An Enhanced-State Reinforcement Learning Algorithm for Multi-Task Fusion in Large-Scale Recommender Systems

As the last key stage of Recommender Systems (RSs), Multi-Task Fusion (MTF) is in charge of combining multiple scores predicted by Multi-Task Learning (MTL) into a final score to maximize user satisfaction, which decides the ultimate recommendation results. In recent years, to maximize long-term user satisfaction within a recommendation session, Reinforcement Learning (RL) is widely used for MTF in large-scale RSs. However, limited by their modeling pattern, all the current RL-MTF methods can only utilize user features as the state to generate actions for each user, but unable to make use of item features and other valuable features, which leads to suboptimal results. Addressing this problem is a challenge that requires breaking through the current modeling pattern of RL-MTF. To solve this problem, we propose a novel method called Enhanced-State RL for MTF in RSs. Unlike the existing methods mentioned above, our method first defines user features, item features, and other valuable features collectively as the enhanced state; then proposes a novel actor and critic learning process to utilize the enhanced state to make much better action for each user-item pair. To the best of our knowledge, this novel modeling pattern is being proposed for the first time in the field of RL-MTF. We conduct extensive offline and online experiments in a large-scale RS. The results demonstrate that our model outperforms other models significantly. Enhanced-State RL has been fully deployed in our RS more than half a year, improving +3.84% user valid consumption and +0.58% user duration time compared to baseline.

P/D-Serve: Serving Disaggregated Large Language Model at Scale

Serving disaggregated large language models (LLMs) over tens of thousands of xPU devices (GPUs or NPUs) with reliable performance faces multiple challenges. 1) Ignoring the diversity (various prefixes and tidal requests), treating all the prompts in a mixed pool is inadequate. To facilitate the similarity per scenario and minimize the inner mismatch on P/D (prefill and decoding) processing, fine-grained organization is required, dynamically adjusting P/D ratios for better performance. 2) Due to inaccurate estimation on workload (queue status or maintained connections), the global scheduler easily incurs unnecessary timeouts in prefill. 3) Block-fixed device-to-device (D2D) KVCache transfer over cluster-level RDMA (remote direct memory access) fails to achieve desired D2D utilization as expected. To overcome previous problems, this paper proposes an end-to-end system P/D-Serve, complying with the paradigm of MLOps (machine learning operations), which models end-to-end (E2E) P/D performance and enables: 1) fine-grained P/D organization, mapping the service with RoCE (RDMA over converged ethernet) as needed, to facilitate similar processing and dynamic adjustments on P/D ratios; 2) on-demand forwarding upon rejections for idle prefill, decoupling the scheduler from regular inaccurate reports and local queues, to avoid timeouts in prefill; and 3) efficient KVCache transfer via optimized D2D access. P/D-Serve is implemented upon Ascend and MindSpore, has been deployed over tens of thousands of NPUs for more than eight months in commercial use, and further achieves 60\%, 42\% and 46\% improvements on E2E throughput, time-to-first-token (TTFT) SLO (service level objective) and D2D transfer time. As the E2E system with optimizations, P/D-Serve achieves 6.7x increase on throughput, compared with aggregated LLMs.

An Off-Policy Reinforcement Learning Algorithm Customized for Multi-Task Fusion in Large-Scale Recommender Systems

Recommender Systems (RSs) are widely used to provide personalized recommendation service. As the last critical stage of RSs, Multi-Task Fusion (MTF) is responsible for combining multiple scores outputted by Multi-Task Learning (MTL) into a final score to maximize user satisfaction, which determines the ultimate recommendation results. Recently, to optimize long-term user satisfaction within a recommendation session, Reinforcement Learning (RL) is used for MTF in the industry. However, the off-policy RL algorithms used for MTF so far have the following severe problems: 1) to avoid out-of-distribution (OOD) problem, their constraints are overly strict, which seriously damage their performance; 2) they are unaware of the exploration policy used for producing training data and never interact with real environment, so only suboptimal policy can be learned; 3) the traditional exploration policies are inefficient and hurt user experience. To solve the above problems, we propose a novel off-policy RL algorithm customized for MTF in large-scale RSs. Our RL-MTF algorithm integrates off-policy RL model with our online exploration policy to relax overstrict and complicated constraints, which significantly improves the performance of our RL model. We also design an extremely efficient exploration policy, which eliminates low-value exploration space and focuses on exploring potential high-value state-action pairs. Moreover, we adopt progressive training mode to further enhance our RL model's performance with the help of our exploration policy. We conduct extensive offline and online experiments in the short video channel of Tencent News. The results demonstrate that our RL-MTF model outperforms other models remarkably. Our RL-MTF model has been fully deployed in the short video channel of Tencent News for about one year. In addition, our solution has been used in other large-scale RSs in Tencent.

Alleviating neighbor bias: augmenting graph self-supervise learning with structural equivalent positive samples

In recent years, using a self-supervised learning framework to learn the general characteristics of graphs has been considered a promising paradigm for graph representation learning. The core of self-supervised learning strategies for graph neural networks lies in constructing suitable positive sample selection strategies. However, existing GNNs typically aggregate information from neighboring nodes to update node representations, leading to an over-reliance on neighboring positive samples, i.e., homophilous samples; while ignoring long-range positive samples, i.e., positive samples that are far apart on the graph but structurally equivalent samples, a problem we call "neighbor bias." This neighbor bias can reduce the generalization performance of GNNs. In this paper, we argue that the generalization properties of GNNs should be determined by combining homogeneous samples and structurally equivalent samples, which we call the "GC combination hypothesis." Therefore, we propose a topological signal-driven self-supervised method. It uses a topological information-guided structural equivalence sampling strategy. First, we extract multiscale topological features using persistent homology. Then we compute the structural equivalence of node pairs based on their topological features. In particular, we design a topological loss function to pull in non-neighboring node pairs with high structural equivalence in the representation space to alleviate neighbor bias. Finally, we use the joint training mechanism to adjust the effect of structural equivalence on the model to fit datasets with different characteristics. We conducted experiments on the node classification task across seven graph datasets. The results show that the model performance can be effectively improved using a strategy of topological signal enhancement.

Augmentation-Free Graph Contrastive Learning of Invariant-Discriminative Representations

The pretasks are mainly built on mutual information estimation, which requires data augmentation to construct positive samples with similar semantics to learn invariant signals and negative samples with dissimilar semantics in order to empower representation discriminability. However, an appropriate data augmentation configuration depends heavily on lots of empirical trials such as choosing the compositions of data augmentation techniques and the corresponding hyperparameter settings. We propose an augmentation-free graph contrastive learning method, invariant-discriminative graph contrastive learning (iGCL), that does not intrinsically require negative samples. iGCL designs the invariant-discriminative loss (ID loss) to learn invariant and discriminative representations. On the one hand, ID loss learns invariant signals by directly minimizing the mean square error between the target samples and positive samples in the representation space. On the other hand, ID loss ensures that the representations are discriminative by an orthonormal constraint forcing the different dimensions of representations to be independent of each other. This prevents representations from collapsing to a point or subspace. Our theoretical analysis explains the effectiveness of ID loss from the perspectives of the redundancy reduction criterion, canonical correlation analysis, and information bottleneck principle. The experimental results demonstrate that iGCL outperforms all baselines on 5 node classification benchmark datasets. iGCL also shows superior performance for different label ratios and is capable of resisting graph attacks, which indicates that iGCL has excellent generalization and robustness.

Curvature Graph Neural Network

Graph neural networks (GNNs) have achieved great success in many graph-based tasks. Much work is dedicated to empowering GNNs with the adaptive locality ability, which enables measuring the importance of neighboring nodes to the target node by a node-specific mechanism. However, the current node-specific mechanisms are deficient in distinguishing the importance of nodes in the topology structure. We believe that the structural importance of neighboring nodes is closely related to their importance in aggregation. In this paper, we introduce discrete graph curvature (the Ricci curvature) to quantify the strength of structural connection of pairwise nodes. And we propose Curvature Graph Neural Network (CGNN), which effectively improves the adaptive locality ability of GNNs by leveraging the structural property of graph curvature. To improve the adaptability of curvature to various datasets, we explicitly transform curvature into the weights of neighboring nodes by the necessary Negative Curvature Processing Module and Curvature Normalization Module. Then, we conduct numerous experiments on various synthetic datasets and real-world datasets. The experimental results on synthetic datasets show that CGNN effectively exploits the topology structure information, and the performance is improved significantly. CGNN outperforms the baselines on 5 dense node classification benchmark datasets. This study deepens the understanding of how to utilize advanced topology information and assign the importance of neighboring nodes from the perspective of graph curvature and encourages us to bridge the gap between graph theory and neural networks.

Graph Information Vanishing Phenomenon inImplicit Graph Neural Networks

One of the key problems of GNNs is how to describe the importance of neighbor nodes in the aggregation process for learning node representations. A class of GNNs solves this problem by learning implicit weights to represent the importance of neighbor nodes, which we call implicit GNNs such as Graph Attention Network. The basic idea of implicit GNNs is to introduce graph information with special properties followed by Learnable Transformation Structures (LTS) which encode the importance of neighbor nodes via a data-driven way. In this paper, we argue that LTS makes the special properties of graph information disappear during the learning process, resulting in graph information unhelpful for learning node representations. We call this phenomenon Graph Information Vanishing (GIV). Also, we find that LTS maps different graph information into highly similar results. To validate the above two points, we design two sets of 70 random experiments on five Implicit GNNs methods and seven benchmark datasets by using a random permutation operator to randomly disrupt the order of graph information and replacing graph information with random values. We find that randomization does not affect the model performance in 93\% of the cases, with about 7 percentage causing an average 0.5\% accuracy loss. And the cosine similarity of output results, generated by LTS mapping different graph information, over 99\% with an 81\% proportion. The experimental results provide evidence to support the existence of GIV in Implicit GNNs and imply that the existing methods of Implicit GNNs do not make good use of graph information. The relationship between graph information and LTS should be rethought to ensure that graph information is used in node representation.

KST-GCN: A Knowledge-Driven Spatial-Temporal Graph Convolutional Network for Traffic Forecasting

When considering the spatial and temporal features of traffic, capturing the impacts of various external factors on travel is an important step towards achieving accurate traffic forecasting. The impacts of external factors on the traffic flow have complex correlations. However, existing studies seldom consider external factors or neglecting the effect of the complex correlations among external factors on traffic. Intuitively, knowledge graphs can naturally describe these correlations, but knowledge graphs and traffic networks are essentially heterogeneous networks; thus, it is a challenging problem to integrate the information in both networks. We propose a knowledge representation-driven traffic forecasting method based on spatiotemporal graph convolutional networks. We first construct a city knowledge graph for traffic forecasting, then use KS-Cells to combine the information from the knowledge graph and the traffic network, and finally, capture the temporal changes of the traffic state with GRU. Testing on real-world datasets shows that the KST-GCN has higher accuracy than the baseline traffic forecasting methods at various prediction horizons. We provide a new way to integrate knowledge and the spatiotemporal features of data for traffic forecasting tasks. Without any loss of generality, the proposed method can also be extended to other spatiotemporal forecasting tasks.

AST-GCN: Attribute-Augmented Spatiotemporal Graph Convolutional Network for Traffic Forecasting

Traffic forecasting is a fundamental and challenging task in the field of intelligent transportation. Accurate forecasting not only depends on the historical traffic flow information but also needs to consider the influence of a variety of external factors, such as weather conditions and surrounding POI distribution. Recently, spatiotemporal models integrating graph convolutional networks and recurrent neural networks have become traffic forecasting research hotspots and have made significant progress. However, few works integrate external factors. Therefore, based on the assumption that introducing external factors can enhance the spatiotemporal accuracy in predicting traffic and improving interpretability, we propose an attribute-augmented spatiotemporal graph convolutional network (AST-GCN). We model the external factors as dynamic attributes and static attributes and design an attribute-augmented unit to encode and integrate those factors into the spatiotemporal graph convolution model. Experiments on real datasets show the effectiveness of considering external information on traffic forecasting tasks when compared to traditional traffic prediction methods. Moreover, under different attribute-augmented schemes and prediction horizon settings, the forecasting accuracy of the AST-GCN is higher than that of the baselines.

RS-MetaNet: Deep meta metric learning for few-shot remote sensing scene classification

Training a modern deep neural network on massive labeled samples is the main paradigm in solving the scene classification problem for remote sensing, but learning from only a few data points remains a challenge. Existing methods for few-shot remote sensing scene classification are performed in a sample-level manner, resulting in easy overfitting of learned features to individual samples and inadequate generalization of learned category segmentation surfaces. To solve this problem, learning should be organized at the task level rather than the sample level. Learning on tasks sampled from a task family can help tune learning algorithms to perform well on new tasks sampled in that family. Therefore, we propose a simple but effective method, called RS-MetaNet, to resolve the issues related to few-shot remote sensing scene classification in the real world. On the one hand, RS-MetaNet raises the level of learning from the sample to the task by organizing training in a meta way, and it learns to learn a metric space that can well classify remote sensing scenes from a series of tasks. We also propose a new loss function, called Balance Loss, which maximizes the generalization ability of the model to new samples by maximizing the distance between different categories, providing the scenes in different categories with better linear segmentation planes while ensuring model fit. The experimental results on three open and challenging remote sensing datasets, UCMerced\_LandUse, NWPU-RESISC45, and Aerial Image Data, demonstrate that our proposed RS-MetaNet method achieves state-of-the-art results in cases where there are only 1-20 labeled samples.