glomerulus detection and bipartite graph matching, and utilizes such correlations to supervise the end-to-end model for multi-domain staining transformation. Experimental results show that the proposed AGMDT achieves a good balance between the precise pixel-level alignment and unpaired domain transfer by exploiting correlations across multi-domain serial pathological slices, and outperforms the state-of-the-art methods in both quantitative measure and morphological details., Comment: BMVC 2023

The ability to precisely edit genomes by deleting or adding genetic information enables the study of biological functions and the building of efficient cell factories. In many unconventional yeasts, such as promising new hosts for cell factory design but also human pathogenic yeasts and food spoilers, this progress has been limited by the fact that most yeasts favor non-homologous end joining (NHEJ) over homologous recombination (HR) as DNA repair mechanism, impairing genetic access to these hosts. In mammalian cells, small molecules that either inhibit proteins involved in NHEJ, enhance protein function in HR, or molecules that arrest the cell cycle in HR-dominant phases are regarded as promising agents for the simple and transient increase of HR-mediated genome editing without the need for a priori host engineering. Only a few of these chemicals have been applied to the engineering of yeast although the targeted proteins are mostly conserved; making chemical agents a yet underexplored area in enhancing yeast engineering. Here, we consolidate knowledge of available small molecules that have been used to improve HR efficiency in mammalian cells and the few ones that have been used in yeast. We include available high throughput (HTP)-compatible NHEJ/HR quantification assays that could be used to screen for and isolate yeast-specific inhibitors.

It is the first step for understanding how RNA structure folds from base sequences that to know how its secondary structure is formed. Traditional energy-based algorithms are short of precision, particularly for non-nested sequences, while learning-based algorithms face challenges in obtaining high-quality training data. Recently, quantum annealer has rapidly predicted the folding of the secondary structure, highlighting that quantum computing is a promising solution to this problem. However, gate model algorithms for universal quantum computing are not available. In this paper, gate-based quantum algorithms will be presented, which are highly flexible and can be applied to various physical devices. Mapped all possible secondary structure to the state of a quadratic Hamiltonian, the whole folding process is described as a quadratic unconstrained binary optimization model. Then the model can be solved through quantum approximation optimization algorithm. We demonstrate the performance with both numerical simulation and experimental realization. Throughout our benchmark dataset, simulation results suggest that our quantum approach is comparable in accuracy to classical methods. For non-nested sequences, our quantum approach outperforms classical energy-based methods. Experimental results also indicate our method is robust in current noisy devices. It is the first instance of universal quantum algorithms being employed to tackle RNA folding problems, and our work provides a valuable model for utilizing universal quantum computers in solving RNA folding problems., Comment: 19 pages, 7 figures

This study presents an ensemble approach that addresses the challenges of identification and analysis of research articles in rapidly evolving fields, using the field of Artificial Intelligence (AI) as a case study. Our approach included using decision tree, sciBERT and regular expression matching on different fields of the articles, and a SVM to merge the results from different models. We evaluated the effectiveness of our method on a manually labeled dataset, finding that our combined approach captured around 97% of AI-related articles in the web of science (WoS) corpus with a precision of 0.92. This presents a 0.15 increase in F1 score compared with existing search term based approach. Following this, we analyzed the publication volume trends and common research themes.We found that compared with existing methods, our ensemble approach revealed an increased degree of interdisciplinarity, and was able to identify more articles in certain subfields like feature extraction and optimization. This study demonstrates the potential of our approach as a tool for the accurate identification of scholarly articles, which is also capable of providing insights into the volume and content of a research area., Comment: 27 pages, 8 figures, 5 tables