Sequential recommendation methods play a crucial role in modern recommender systems because of their ability to capture user's dynamic interest from her/his historical interactions. Despite success, we argue that these approaches usually rely on the sequential prediction task optimize huge amounts parameters. They suffer data sparsity problem, which makes it difficult for them learn high-quality user representations. To tackle that, inspired by recent advances contrastive learning techniques...

Newly emerging location-based and event-based social network services provide us with a new platform to understand users' preferences based on their activity history. A user can only visit limited number of venues/events most them are within distance range, so the user-item matrix is very sparse, which creates big challenge for traditional collaborative filtering-based recommender systems. The problem becomes more challenging when people travel city where they have no

The success of "infinite-inventory" retailers such as Amazon.com and Netflix has been largely attributed to a "long tail" phenomenon. Although the majority their inventory is not in high demand, these niche products, unavailable at limited-inventory competitors, generate significant fraction total revenue aggregate. In addition, tail product availability can boost head sales by offering consumers convenience "one-stop shopping" for both mainstream tastes. However, most existing recommender...

Recently, zero-shot and few-shot learning via Contrastive Vision-Language Pre-training (CLIP) have shown inspirational performance on 2D visual recognition, which learns to match images with their corresponding texts in open-vocabulary settings. However, it remains under explored that whether CLIP, pre-trained by large-scale image-text pairs 2D, can be generalized 3D recognition. In this paper, we identify such a setting is feasible proposing PointCLIP, conducts alignment between...

Point-of-Interest recommendation is an essential means to help people discover attractive locations, especially when travel out of town or unfamiliar regions. While a growing line research has focused on modeling user geographical preferences for POI recommendation, they ignore the phenomenon interest drift across regions, i.e., users tend have different interests in which discounts quality existing methods, out-of-town users. In this paper, we propose latent class probabilistic generative...

We study distributed machine learning in heterogeneous environments this work. first conduct a systematic of existing systems running stochastic gradient descent; we find that, although these work well homogeneous environments, they can suffer performance degradation, sometimes up to 10x, where stragglers are common because their synchronization protocols cannot fit setting. Our contribution is heterogeneity-aware algorithm that uses constant rate schedule for updates before adding them the...

Recommender systems play a significant role in information filtering and have been utilized different scenarios, such as e-commerce social media. With the prosperity of deep learning, recommender show superior performance by capturing non-linear item-user relationships. However, design heavily relies on human experiences expert knowledge. To tackle this problem, Automated Machine Learning (AutoML) is introduced to automatically search for proper candidates parts systems. This survey performs...

The development of Artificial Intelligence Generated Content (AIGC) has been facilitated by advancements in model algorithms, the increasing scale foundation models, and availability ample high-quality datasets. While AIGC achieved remarkable performance, it still faces several challenges, such as difficulty maintaining up-to-date long-tail knowledge, risk data leakage, high costs associated with training inference. Retrieval-Augmented Generation(RAG) recently emerged a paradigm to address...

Point-of-Interest (POI) recommendation has become an important means to help people discover attractive and interesting places, especially when users travel out of town. However, the extreme sparsity a user-POI matrix creates severe challenge. To cope with this challenge, we propose unified probabilistic generative model, Topic-Region Model (TRM) , simultaneously semantic, temporal, spatial patterns users’ check-in activities, model their joint effect on decision making for selection POIs...

Social media provides valuable resources to analyze user behaviors and capture preferences. This article focuses on analyzing in social systems designing a latent class statistical mixture model, named temporal context-aware model (TCAM), account for the intentions preferences behind behaviors. Based observation that of are generally influenced by intrinsic interest as well context (e.g., public's attention at time), TCAM simultaneously models topics related users' interests then combines...

Newly emerging location-based and event-based social network services provide us with a new platform to understand users' preferences based on their activity history. A user can only visit limited number of venues/events most them are within distance range, so the user-item matrix is very sparse, which creates big challenge traditional collaborative filtering-based recommender systems. The problem becomes even more challenging when people travel city where they have no information. In this...

This article proposes LA-LDA, a location-aware probabilistic generative model that exploits location-based ratings to user profiles and produce recommendations. Most of the existing recommendation models do not consider spatial information users or items; however, LA-LDA supports three classes ratings, namely for nonspatial items, items. consists two components, ULA-LDA ILA-LDA, which are designed take into account item location information, respectively. The component explicitly...

Social media provides valuable resources to analyze user behaviors and capture preferences. This paper focuses on analyzing in social systems designing a latent class statistical mixture model, named temporal context-aware model (TCAM), account for the intentions preferences behind behaviors. Based observation that of are generally influenced by intrinsic interest as well context (e.g., public's attention at time), TCAM simultaneously models topics related users' interests then combines...

Subgraph listing is a fundamental operation to many graph and network analyses. The problem itself computationally expensive well-studied in centralized processing algorithms. However, the solutions cannot scale well large graphs. Recently, several parallel approaches are introduced handle Unfortunately, these still rely on join operations, thus achieve high performance. In this paper, we design novel subgraph framework, named PSgL. PSgL iteratively enumerates instances solves...

Approximate stream processing algorithms, such as Count-Min sketch, Space-Saving, etc., support numerous applications in databases, storage systems, networking, and other domains. However, the unbalanced distribution real data streams poses great challenges to existing algorithms. To enhance these we propose a meta-framework, called Cold Filter (CF), that enables faster more accurate processing.

With the explosive growth of online information, recommender systems play a key role to alleviate such information overload. Due important application value systems, there have always been emerging works in this field. In main challenge is learn effective user/item representations from their interactions and side (if any). Recently, graph neural network (GNN) techniques widely utilized since most essentially has structure GNN superiority representation learning. This article aims provide...

To address the challenge of explosive big data, distributed machine learning (ML) has drawn interests many researchers. Since ML algorithms trained by stochastic gradient descent (SGD) involve communicating gradients through network, it is important to compress transferred gradient. A category low-precision can significantly reduce size gradients, at expense some precision loss. However, existing methods are not suitable for cases where sparse and nonuniformly distributed. In this paper, we...

E-commerce platforms are becoming a primary place for people to find, compare and ultimately purchase products. One of the fundamental questions that arises in e-commerce is predict user purchasing intent, which an important part understanding allows providing better services both sellers customers. However, previous work cannot real-time intent with high accuracy, limited by representation capability traditional browse-interactive behavior adopted. In this paper, we propose novel end-to-end...

Modern database management systems (DBMS) contain tens to hundreds of critical performance tuning knobs that determine the system runtime behaviors. To reduce total cost ownership, cloud providers put in drastic effort automatically optimize resource utilization by these knobs. There are two challenges. First, should always abide service level agreement (SLA) while optimizing utilization, which imposes strict constrains on process. Second, time be reasonably acceptable since time-consuming...

Graph neural networks (GNNs) have achieved great success in many graph-based applications. However, the enormous size and high sparsity level of graphs hinder their applications under industrial scenarios. Although some scalable GNNs are proposed for large-scale graphs, they adopt a fixed $K$-hop neighborhood each node, thus facing over-smoothing issue when adopting large propagation depths nodes within sparse regions. To tackle above issue, we propose new GNN architecture -- Attention...

Contrastive Language-Image Pre-training (CLIP) has been shown to learn visual representations with promising zero-shot performance. To further improve its downstream accuracy, existing works propose additional learnable modules upon CLIP and fine-tune them by few-shot training sets. However, the resulting extra cost data requirement severely hinder efficiency for model deployment knowledge transfer. In this paper, we introduce a free-lunch enhancement method, CALIP, boost CLIP's performance...