Images are typically sampled on a uniform grid,despite their non-uniform information distribution—some regions are rich in content while others are not. The mismatch leads to inefficient computation allocation in deep learning models. To address this, recent studies have proposed predictive downsampling methodsthat adaptively downsample images based on predicted per-pixel importance, allocating more pixels to informative areas. However,these methods require high-resolution processing to accurately estimate importance, which undermines their efficiency:the prediction itself must process the full-resolution image,consuming most of the computational budget. This high-resolution importance prediction is necessary because each input may differ significantly in structure and content. In this paper, we take a different approach and introduce a learn-to-downsample paradigmtailored for aligned vision recognition tasks, such as face recognition and palmprint recognition, where input alignment ensures consistent spatial structure across images. This alignment ensures structural consistency across images, allowing a shared, input-agnostic downsampling template applicable to all inputs. Furthermore, instead of relying on implicit importance maps, we introduce a flow-based representation that explicitly models the spatial warping from the original image to the downsampled version. The flow representation is not only more efficient but also more controllable: we regularize the flow using its Jacobian determinant to precisely control the sampling density and coverage,enabling interpretable and tunable sampling patterns. Extensive experiments on two aligned recognition tasks, face and palmprint recognition, demonstrate that our method substantially reduces computational cost with minimal accuracy degradation, achieving a significantly better performance-efficiency trade-off than existing predictive downsampling methods.

Open-vocabulary camouflaged object segmentation (OVCOS) seeks to segment and classify camouflaged objects in arbitrary categories, presenting unique challenges due to visual ambiguity and unseen categories. Recent approaches typically adopt a two-stage paradigm: they first segment objects, and then classify the segmented regions using vision language models (VLMs). However, such methods (i) suffer from a domain gap caused by the mismatch between VLMs' full-image training and cropped-region inferencing, and (ii) depend on generic segmentation models optimized for well-delineated objects which are less effective for camouflaged objects. Without explicit guidance, generic segmentation models often overlook subtle boundaries, leading to imprecise segmentation. In this paper, we introduce a novel VLM-guided cascaded framework to address these issues in OVCOS. For segmentation, we leverage the segment anything model (SAM), guided by the VLM. Our framework uses VLM-derived features as explicit prompts to SAM, effectively directing attention to camouflaged regions and significantly improving localization accuracy. For classification, we avoid the domain gap introduced by hard cropping. Instead, we treat the segmentation output as a soft spatial prior using the alpha channel. This retains the full image context while providing precise spatial guidance, leading to more accurate and context-aware classification of camouflaged objects. The same VLM is shared between segmentation and classification to ensure efficiency and semantic consistency. Extensive experiments on both OVCOS and conventional camouflaged object segmentation benchmarks demonstrate the clear superiority of our method, highlighting the effectiveness of leveraging rich VLM semantics for both segmentation and classification of camouflaged objects. Our code and models are open-sourced at https://github.com/intcomp/camouflaged-vlm.

The rapid proliferation of wireless technologies and dynamic spectrum sharing in Internet of Things (IoT) and cyber–physical systems has increased the vulnerability of radio-frequency (RF) environments to anomalies such as unauthorized transmissions, interference, and spoofing. Reliable spectrum anomaly detection (SAD) is therefore critical for secure and dependable wireless operation. However, existing unsupervised approaches, including autoencoders (AEs), memory-augmented AEs (MemAEs), and distillation-based methods such as UDMA, often suffer performance degradation in low signal-to-noise ratio (SNR) regimes and inadequately capture the structural dependencies inherent in spectral data. This paper proposes GRETEL, a fully unsupervised graph based framework for spectrum anomaly detection. GRETEL transforms power spectral density (PSD) sequences into graph representations, where nodes encode localized spectral segments augmented with frequency positional embeddings and SNR-aware features. A graph attention network is combined with a gated cross-attention mechanism over a learnable prototype memory to jointly model local spectral patterns and global dependencies, enabling robust detection of subtle deviations from normal spectrum behavior. Extensive evaluations on both synthetic and real-world RF datasets demonstrate that GRETEL consistently outperforms AE, MemAE, and UDMA-based detectors, as well as Isolation Forest and One-Class SVM baselines, in terms of F1-score and AUROC, particularly under severe noise conditions with SNRs as low as -35 dB. Furthermore, a comprehensive computational analysis reports the framework’s complexity in terms of FLOPs, memory footprint, and inference latency on representative edge devices, highlighting its suitability for resource-constrained IoT deployments. The attention-based design also provides interpretability by identifying the spectral regions that most influence anomaly decisions.

In intelligent education, cognitive diagnostic assessment is a fundamental task for evaluating students’ mastery of knowledge attributes based on their response logs. However, existing cognitive diagnostic models face several limitations. For example, the Q-matrix heavily relies on expert-defined prior knowledge and remains difficult to optimize automatically. In addition, most models assume static knowledge states, limiting their ability to capture the fine-grained dynamics of student cognition. To address these issues, this paper proposes a dynamic cognitive diagnosis model enhanced by genetic evolution (GE-QCDM). Specifically, latent knowledge attributes are discovered through a genetic algorithm, and a refined Q-matrix is generated via discrete-space search to incorporate these potential attributes. A Q-matrix enhancement strategy is further introduced to retain the accuracy of explicit attributes while improving interpretability by uncovering hidden associations. Furthermore, a dynamic attention mechanism is designed to model the interaction between learner states and question characteristics, thereby enabling a more comprehensive evaluation of students’ mastery levels. Experimental results on multiple real-world datasets demonstrate that the proposed model consistently outperforms existing methods in both accuracy and interpretability.

The rapid development of Integrated Sensing and Communication (ISAC) has driven the need for robust and adaptable indoor positioning systems. Channel State Information (CSI)-based fingerprint localization has emerged as a promising solution, but existing methods face significant challenges as their sensitivity to noise often leads to poor localization accuracy and deep neural networks suffer from computational inefficiency. In this paper, we propose KFBK, a novel framework that integrates wavelet decomposition, Kolmogorov-Arnold Network (KAN)-based feature learning, and a dynamic fusion mechanism of the Broad Learning System (BLS) and KAN. It first applies a hybrid wavelet denoising strategy with decomposition and adaptive thresholding to suppress noise while preserving critical CSI patterns. Then, a KAN-based feature extractor with splineoptimized activation functions captures complex nonlinear spatiotemporal dependencies in high-dimensional CSI data, enabling dimensionality reduction without losing essential information. Finally, a temperature-controlled dynamic weighting mechanism adaptively adjusts the contributions of BLS and KAN to improve overall localization performance under varying conditions. Extensive evaluations in two real-world environments demonstrate that KFBK outperforms state-of-the-art methods in localization accuracy and environmental adaptability while maintaining realtime responsiveness and computational efficiency.

To address the challenges of small target features being less prominent, susceptibility to background interference, and sample imbalance in road traffic sign detection, which leads to insufficient model detection accuracy, as well as the high complexity of current object detection models that struggle to operate efficiently on resource-constrained edge devices, we propose a traffic sign detection model based on GCA-YOLO. By adding small target detection layers and removing large target detection layers, the model enhances its small target detection capabilities and reduces its parameter size. The introduction of the T-BiFPN (Tiny-BiFPN) structure improves multi-scale feature fusion, while the C2f-CP module increases computational efficiency on edge devices. The GCA (Global Coordinate Attention) mechanism enhances feature extraction, and the Focaler-CIoU loss function enables the model to focus more on difficult samples and accelerate the convergence of bounding boxes. Experimental results show the superiorities of the proposed GCA-YOLO that compared to YOLOv8n, GCA-YOLO improves precision, recall, mAP@50, and mAP@50:95 by 8.6%, 6.1%, 8.7%, and 6.2%, respectively, while reducing the model's parameter count and size by 38.57% and 33.21%, respectively.

Fingerprint localization methods typically require a substantial amount of manual effort to collect fingerprint data from various scenarios to construct an accurate radio map. While some existing research has attempted to use path planning strategies to save on labor costs, these approaches often suffer from being time-consuming and prone to locally optimal solutions. To address these shortcomings, our paper proposes a novel approach that utilizes imitation learning to construct and update a highly accurate radio map with minimal manual intervention in dynamic environments. Specifically, we employ a multivariate Gaussian process model to fit a rough standby fingerprint database with only a few pilot data points. We then utilize a state space model to calculate the variation range of the pilot data, which forms the CSI error band used to filter the rough radio map. Imitation learning and a confidence coefficient are utilized to predict and calibrate the global CSI data distribution. And we utilize the K-nearest neighbor algorithm to achieve the real-time localization function. Experimental results show that our proposed algorithm outperforms several state-of-the-art approaches in most test cases, exhibiting low computation complexity, lower localization error, and saving 73.3% of the manual workload.

With the advent of the 5.5G era, integrated sensing and communication (ISAC) technology has emerged, demonstrating its high-precision sensing capabilities across industries like wireless communication, intelligent transportation, and smart homes. Researchers are particularly exploring the use of channel state information (CSI) from WiFi signals for human activity recognition. However, current research primarily relies on complex network models to manage vast CSI data for developing accurate recognition models. It confronts two main challenges in time-sensitive applications: a limited number of samples and the need for timely solutions. In this paper, we present a prototype network model that combines CNN and self-attention to address the issues raised above, namely Proto-CSNet. First, we use time-frequency analysis and filter the data to transfer the data from the time domain to the frequency domain and eliminate noise and outliers from the samples. Then, based on the ACmix model, we modify and build a novel custom feature extraction module that retains the fused CNN's local feature capture capabilities as well as self-attention's global information processing advantage. Finally, we merge the custom modules with CNNs to form a prototype network. A real-world dataset is constructed including 11 different human activities, and the extensive experimental resutls demonstrate that Proto-CSNet outperforms existing algorithms with a model inference accuracy exceeding 95%.

With the rapid development of Integrated Sensing and Communication in 6G, Channel State Information (CSI)-based fingerprint indoor localization technology is becoming crucial. However, during the offline phase, the fingerprint database update process using crowdsourcing techniques is prone to noise interference and incomplete coverage, and fitting Gaussian regression models requires extensive computational resources. In this paper, we propose a fine-grained CSI fingerprint database update method based on Deep-Broad Learning system (DeepBLS) and Generative Adversarial Networks (GAN), termed DBLG. Firstly, we employ the combined DeepBLS network for the initial construction of the global CSI fingerprint database. Subsequently, we utilize GAN to extract features from the raw data, predict and update the global CSI fingerprint database, and construct a high-precision fingerprint database using confidence coefficients. Finally, We implement the proposed algorithm in two real-world environments and conduct extensive experiments to verify its performance. Compared to several existing methods, our approach shows superior performance in updating the CSI finger-print database, achieving a 46.78 % improvement in localization accuracy.

Abstract 3D reconstruction is a long‐standing problem. Recently, a number of studies have emerged that utilize transformers for 3D reconstruction, and these approaches have demonstrated strong performance. However, transformer‐based 3D reconstruction methods tend to establish the transformation relationship between the 2D image and the 3D voxel space directly using transformers or rely solely on the powerful feature extraction capabilities of transformers. They ignore the crucial role played by deep multi‐scale representation of the object in the voxel feature domain, which can provide extensive global shape and local detail information about the object in a multi‐scale manner. In this article, we propose a novel framework TMSDNet (transformer with multi‐scale dense network) for single‐view and multi‐view 3D reconstruction with transformer to solve this problem. Based on our well‐designed combined‐transformer Block, which is canonical encoder–decoder architecture, voxel features with spatial order can be extracted from the input image, which are used to further extract multi‐scale global features in parallel using a multi‐scale residual attention module. Furthermore, a residual dense attention block is introduced for deep local features extraction and adaptive fusion. Finally, the reconstructed objects are produced with the voxel reconstruction block. Experiment results on the benchmarks such as ShapeNet and Pix3D datasets demonstrate that TMSDNet outperforms the existing state‐of‐the‐art reconstruction methods substantially.

Recent advances in NeRF-based 3D-aware GANs have achieved outstanding performance, especially in the realm of human facial representations, making projection of facial images back into their latent space superior and preferable compared to 2D GAN inversion. However, the direct application of 2DGAN inversion techniques to 3DGAN raises challenges due to potential appearance distortions and geometric inconsistences. To tackle these issues, this work presents a novel integrated framework that combines a composite inversion pipeline in both the SS and W+ spaces and integrates a contrastive-based training strategy, ensuring proficient disentanglement within the module. Moreover, we design a facial semantic manipulation technique based on dimensional analysis of the latent code, which is fully compatible with the proposed 3DGAN inversion pipeline. Comprehensive experimental validations substantiate the effectiveness of the proposed approach in executing 3d-aware face inversion and semantic editing tasks, presenting a robust technological solution for a diverse array of digital human modeling applications in the downstream.

Neuron tracing, alternately referred to as neuron reconstruction, is the procedure for extracting the digital representation of the three-dimensional neuronal morphology from stacks of microscopic images. Achieving accurate neuron tracing is critical for profiling the neuroanatomical structure at single-cell level and analyzing the neuronal circuits and projections at whole-brain scale. However, the process often demands substantial human involvement and represents a nontrivial task. Conventional solutions towards neuron tracing often contend with challenges such as non-intuitive user interactions, suboptimal data generation throughput, and ambiguous visualization. In this paper, we introduce a novel method that leverages the power of extended reality (XR) for intuitive and progressive semi-automatic neuron tracing in real time. In our method, we have defined a set of interactors for controllable and efficient interactions for neuron tracing in an immersive environment. We have also developed a GPU-accelerated automatic tracing algorithm that can generate updated neuron reconstruction in real time. In addition, we have built a visualizer for fast and improved visual experience, particularly when working with both volumetric images and 3D objects. Our method has been successfully implemented with one virtual reality (VR) headset and one augmented reality (AR) headset with satisfying results achieved. We also conducted two user studies and proved the effectiveness of the interactors and the efficiency of our method in comparison with other approaches for neuron tracing.

extended-abstract Share on International Workshop on Deep Learning Practice for High-Dimensional Sparse Data with RecSys 2023 Authors: Ruiming Tang Huawei Noah's Ark Lab, China Huawei Noah's Ark Lab, China 0000-0002-9224-2431View Profile , Xiaoqiang Zhu Mobvista Group, China Mobvista Group, China 0000-0001-7486-0853View Profile , Junfeng Ge Alibaba Group, China Alibaba Group, China 0000-0001-8435-0443View Profile , Kuang-chih Lee Alibaba Group, USA Alibaba Group, USA 0009-0007-5198-9866View Profile , Biye Jiang Alibaba Group, China Alibaba Group, China 0009-0001-5814-1581View Profile , Xingxing Wang Meituan, China Meituan, China 0000-0002-2655-3928View Profile , Han Zhu Alibaba Group, China Alibaba Group, China 0000-0002-9522-5637View Profile , Tao Zhuang Alibaba Group, China Alibaba Group, China 0000-0002-7408-8514View Profile , Weiwen Liu Huawei Noah's Ark Lab, China Huawei Noah's Ark Lab, China 0000-0002-9148-3997View Profile , Kan Ren Microsoft Research, China Microsoft Research, China 0000-0002-4032-9615View Profile , Weinan Zhang Shanghai Jiao Tong University, China Shanghai Jiao Tong University, China 0000-0002-0127-2425View Profile , Xiangyu Zhao City University of Hong Kong, China City University of Hong Kong, China 0000-0003-2926-4416View Profile Authors Info & Claims RecSys '23: Proceedings of the 17th ACM Conference on Recommender SystemsSeptember 2023Pages 1276–1280https://doi.org/10.1145/3604915.3608765Published:14 September 2023Publication History 0citation67DownloadsMetricsTotal Citations0Total Downloads67Last 12 Months67Last 6 weeks67 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

With the rapid growth in the demand for location-based services in indoor environments, wireless fingerprint localization has attracted increasing attention because of its high precision and easy implementation. However, an effective method does not exist owing to the problems of data loss, noise interference in the fingerprint database, and being time-consuming during the offline training phase. Therefore, this paper presents a novel indoor wireless fingerprint localization algorithm, termed BLS-Location, based on a broad learning system (BLS) that utilizes channel state information (CSI) to overcome the aforementioned problems. It includes an offline training phase and an online localization phase. In the offline training phase, the Kalman filter and the expectation-maximization (EM) algorithm are utilized for completing and denoising the data. Moreover, principal component analysis (PCA) is used to reconstruct the CSI data to reduce complexity and train the weights by BLS. In the online localization phase, we employ a novel probabilistic method based on the regression results of BLS to obtain the estimated location. The experimental results show that BLS-Location can significantly reduce the training time with a high accuracy, compared to several machine learning algorithms and four existing methods in two representative indoor environments.

With the growing demand of Location-Based Service, the fingerprint localization based on Channel State Information (CSI) has become a vital positioning technology because it has easy implementation, low device cost and adequate accuracy which benefits from fine-grained information provided by CSI. However, the main drawback is that the approach has to construct the fingerprint map manually during the off-line stage, which is tedious and time-consuming. In this paper, we propose a novel data collection strategy for path planning based on reinforcement learning, namely Asynchronous Advantage Actor-Critic (A3C). Given the limited exploration step length, it needs to maximize the informative CSI data for reducing manual cost. We collect a small amount of real data in advance to predict the rewards of all sampling points by multivariate Gaussian process and mutual information. Then the optimization problem is transformed into a sequential decision process, which can exploit the informative path by A3C. We complete the proposed algorithm in two real-world dynamic environments and extensive experiments verify its performance. Compared to coverage path planning and several existing algorithms, our system not only can achieve similar indoor localization accuracy, but also reduce the CSI collection task.

Recently, we have witnessed that deep learning-based approaches have been widely applied. Particularly, some applications involve data that are high dimensional, sparse or imbalanced, which are different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice and theory of deep learning on high-dimensional, sparse and imbalanced data.

An important technology for copyright protection is digital watermarking, which generally embeds copyright infor-mation into the carrier invisibly. In the case of various attacks, the identity of the watermark can also be extracted to prove its copyright from the carrier of the watermark. Most of the existing watermarking technologies use natural images as carriers. Unlike natural images, document images are not rich in color and texture, and thus carry less redundant information for watermarks. This paper proposes a adaptive positioning document watermarking algorithm based on Discrete Wavelet Transform(DWT) and Discrete Cosine Transform(DCT). The method integrates adaptive positioning algorithm, discrete wavelet transform, discrete cosine transform and other algorithms. In our case, we first scan the document image to adaptively embed the keypoints information, so that even if a part of the document is recorded, the complete watermark area can be accurately located. Secondly, we construct the watermark embedding regions centered at keypoints. Third, the DWT and DCT algorithms are used to embed the watermark, which achieve strong robustness to the capture process attack. Finally, the watermark can be extracted from at least one region in the extraction stage. Compared with the previous watermarking algorithms, the proposed approach not only ensures the independence of embedded content, but also is robust to various noise attacks. In addition, the method can meet the extraction accuracy requirements even when the document image is incomplete.

Feature interaction has been recognized as an important problem in machine learning, which is also very essential for click-through rate (CTR) prediction tasks. In recent years, Deep Neural Networks (DNNs) can automatically learn implicit nonlinear interactions from original sparse features, and therefore have been widely used in industrial CTR prediction tasks. However, the implicit feature interactions learned in DNNs cannot fully retain the complete representation capacity of the original and empirical feature interactions (e.g., cartesian product) without loss. For example, a simple attempt to learn the combination of feature A and feature B < A, B > as the explicit cartesian product representation of new features can outperform previous implicit feature interaction models including factorization machine (FM)-based models and their variations. This indicates there is still a big gap between explicit and implicit feature interaction models. However, to learn all the explicit feature interaction (cartesian product) representations requires a very large sample size along with N times of original parameter space (where N is quite large in most industrial applications). In this paper, we propose a Co-Action Network (CAN) to approximate the explicit pairwise feature interactions without introducing too many additional parameters. More specifically, giving feature A and its associated feature B, their feature interaction is modeled by learning two sets of parameters: 1) the embedding of feature A, and 2) a Multi-Layer Perceptron (MLP) to represent feature B. The approximated feature interaction can be obtained by passing the embedding of feature A through the MLP network of feature B. We refer to such pairwise feature interaction as feature co-action, and such a Co-Action Network unit can provide a very powerful capacity to fitting complex feature interactions. In addition, FM can be viewed as a special case of the CAN unit when the MLP is a single layer with only one output. Experimental results on public and industrial datasets show that CAN outperforms state-of-the-art CTR models and the cartesian product method. Moreover, CAN has been deployed in the display advertisement system in Alibaba, obtaining 12% improvement on CTR and 8% on Revenue Per Mille (RPM), which is a great improvement to the business. The code for experiments in this paper is open-sourced\footnotehttps://github.com/CAN-Paper/Co-Action-Network.

In online advertising, auto-bidding has become an essential tool for advertisers to optimize their preferred ad performance metrics by simply expressing high-level campaign objectives and constraints. Previous works designed auto-bidding tools from the view of single-agent, without modeling the mutual influence between agents. In this paper, we instead consider this problem from a distributed multi-agent perspective, and propose a general \underlineM ulti-\underlineA gent reinforcement learning framework for \underlineA uto-\underlineB idding, namely MAAB, to learn the auto-bidding strategies. First, we investigate the competition and cooperation relation among auto-bidding agents, and propose a temperature-regularized credit assignment to establish a mixed cooperative-competitive paradigm. By carefully making a competition and cooperation trade-off among agents, we can reach an equilibrium state that guarantees not only individual advertiser's utility but also the system performance (i.e., social welfare). Second, to avoid the potential collusion behaviors of bidding low prices underlying the cooperation, we further propose bar agents to set a personalized bidding bar for each agent, and then alleviate the revenue degradation due to the cooperation. Third, to deploy MAAB in the large-scale advertising system with millions of advertisers, we propose a mean-field approach. By grouping advertisers with the same objective as a mean auto-bidding agent, the interactions among the large-scale advertisers are greatly simplified, making it practical to train MAAB efficiently. Extensive experiments on the offline industrial dataset and Alibaba advertising platform demonstrate that our approach outperforms several baseline methods in terms of social welfare and revenue.

Traditional industry recommendation systems usually use data in a single domain to train models and then serve the domain. However, a large-scale commercial platform often contains multiple domains, and its recommendation system often needs to make click-through rate (CTR) predictions for multiple domains. Generally, different domains may share some common user groups and items, and each domain may have its own unique user groups and items. Moreover, even the same user may have different behaviors in different domains. In order to leverage all the data from different domains, a single model can be trained to serve all domains. However, it is difficult for a single model to capture the characteristics of various domains and serve all domains well. On the other hand, training an individual model for each domain separately does not fully use the data from all domains. In this paper, we propose the Star Topology Adaptive Recommender (STAR) model to train a single model to serve all domains by leveraging data from all domains simultaneously, capturing the characteristics of each domain, and modeling the commonalities between different domains. Essentially, the net- work of each domain consists of two factorized networks: one centered network shared by all domains and the domain-specific network tailored for each domain. For each domain, we combine these two factorized networks and generate a unified network by element-wise multiplying the weights of the shared network and those of the domain-specific network, although these two factorized networks can be combined using other functions, which is open for further research. Most importantly, STAR can learn the shared network from all the data and adapt domain-specific parameters according to the characteristics of each domain. The experimental results from production data validate the superiority of the proposed STAR model. Since late 2020, STAR has been deployed in the display advertising system of Alibaba, obtaining 8.0% improvement on CTR and 6.0% increase on RPM (Revenue Per Mille).

Online display advertising has become a vital business for large-scale E-commerce markets. As the main goal of advertisers is to reach interested customer prospects, accurate conversion prediction is essential for successful online display advertising. A particular challenge for conversion prediction is that conversions may occur long after the click events. Such delayed feedback makes it a non-trivial task to keep conversion prediction models updated and consistent with the latest customer distribution. Although several studies have been conducted to tackle the delayed feedback issue, the relationship between the early conversion and full term conversion has not been fully exploited to improve conversion prediction. In this paper, we consider conversion prediction as a multi-task learning problem by leveraging multiple conversion labels after different observation intervals. Specifically, we propose a multi-task model with an end-to-end architecture for conversion prediction. Our approach is guided by theoretical and probabilistic analysis of the early and full term conversions. Our mixture-of-experts module can integrate distinct characteristics of input features and optimize the task-specific experts. In addition, the multiple tasks are jointly learned with a regularization term to ensure the embedding consistency between tasks and prevent potential overfitting issues. In comparison with competitive benchmarks, our approach can significantly improve conversion prediction with delayed feedback and improve business performance of online display advertising.

Uplift modeling is an important task for online advertising and marketing. Advertisers rely on accurate modeling of the uplift effect to formulate, plan and implement their advertising and marketing strategies. Therefore, the capability of effective and efficient uplift modeling is essential for advertising platforms to attract and satisfy their customers (i.e., advertisers). In practical advertising applications, uplift modeling focuses on the estimation of the uplift effect caused by ad exposure. It is not a trivial task to estimate such causal impact of ad exposure at the individual level. In this paper, we propose an end-to-end approach for explicit uplift modeling, using data collected from Randomized Controlled Trials (RCTs) in large-scale real-world advertising platforms. More specifically, we first introduce the Explicit Uplift Effect Network (EUEN) to explicitly model the uplift effect and demonstrate its advantages in uplift modeling. Then for the exposure uplift effect modeling, we further propose the Explicit Exposure Uplift Effect Network (EEUEN), which can correct the exposure bias for uplift modeling. We evaluate our proposed approach with both public data sets as well as data sets collected from our advertising platform. The significant improvements with respect to various performance metrics demonstrate the advantages of our approach.

Modern online advertising systems inevitably rely on personalization methods, such as click-through rate (CTR) prediction. Recent progress in CTR prediction enjoys the rich representation capabilities of deep learning and achieves great success in large-scale industrial applications. However, these methods can suffer from lack of exploration. Another line of prior work addresses the exploration-exploitation trade-off problem with contextual bandit methods, which are recently less studied in the industry due to the difficulty in extending their flexibility with deep models. In this paper, we propose a novel Deep Uncertainty-Aware Learning (DUAL) method to learn CTR models based on Gaussian processes, which can provide predictive uncertainty estimations while maintaining the flexibility of deep neural networks. DUAL can be easily implemented on existing models and deployed in real-time systems with minimal extra computational overhead. By linking the predictive uncertainty estimation ability of DUAL to well-known bandit algorithms, we further present DUAL-based Ad-ranking strategies to boost up long-term utilities such as the social welfare in advertising systems. Experimental results on several public datasets demonstrate the effectiveness of our methods. Remarkably, an online A/B test deployed in the Alibaba display advertising platform shows an 8.2% social welfare improvement and an 8.0% revenue lift.

One of the difficulties of conversion rate (CVR) prediction is that the conversions can delay and take place long after the clicks. The delayed feedback poses a challenge: fresh data are beneficial to continuous training but may not have complete label information at the time they are ingested into the training pipeline. To balance model freshness and label certainty, previous methods set a short waiting window or even do not wait for the conversion signal. If conversion happens outside the waiting window, this sample will be duplicated and ingested into the training pipeline with a positive label. However, these methods have some issues. First, they assume the observed feature distribution remains the same as the actual distribution. But this assumption does not hold due to the ingestion of duplicated samples. Second, the certainty of the conversion action only comes from the positives. But the positives are scarce as conversions are sparse in commercial systems. These issues induce bias during the modeling of delayed feedback. In this paper, we propose DElayed FEedback modeling with Real negatives (DEFER) method to address these issues. The proposed method ingests real negative samples into the training pipeline. The ingestion of real negatives ensures the observed feature distribution is equivalent to the actual distribution, thus reducing the bias. The ingestion of real negatives also brings more certainty information of the conversion. To correct the distribution shift, DEFER employs importance sampling to weigh the loss function. Experimental results on industrial datasets validate the superiority of DEFER. DEFER have been deployed in the display advertising system of Alibaba, obtaining over 6.0% improvement on CVR in several scenarios. The code and data in this paper are now open-sourced (https://github.com/gusuperstar/defer.git).

Advertising expenditures have become the major source of revenue for e-commerce platforms. Providing good advertising experiences for advertisers by reducing their costs of trial and error in discovering the optimal advertising strategies is crucial for the long-term prosperity of online advertising. To achieve this goal, the advertising platform needs to identify the advertiser's optimization objectives, and then recommend the corresponding strategies to fulfill the objectives. In this work, we first deploy a prototype of strategy recommender system on Taobao display advertising platform, which indeed increases the advertisers' performance and the platform's revenue, indicating the effectiveness of strategy recommendation for online advertising. We further augment this prototype system by explicitly learning the advertisers' preferences over various advertising performance indicators and then optimization objectives through their adoptions of different recommending advertising strategies. We use contextual bandit algorithms to efficiently learn the advertisers' preferences and maximize the recommendation adoption, simultaneously. Simulation experiments based on Taobao online bidding data show that the designed algorithms can effectively optimize the strategy adoption rate of advertisers.

In online display advertising, advertisers usually participate in real-time bidding to acquire ad impression opportunities. In most advertising platforms, a typical impression acquiring demand of advertisers is to maximize the sum value of winning impressions under budget and some key performance indicators constraints, (e.g. maximizing clicks with the constraints of budget and cost per click upper bound). The demand can be various in value type (e.g. ad exposure/click), constraint type (e.g. cost per unit value) and constraint number. Existing works usually focus on a specific demand or hardly achieve the optimum. In this paper, we formulate the demand as a constrained bidding problem, and deduce a unified optimal bidding function on behalf of an advertiser. The optimal bidding function facilitates an advertiser calculating bids for all impressions with only m parameters, where m is the constraint number. However, in real application, it is non-trivial to determine the parameters due to the non-stationary auction environment. We further propose a reinforcement learning (RL) method to dynamically adjust parameters to achieve the optimum, whose converging efficiency is significantly boosted by the recursive optimization property in our formulation. We name the formulation and the RL method, together, as Unified Solution to Constrained Bidding (USCB). USCB is verified to be effective on industrial datasets and is deployed in Alibaba display advertising platform.

In e-commerce advertising, it is crucial to jointly consider various performance metrics, e.g., user experience, advertiser utility, and platform revenue. Traditional auction mechanisms, such as GSP and VCG auctions, can be suboptimal due to their fixed allocation rules to optimize a single performance metric (e.g., revenue or social welfare). Recently, data-driven auctions, learned directly from auction outcomes to optimize multiple performance metrics, have attracted increasing research interests. However, the procedure of auction mechanisms involves various discrete calculation operations, making it challenging to be compatible with continuous optimization pipelines in machine learning. In this paper, we design Deep Neural Auctions (DNAs) to enable end-to-end auction learning by proposing a differentiable model to relax the discrete sorting operation, a key component in auctions. We optimize the performance metrics by developing deep models to efficiently extract contexts from auctions, providing rich features for auction design. We further integrate the game theoretical conditions within the model design, to guarantee the stability of the auctions. DNAs have been successfully deployed in the e-commerce advertising system at Taobao. Experimental evaluation results on both large-scale data set as well as online A/B test demonstrated that DNAs significantly outperformed other mechanisms widely adopted in industry.

Recently, we have witnessed that deep learning-based approaches has been widely applied to empower many internet-scale applications. However, the data in these internet-scale applications are high dimensional and extremely sparse, which makes it different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice of deep learning on high-dimensional sparse data.

Rich user behavior data has been proven to be of great value for click-through rate prediction tasks, especially in industrial applications such as recommender systems and online advertising. Both industry and academy have paid much attention to this topic and propose different approaches to modeling with long sequential user behavior data. Among them, memory network based model MIMN proposed by Alibaba, achieves SOTA with the co-design of both learning algorithm and serving system. MIMN is the first industrial solution that can model sequential user behavior data with length scaling up to 1000. However, MIMN fails to precisely capture user interests given a specific candidate item when the length of user behavior sequence increases further, say, by 10 times or more. This challenge exists widely in previously proposed approaches.

Position bias is a critical problem in information retrieval when dealing with implicit yet biased user feedback data. Unbiased ranking methods typically rely on causality models and debias the user feedback through inverse propensity weighting. While practical, these methods still suffer from two major problems. First, when infer a user click, the impact of the contextual information, such as documents that have been examined, is often ignored. Second, only the position bias is considered but other issues resulted from user browsing behaviors are overlooked. In this paper, we propose an end-to-end Deep Recurrent Survival Ranking (DRSR), a unified framework to jointly model user's various behaviors, to (i) consider the rich contextual information in the ranking list; and (ii) address the hidden issues underlying user behaviors, i.e., to mine observe pattern in queries without any click (non-click queries), and to model tracking logs which cannot truly reflect the user browsing intents (untrusted observation). Specifically, we adopt a recurrent neural network to model the contextual information and estimates the conditional likelihood of user feedback at each position. We then incorporate survival analysis techniques with the probability chain rule to mathematically recover the unbiased joint probability of one user's various behaviors. DRSR can be easily incorporated with both point-wise and pair-wise learning objectives. The extensive experiments over two large-scale industrial datasets demonstrate the significant performance gains of our model comparing with the state-of-the-arts.

Click-through rate (CTR) prediction, whose goal is to estimate the probability of a user clicking on the item, has become one of the core tasks in the advertising system. For CTR prediction model, it is necessary to capture the latent user interest behind the user behavior data. Besides, considering the changing of the external environment and the internal cognition, user interest evolves over time dynamically. There are several CTR prediction methods for interest modeling, while most of them regard the representation of behavior as the interest directly, and lack specially modeling for latent interest behind the concrete behavior. Moreover, little work considers the changing trend of the interest. In this paper, we propose a novel model, named Deep Interest Evolution Network (DIEN), for CTR prediction. Specifically, we design interest extractor layer to capture temporal interests from history behavior sequence. At this layer, we introduce an auxiliary loss to supervise interest extracting at each step. As user interests are diverse, especially in the e-commerce system, we propose interest evolving layer to capture interest evolving process that is relative to the target item. At interest evolving layer, attention mechanism is embedded into the sequential structure novelly, and the effects of relative interests are strengthened during interest evolution. In the experiments on both public and industrial datasets, DIEN significantly outperforms the state-of-the-art solutions. Notably, DIEN has been deployed in the display advertisement system of Taobao, and obtained 20.7% improvement on CTR.

Click-through rate (CTR) prediction is critical for industrial applications such as recommender system and online advertising. Practically, it plays an important role for CTR modeling in these applications by mining user interest from rich historical behavior data. Driven by the development of deep learning, deep CTR models with ingeniously designed architecture for user interest modeling have been proposed, bringing remarkable improvement of model performance over offline metric. However, great efforts are needed to deploy these complex models to online serving system for realtime inference, facing massive traffic request. Things turn to be more difficult when it comes to long sequential user behavior data, as the system latency and storage cost increase approximately linearly with the length of user behavior sequence.

User response prediction, which models the user preference w.r.t. the presented items, plays a key role in online services. With two-decade rapid development, nowadays the cumulated user behavior sequences on mature Internet service platforms have become extremely long since the user's first registration. Each user not only has intrinsic tastes, but also keeps changing her personal interests during lifetime. Hence, it is challenging to handle such lifelong sequential modeling for each individual user. Existing methodologies for sequential modeling are only capable of dealing with relatively recent user behaviors, which leaves huge space for modeling long-term especially lifelong sequential patterns to facilitate user modeling. Moreover, one user's behavior may be accounted for various previous behaviors within her whole online activity history, i.e., long-term dependency with multi-scale sequential patterns. In order to tackle these challenges, in this paper, we propose a Hierarchical Periodic Memory Network for lifelong sequential modeling with personalized memorization of sequential patterns for each user. The model also adopts a hierarchical and periodical updating mechanism to capture multi-scale sequential patterns of user interests while supporting the evolving user behavior logs. The experimental results over three large-scale real-world datasets have demonstrated the advantages of our proposed model with significant improvement in user response prediction performance against the state-of-the-arts.

Abstract Although different approaches of segmenting and assembling geometric models for 3D printing have been proposed, it is difficult to find any research studies, which investigate model segmentation and assembly in head‐mounted display (HMD) virtual reality (VR) environments for 3D printing. In this work, we propose a novel and interactive segmentation method for screwing assembly in the environments to tackle this problem. Our approach divides a large model into semantic parts with a screwing interface for repeated tight assembly. Specifically, after a user places the cutting interface, our algorithm computes the bounding box of the current part automatically for subsequent multicomponent semantic Boolean segmentations. Afterwards, the bolt is positioned with an improved K3M image thinning algorithm and is used for merging paired components with union and subtraction Boolean operations respectively. Moreover, we introduce a swept Boolean‐based rotation collision detection and location method to guarantee a collision‐free screwing assembly. Experiments show that our approach provides a new interactive multicomponent semantic segmentation tool that supports not only repeated installation and disassembly but also tight and aligned assembly.

Models applied on real time response tasks, like click-through rate (CTR) prediction model, require high accuracy and rigorous response time. Therefore, top-performing deep models of high depth and complexity are not well suited for these applications with the limitations on the inference time. In order to get neural networks of better performance given the time limitations, we propose a universal framework that exploits a booster net to help train the lightweight net for prediction. We dub the whole process rocket launching, where the booster net is used to guide the learning of our light net throughout the whole training process. We analyze different loss functions aiming at pushing the light net to behave similarly to the booster net. Besides, we use one technique called gradient block to improve the performance of light net and booster net further. Experiments on benchmark datasets and real-life industrial advertisement data show the effectiveness of our proposed method.

In Taobao, the largest e-commerce platform in China, billions of items are provided and typically displayed with their images.For better user experience and business effectiveness, Click Through Rate (CTR) prediction in online advertising system exploits abundant user historical behaviors to identify whether a user is interested in a candidate ad. Enhancing behavior representations with user behavior images will help understand user's visual preference and improve the accuracy of CTR prediction greatly. So we propose to model user preference jointly with user behavior ID features and behavior images. However, training with user behavior images brings tens to hundreds of images in one sample, giving rise to a great challenge in both communication and computation. To handle these challenges, we propose a novel and efficient distributed machine learning paradigm called Advanced Model Server (AMS). With the well-known Parameter Server (PS) framework, each server node handles a separate part of parameters and updates them independently. AMS goes beyond this and is designed to be capable of learning a unified image descriptor model shared by all server nodes which embeds large images into low dimensional high level features before transmitting images to worker nodes. AMS thus dramatically reduces the communication load and enables the arduous joint training process. Based on AMS, the methods of effectively combining the images and ID features are carefully studied, and then we propose a Deep Image CTR Model. Our approach is shown to achieve significant improvements in both online and offline evaluations, and has been deployed in Taobao display advertising system serving the main traffic.

Click-through rate prediction is an essential task in industrial applications, such as online advertising. Recently deep learning based models have been proposed, which follow a similar Embedding&MLP paradigm. In these methods large scale sparse input features are first mapped into low dimensional embedding vectors, and then transformed into fixed-length vectors in a group-wise manner, finally concatenated together to fed into a multilayer perceptron (MLP) to learn the nonlinear relations among features. In this way, user features are compressed into a fixed-length representation vector, in regardless of what candidate ads are. The use of fixed-length vector will be a bottleneck, which brings difficulty for Embedding&MLP methods to capture user's diverse interests effectively from rich historical behaviors. In this paper, we propose a novel model: Deep Interest Network (DIN) which tackles this challenge by designing a local activation unit to adaptively learn the representation of user interests from historical behaviors with respect to a certain ad. This representation vector varies over different ads, improving the expressive ability of model greatly. Besides, we develop two techniques: mini-batch aware regularization and data adaptive activation function which can help training industrial deep networks with hundreds of millions of parameters. Experiments on two public datasets as well as an Alibaba real production dataset with over 2 billion samples demonstrate the effectiveness of proposed approaches, which achieve superior performance compared with state-of-the-art methods. DIN now has been successfully deployed in the online display advertising system in Alibaba, serving the main traffic.

Estimating post-click conversion rate (CVR) accurately is crucial for ranking systems in industrial applications such as recommendation and advertising. Conventional CVR modeling applies popular deep learning methods and achieves state-of-the-art performance. However it encounters several task-specific problems in practice, making CVR modeling challenging. For example, conventional CVR models are trained with samples of clicked impressions while utilized to make inference on the entire space with samples of all impressions. This causes a sample selection bias problem. Besides, there exists an extreme data sparsity problem, making the model fitting rather difficult. In this paper, we model CVR in a brand-new perspective by making good use of sequential pattern of user actions, i.e., impression -> click -> conversion. The proposed Entire Space Multi-task Model (ESMM) can eliminate the two problems simultaneously by i) modeling CVR directly over the entire space, ii) employing a feature representation transfer learning strategy. Experiments on dataset gathered from Taobao's recommender system demonstrate that ESMM significantly outperforms competitive methods. We also release a sampling version of this dataset to enable future research. To the best of our knowledge, this is the first public dataset which contains samples with sequential dependence of click and conversion labels for CVR modeling.

Abstract Easy and efficient 3D modelling in virtual environments is an important and unsolved topic. This paper proposes a new modelling approach to tackle this issue. It invents convolution surface‐based brushes to directly draw 3D models in head‐mounted display‐based virtual environments. In order to maximize the efficiency, flexibility, and capacity of our proposed modelling approach, we propose three different skeleton‐based convolution surfaces to tackle different modelling tasks: point skeleton‐based convolution surfaces for metaball shapes, line skeleton‐based convolution surfaces for cylindrical shapes, and polygon skeleton‐based convolution surfaces for planar surfaces. Their combination makes 3D modelling more flexible and powerful. The high efficiency is further raised by our developed closed‐form solutions for point skeletons, ends of line skeletons, and edges of polygonal skeletons. Different user‐friendly sweeping schemes are provided to facilitate intuitive inputs for various complex shape generation. Unlike Google's Tilt Brush, which is used to create disconnected sheet‐like surfaces only, our proposed convolution surface‐based brushes can produce smoothly blended manifold surfaces, and novice users can easily learn and use them to create various interesting 3D models efficiently.

Abstract We present a novel approach to efficiently modelling branch structures with high‐quality meshes. Our approach has the following advantages. First, the limit surface can fit the target skeleton models as tightly as possible by reversely calculating the control vertices of subdivision surfaces. Second, high performance is achieved through our proposed analytical solutions and the parallel subdivision scheme on a graphics processing unit. Third, a smooth manifold quad‐only mesh is produced from the adopted Catmull–Clark scheme. A number of examples are given to demonstrate applications of our approach in various branch structures, such as tree branches, animal torsos, and vasculatures. Copyright © 2013 John Wiley & Sons, Ltd.

Particle systems' greatest advantage is well suited for modeling complex fuzzy phenomena, such as explosions, fountain, tornado and fireworks, etc. in 3D graphics. With the increasing requirements on the number of particles and particle-particle interactions, the computational complexity of simulation in particle systems has increased rapidly. Particle systems are traditionally implemented on a general-purpose CPU, and the computational complexity of particle systems limits the number of particles that can be computed at interactive rates. This paper focuses on real-time simulation of large-scale particle systems. We discuss optional integration algorithms based on CUDA (Compute Unified Device Architecture) for both graphic and scientific simulation. The speed of particle systems has been greatly improved, with parallel-core GPUs working in tandem with multi-core CPUs. In order to provide a scalable and portable API library, the object-oriented programming method is adopted to encapsulate the functions of parallel particle system. Results show that our proposed APIs are user-friendly and the parallel implementations are significantly efficient.

In this paper, we show how to learn a good similarity metric for SVM classification. We present a novel approach to simultaneously learn a Mahalanobis similarity metric and an SVM classifier. Different from previous approaches, we optimize the Mahalanobis metric directly for minimizing the SVM classification error. Our formulation generalizes the traditional large margin principle used in standard SVM, that is, we maximize the margin-radius-ratio. The learned similarity metric significantly improves the classification performance of standard SVM. Empirical studies on real datasets show the proposed approach achieves higher or comparable classification accuracies compared with state-of-the-art similarity learning methods.