We present, GeniePath, a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In we propose an path layer consists two complementary functions designed breadth and depth exploration respectively, where the former learns importance different sized neighborhoods, while latter extracts filters signals aggregated from neighbors hops away. Our method works in both transductive inductive settings, extensive experiments compared...

Points of interest (POI) recommendation has been drawn much attention recently due to the increasing popularity location-based networks, e.g., Foursquare and Yelp. Among existing approaches POI recommendation, Matrix Factorization (MF) based techniques have proven be effective. However, MF suffer from two major problems: (1) Expensive computations storages centralized model training mechanism: learners maintain whole user-item rating matrix, potentially huge low rank matrices. (2) Privacy...

Factorization Machines offer good performance and useful embeddings of data. However, they are costly to scale large amounts data numbers features. In this paper we describe DiFacto, which uses a refined Machine model with sparse memory adaptive constraints frequency regularization. We show how distribute DiFacto over multiple machines using the Parameter Server framework by computing distributed subgradients on minibatches asynchronously. analyze its convergence demonstrate efficiency in...

Large language models (LLMs) open up new horizons for sequential recommendations, owing to their remarkable comprehension and generation capabilities. However, there are still numerous challenges that should be addressed successfully implement recommendations empowered by LLMs. Firstly, user behavior patterns often complex, relying solely on one-step reasoning from LLMs may lead incorrect or task-irrelevant responses. Secondly, the prohibitively resource requirements of LLM (e.g.,...

In order to solve the problems of low efficiency and long running time caused by traditional Zernike moment method for convolution calculation whole image, this paper combines canny detection algorithm with method. First, edge algorithm, which combined Otsu threshold method, is used extract pixel image. Then an improved Hough transform fit geometric in Based on this, applied realize sub-pixel positioning images. The improves deficiencies direct detection, improving accuracy reducing time. To...

Several sampling algorithms with variance reduction have been proposed for accelerating the training of Graph Convolution Networks (GCNs). However, due to intractable computation optimal distribution, these are suboptimal GCNs and not applicable more general graph neural networks (GNNs) where message aggregator contains learned weights rather than fixed weights, such as Attention (GAT). The fundamental reason is that embeddings neighbors or involved in distribution changing during known a...

Recommendation systems are widely used in e-commerce websites and online platforms to address information overload. However, existing primarily rely on historical data user feedback, making it difficult capture intent transitions. Recently, Knowledge Base (KB)-based models proposed incorporate expert knowledge, but struggle adapt new items the evolving environment. To these challenges, we propose a novel Large Language Model based Complementary Enhanced System (LLM-KERec). It introduces an...

Mobile payment such as Alipay has been widely used in our daily lives. To further promote the mobile activities, it is important to run marketing campaigns under a limited budget by providing incentives coupons, commissions merchants. As result, incentive optimization key maximizing commercial objective of campaign. With analyses online experiments, we found that transaction network can subtly describe similarity merchants' responses different incentives, which great use problem. In this...

Abstract Prostate cancer remains a complex and challenging disease, necessitating innovative approaches for prognosis therapeutic guidance. This study integrates machine learning techniques to develop novel mitophagy-related long non-coding RNA (lncRNA) signature predicting the progression of prostate cancer. Leveraging TCGA-PRAD dataset, we identify set four key lncRNAs formulate riskscore, revealing its potential as prognostic indicator. Subsequent analyses unravel intricate connections...

Recent studies have revealed that GNNs are vulnerable to adversarial attacks. Most existing robust graph learning methods measure model robustness based on label information, rendering them infeasible when information is not available. A straightforward direction employ the widely used Infomax technique from typical Unsupervised Graph Representation Learning (UGRL) learn unsupervised representations. Nonetheless, directly transplanting UGRL may involve a biased assumption. In light of...

We study feature propagation on graph, an inference process involved in graph representation learning tasks. It's to spread the features over whole $t$-th orders, thus expand end's features. The has been successfully adopted embedding or neural networks, however few works studied convergence of propagation. Without guarantees, it may lead unexpected numerical overflows and task failures. In this paper, we first define concept formally, then its conditions equilibrium states. further link...

In this paper, we study the digital marketing where officers (MOs) have to commit creating brand new promotion ads/contents based on understandings of users' needs or preferences. Users' behaviors are typically high dimensional and hard understand. Therefore, dimension reduction from dimensions explainability important help MOs launch operation-friendly marketings. As such, it is natural exploit topic models understand intents (e.g., user-item visits) in case treat each user as a document...

A large body of work in Computer Science and Operations Research study online algorithms for stochastic resource allocation problems. The most common assumption is that the requests have randomly generated i.i.d. types. This well justified static markets and/or relatively short time periods. We consider dynamic markets, whose states evolve as a random walk market-specific Markov Chain. new model generalizes previous settings. identify important parameters chain crucial obtaining good...

Improving the performance of large language models (LLMs) in complex question-answering (QA) scenarios has always been a research focal point. Recent studies have attempted to enhance LLMs' by combining step-wise planning with external retrieval. While effective for advanced like GPT-3.5, smaller LLMs face challenges decomposing questions, necessitating supervised fine-tuning. Previous work relied on manual annotation and knowledge distillation from teacher LLMs, which are time-consuming not...

Semi-supervised graph learning aims to improve performance by leveraging unlabeled nodes. Typically, it can be approached in two different ways, including <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">predictive representation learning</i> (PRL) where data provide clues on input distribution and xmlns:xlink="http://www.w3.org/1999/xlink">label-dependent regularization</i> (LDR) which smooths the output with nodes generalization. However,...

Recommendation systems are widely used in e-commerce websites and online platforms to address information overload. However, existing primarily rely on historical data user feedback, making it difficult capture intent transitions. Recently, Knowledge Base (KB)-based models proposed incorporate expert knowledge, but struggle adapt new items the evolving environment. To these challenges, we propose a novel Large Language Model based Complementary Enhanced System (LLM-KERec). It introduces an...

In this paper, the automatic pricing and replenishment decision-making problem for vegetable category goods is studied. Using K-means clustering model, random forest, LSTM long short-term memory nonlinear programming, distribution law interrelationship of each single product vegetables are analyzed in detail based on commodity information, sales flow details, wholesale price, loss rate data category. Firstly, correlation between volume categories seasons was revealed by analyzing total time,...

The causal dependence in data is often characterized by Directed Acyclic Graphical (DAG) models, widely used many areas. Causal discovery aims to recover the DAG structure using observational data. This paper focuses on with multi-variate count We are motivated real-world web visit data, recording individual user visits multiple websites. Building a diagram can help understand behavior transitioning between websites, inspiring operational strategy. A challenge modeling heterogeneity, as...