Search

Large language models (LLM) have prioritized expanding the context window from which models can incorporate more information. However, training models to handle long contexts presents significant challenges. These include the scarcity of high-quality natural long-context data, the potential for performance degradation on short-context tasks, and the reduced training efficiency associated with attention mechanisms. In this paper, we introduce Untie the Knots (\textbf{UtK}), a novel data augmentation strategy employed during the continue pre-training phase, designed to efficiently enable LLMs to gain long-context capabilities without the need to modify the existing data mixture. In particular, we chunk the documents, shuffle the chunks, and create a complex and knotted structure of long texts; LLMs are then trained to untie these knots and identify relevant segments within seemingly chaotic token sequences. This approach greatly improves the model's performance by accurately attending to relevant information in long context and the training efficiency is also largely increased. We conduct extensive experiments on models with 7B and 72B parameters, trained on 20 billion tokens, demonstrating that UtK achieves 75\% and 84.5\% accurracy on RULER at 128K context length, significantly outperforming other long context strategies. The trained models will open-source for further research.

We investigate the problem of carving an $n$-face triangulated three-dimensional polytope using a tool to make cuts modelled by either a half-plane or sweeps from an infinite ray. In the case of half-planes cuts, we present a deterministic algorithm running in $O(n^2)$ time and a randomized algorithm running in $O(n^{3/2+\varepsilon})$ expected time for any $\varepsilon>0$. In the case of cuts defined by sweeps of infinite rays, we present an algorithm running in $O(n^5)$ time., Comment: Appeared in CCCG 2024. 10 pages, 7 figures

We introduce a new balanced separator theorem for unit-disk graphs involving two shortest paths combined with the 1-hop neighbours of those paths and two other vertices. This answers an open problem of Yan, Xiang and Dragan [CGTA '12] and improves their result that requires removing the 3-hop neighborhood of two shortest paths. Our proof uses very different ideas, including Delaunay triangulations and a generalization of the celebrated balanced separator theorem of Lipton and Tarjan [J. Appl. Math. '79] to systems of non-intersecting paths., Comment: To appear in ESA 2024. 15 pages, 7 figures

In the Directed Steiner Tree (DST) problem the input is a directed edge-weighted graph $G=(V,E)$, a root vertex $r$ and a set $S \subseteq V$ of $k$ terminals. The goal is to find a min-cost subgraph that connects $r$ to each of the terminals. DST admits an $O(\log^2 k/\log \log k)$-approximation in quasi-polynomial time, and an $O(k^{\epsilon})$-approximation for any fixed $\epsilon > 0$ in polynomial-time. Resolving the existence of a polynomial-time poly-logarithmic approximation is a major open problem in approximation algorithms. In a recent work, Friggstad and Mousavi [ICALP 2023] obtained a simple and elegant polynomial-time $O(\log k)$-approximation for DST in planar digraphs via Thorup's shortest path separator theorem. We build on their work and obtain several new results on DST and related problems. - We develop a tree embedding technique for rooted problems in planar digraphs via an interpretation of the recursion in Friggstad and Mousavi [ICALP 2023]. Using this we obtain polynomial-time poly-logarithmic approximations for Group Steiner Tree, Covering Steiner Tree, and the Polymatroid Steiner Tree problems in planar digraphs. All these problems are hard to approximate to within a factor of $\Omega(\log^2 n/\log \log n)$ even in trees. - We prove that the natural cut-based LP relaxation for DST has an integrality gap of $O(\log^2 k)$ in planar graphs. This is in contrast to general graphs where the integrality gap of this LP is known to be $\Omega(k)$ and $\Omega(n^{\delta})$ for some fixed $\delta > 0$. - We combine the preceding results with density based arguments to obtain poly-logarithmic approximations for the multi-rooted versions of the problems in planar digraphs. For DST our result improves the $O(R + \log k)$ approximation of Friggstad and Mousavi [ICALP 2023] when $R= \omega(\log^2 k)$.

Text-rich graphs, prevalent in data mining contexts like e-commerce and academic graphs, consist of nodes with textual features linked by various relations. Traditional graph machine learning models, such as Graph Neural Networks (GNNs), excel in encoding the graph structural information, but have limited capability in handling rich text on graph nodes. Large Language Models (LLMs), noted for their superior text understanding abilities, offer a solution for processing the text in graphs but face integration challenges due to their limitation for encoding graph structures and their computational complexities when dealing with extensive text in large neighborhoods of interconnected nodes. This paper introduces ``Hierarchical Compression'' (HiCom), a novel method to align the capabilities of LLMs with the structure of text-rich graphs. HiCom processes text in a node's neighborhood in a structured manner by organizing the extensive textual information into a more manageable hierarchy and compressing node text step by step. Therefore, HiCom not only preserves the contextual richness of the text but also addresses the computational challenges of LLMs, which presents an advancement in integrating the text processing power of LLMs with the structural complexities of text-rich graphs. Empirical results show that HiCom can outperform both GNNs and LLM backbones for node classification on e-commerce and citation graphs. HiCom is especially effective for nodes from a dense region in a graph, where it achieves a 3.48% average performance improvement on five datasets while being more efficient than LLM backbones.

Graph machine learning (GML) is effective in many business applications. However, making GML easy to use and applicable to industry applications with massive datasets remain challenging. We developed GraphStorm, which provides an end-to-end solution for scalable graph construction, graph model training and inference. GraphStorm has the following desirable properties: (a) Easy to use: it can perform graph construction and model training and inference with just a single command; (b) Expert-friendly: GraphStorm contains many advanced GML modeling techniques to handle complex graph data and improve model performance; (c) Scalable: every component in GraphStorm can operate on graphs with billions of nodes and can scale model training and inference to different hardware without changing any code. GraphStorm has been used and deployed for over a dozen billion-scale industry applications after its release in May 2023. It is open-sourced in Github: https://github.com/awslabs/graphstorm.

Text-rich graphs, which exhibit rich textual information on nodes and edges, are prevalent across a wide range of real-world business applications. Large Language Models (LLMs) have demonstrated remarkable abilities in understanding text, which also introduced the potential for more expressive modeling in text-rich graphs. Despite these capabilities, efficiently applying LLMs to representation learning on graphs presents significant challenges. Recently, parameter-efficient fine-tuning methods for LLMs have enabled efficient new task generalization with minimal time and memory consumption. Inspired by this, we introduce Graph-aware Parameter-Efficient Fine-Tuning - GPEFT, a novel approach for efficient graph representation learning with LLMs on text-rich graphs. Specifically, we utilize a graph neural network (GNN) to encode structural information from neighboring nodes into a graph prompt. This prompt is then inserted at the beginning of the text sequence. To improve the quality of graph prompts, we pre-trained the GNN to assist the frozen LLM in predicting the next token in the node text. Compared with existing joint GNN and LMs, our method directly generate the node embeddings from large language models with an affordable fine-tuning cost. We validate our approach through comprehensive experiments conducted on 8 different text-rich graphs, observing an average improvement of 2% in hit@1 and Mean Reciprocal Rank (MRR) in link prediction evaluations. Our results demonstrate the efficacy and efficiency of our model, showing that it can be smoothly integrated with various large language models, including OPT, LLaMA and Falcon.

Polynomial partitioning techniques have recently led to improved geometric data structures for a variety of fundamental problems related to semialgebraic range searching and intersection searching in 3D and higher dimensions (e.g., see [Agarwal, Aronov, Ezra, and Zahl, SoCG 2019; Ezra and Sharir, SoCG 2021; Agarwal, Aronov, Ezra, Katz, and Sharir, SoCG 2022]). They have also led to improved algorithms for offline versions of semialgebraic range searching in 2D, via lens-cutting [Sharir and Zahl (2017)]. In this paper, we show that these techniques can yield new data structures for a number of other 2D problems even for online queries: 1. Semialgebraic range stabbing. We present a data structure for $n$ semialgebraic ranges in 2D of constant description complexity with $O(n^{3/2+\varepsilon})$ preprocessing time and space, so that we can count the number of ranges containing a query point in $O(n^{1/4+\varepsilon})$ time, for an arbitrarily small constant $\varepsilon>0$. 2. Ray shooting amid algebraic arcs. We present a data structure for $n$ algebraic arcs in 2D of constant description complexity with $O(n^{3/2+\varepsilon})$ preprocessing time and space, so that we can find the first arc hit by a query (straight-line) ray in $O(n^{1/4+\varepsilon})$ time. 3. Intersection counting amid algebraic arcs. We present a data structure for $n$ algebraic arcs in 2D of constant description complexity with $O(n^{3/2+\varepsilon})$ preprocessing time and space, so that we can count the number of intersection points with a query algebraic arc of constant description complexity in $O(n^{1/2+\varepsilon})$ time. In particular, this implies an $O(n^{3/2+\varepsilon})$-time algorithm for counting intersections between two sets of $n$ algebraic arcs in 2D., Comment: SOCG 2024

Given a node-attributed graph, and a graph task (link prediction or node classification), can we tell if a graph neural network (GNN) will perform well? More specifically, do the graph structure and the node features carry enough usable information for the task? Our goals are (1) to develop a fast tool to measure how much information is in the graph structure and in the node features, and (2) to exploit the information to solve the task, if there is enough. We propose NetInfoF, a framework including NetInfoF_Probe and NetInfoF_Act, for the measurement and the exploitation of network usable information (NUI), respectively. Given a graph data, NetInfoF_Probe measures NUI without any model training, and NetInfoF_Act solves link prediction and node classification, while two modules share the same backbone. In summary, NetInfoF has following notable advantages: (a) General, handling both link prediction and node classification; (b) Principled, with theoretical guarantee and closed-form solution; (c) Effective, thanks to the proposed adjustment to node similarity; (d) Scalable, scaling linearly with the input size. In our carefully designed synthetic datasets, NetInfoF correctly identifies the ground truth of NUI and is the only method being robust to all graph scenarios. Applied on real-world datasets, NetInfoF wins in 11 out of 12 times on link prediction compared to general GNN baselines., Comment: Accepted to ICLR 2024 (Spotlight)

The question of whether complex numbers play a fundamental role in quantum theory has been debated since the inception of quantum mechanics. Recently, a feasible proposal to differentiate between real and complex quantum theories based on the technique of testing Bell nonlocalities has emerged [Nature 600, 625-629 (2021)]. Based on this method, the real quantum theory has been falsified experimentally in both photonic and superconducting quantum systems [Phys. Rev. Lett. 128, 040402 (2022), Phys. Rev. Lett. 128, 040403 (2022)]. The quantum networks with multiple independent sources which are not causally connected have gained significant interest as they offer new perspective on studying the nonlocalities. The independence of these sources imposes additional constraints on observable covariances and leads to new bounds for classical and quantum correlations. In this study, we examine the discrimination between the real and complex quantum theories with an entanglement swapping scenario under a stronger assumption that the two sources are causally independent, which wasn't made in previous works. Using a revised Navascu\'es-Pironio-Ac\'in method and Bayesian optimization, we find a proposal with optimal coefficients of the correlation function which could give a larger discrimination between the real and quantum theories comparing with the existing proposals. This work opens up avenues for further exploration of the discrimination between real and complex quantum theories within intricate quantum networks featuring causally independent parties.

We present the first optimal randomized algorithm for constructing the order-$k$ Voronoi diagram of $n$ points in two dimensions. The expected running time is $O(n\log n + nk)$, which improves the previous, two-decades-old result of Ramos (SoCG'99) by a $2^{O(\log^*k)}$ factor. To obtain our result, we (i) use a recent decision-tree technique of Chan and Zheng (SODA'22) in combination with Ramos's cutting construction, to reduce the problem to verifying an order-$k$ Voronoi diagram, and (ii) solve the verification problem by a new divide-and-conquer algorithm using planar-graph separators. We also describe a deterministic algorithm for constructing the $k$-level of $n$ lines in two dimensions in $O(n\log n + nk^{1/3})$ time, and constructing the $k$-level of $n$ planes in three dimensions in $O(n\log n + nk^{3/2})$ time. These time bounds (ignoring the $n\log n$ term) match the current best upper bounds on the combinatorial complexity of the $k$-level. Previously, the same time bound in two dimensions was obtained by Chan (1999) but with randomization., Comment: To appear in SODA 2024. 16 pages, 1 figure

Graph Neural Networks (GNNs) are becoming increasingly popular due to their superior performance in critical graph-related tasks. While quantization is widely used to accelerate GNN computation, quantized training faces unprecedented challenges. Current quantized GNN training systems often have longer training times than their full-precision counterparts for two reasons: (i) addressing the accuracy challenge leads to excessive overhead, and (ii) the optimization potential exposed by quantization is not adequately leveraged. This paper introduces Tango which re-thinks quantization challenges and opportunities for graph neural network training on GPUs with three contributions: Firstly, we introduce efficient rules to maintain accuracy during quantized GNN training. Secondly, we design and implement quantization-aware primitives and inter-primitive optimizations that can speed up GNN training. Finally, we integrate Tango with the popular Deep Graph Library (DGL) system and demonstrate its superior performance over state-of-the-art approaches on various GNN models and datasets.

Memory-based Temporal Graph Neural Networks are powerful tools in dynamic graph representation learning and have demonstrated superior performance in many real-world applications. However, their node memory favors smaller batch sizes to capture more dependencies in graph events and needs to be maintained synchronously across all trainers. As a result, existing frameworks suffer from accuracy loss when scaling to multiple GPUs. Evenworse, the tremendous overhead to synchronize the node memory make it impractical to be deployed to distributed GPU clusters. In this work, we propose DistTGL -- an efficient and scalable solution to train memory-based TGNNs on distributed GPU clusters. DistTGL has three improvements over existing solutions: an enhanced TGNN model, a novel training algorithm, and an optimized system. In experiments, DistTGL achieves near-linear convergence speedup, outperforming state-of-the-art single-machine method by 14.5% in accuracy and 10.17x in training throughput., Comment: SC'23

Model pre-training on large text corpora has been demonstrated effective for various downstream applications in the NLP domain. In the graph mining domain, a similar analogy can be drawn for pre-training graph models on large graphs in the hope of benefiting downstream graph applications, which has also been explored by several recent studies. However, no existing study has ever investigated the pre-training of text plus graph models on large heterogeneous graphs with abundant textual information (a.k.a. large graph corpora) and then fine-tuning the model on different related downstream applications with different graph schemas. To address this problem, we propose a framework of graph-aware language model pre-training (GALM) on a large graph corpus, which incorporates large language models and graph neural networks, and a variety of fine-tuning methods on downstream applications. We conduct extensive experiments on Amazon's real internal datasets and large public datasets. Comprehensive empirical results and in-depth analysis demonstrate the effectiveness of our proposed methods along with lessons learned., Comment: To be published in the KDD 2023 proceedings as a full paper

The maximization of submodular functions have found widespread application in areas such as machine learning, combinatorial optimization, and economics, where practitioners often wish to enforce various constraints; the matroid constraint has been investigated extensively due to its algorithmic properties and expressive power. Recent progress has focused on fast algorithms for important classes of matroids given in explicit form. Currently, nearly-linear time algorithms only exist for graphic and partition matroids [ICALP '19]. In this work, we develop algorithms for monotone submodular maximization constrained by graphic, transversal matroids, or laminar matroids in time near-linear in the size of their representation. Our algorithms achieve an optimal approximation of $1-1/e-\epsilon$ and both generalize and accelerate the results of Ene and Nguyen [ICALP '19]. In fact, the running time of our algorithm cannot be improved within the fast continuous greedy framework of Badanidiyuru and Vondr\'ak [SODA '14]. To achieve near-linear running time, we make use of dynamic data structures that maintain bases with approximate maximum cardinality and weight under certain element updates. These data structures need to support a weight decrease operation and a novel FREEZE operation that allows the algorithm to freeze elements (i.e. force to be contained) in its basis regardless of future data structure operations. For the laminar matroid, we present a new dynamic data structure using the top tree interface of Alstrup, Holm, de Lichtenberg, and Thorup [TALG '05] that maintains the maximum weight basis under insertions and deletions of elements in $O(\log n)$ time. For the transversal matroid the FREEZE operation corresponds to requiring the data structure to keep a certain set $S$ of vertices matched, a property that we call $S$-stability., Comment: 38 pages. Abstract shortened for arxiv. To appear in ICALP 2023

How can we learn effective node representations on textual graphs? Graph Neural Networks (GNNs) that use Language Models (LMs) to encode textual information of graphs achieve state-of-the-art performance in many node classification tasks. Yet, combining GNNs with LMs has not been widely explored for practical deployments due to its scalability issues. In this work, we tackle this challenge by developing a Graph-Aware Distillation framework (GRAD) to encode graph structures into an LM for graph-free, fast inference. Different from conventional knowledge distillation, GRAD jointly optimizes a GNN teacher and a graph-free student over the graph's nodes via a shared LM. This encourages the graph-free student to exploit graph information encoded by the GNN teacher while at the same time, enables the GNN teacher to better leverage textual information from unlabeled nodes. As a result, the teacher and the student models learn from each other to improve their overall performance. Experiments in eight node classification benchmarks in both transductive and inductive settings showcase GRAD's superiority over existing distillation approaches for textual graphs.

We consider the massively parallel computation (MPC) model, which is a theoretical abstraction of large-scale parallel processing models such as MapReduce. In this model, assuming the widely believed 1-vs-2-cycles conjecture, solving many basic graph problems in $O(1)$ rounds with a strongly sublinear memory size per machine is impossible. We improve on the recent work of Holm and T\v{e}tek [SODA 2023] that bypass this barrier for problems when a planar embedding of the graph is given. In the previous work, on graphs of size $n$ with $O(n/\mathcal{S})$ machines, the memory size per machine needs to be at least $\mathcal{S} = n^{2/3+\Omega(1)}$, whereas we extend their work to the fully scalable regime, where the memory size per machine can be $\mathcal{S} = n^{\delta}$ for any constant $0< \delta < 1$. We give the first constant round fully scalable algorithms for embedded planar graphs for the problems of (i) connectivity and (ii) minimum spanning tree (MST). Moreover, we show that the $\varepsilon$-emulator of Chang, Krauthgamer, and Tan [STOC 2022] can be incorporated into our recursive framework to obtain constant-round $(1+\varepsilon)$-approximation algorithms for the problems of computing (iii) single source shortest path (SSSP), (iv) global min-cut, and (v) $st$-max flow. All previous results on cuts and flows required linear memory in the MPC model. Furthermore, our results give new algorithms for problems that implicitly involve embedded planar graphs. We give as corollaries constant round fully scalable algorithms for (vi) 2D Euclidean MST using $O(n)$ total memory and (vii) $(1+\varepsilon)$-approximate weighted edit distance using $\widetilde{O}(n^{2-\delta})$ memory. Our main technique is a recursive framework combined with novel graph drawing algorithms to compute smaller embedded planar graphs in constant rounds in the fully scalable setting., Comment: To appear in SODA24. 55 pages, 9 figures, 1 table. Added section on weighted edit distance and shortened abstract

CG:SHOP is an annual geometric optimization challenge and the 2022 edition proposed the problem of coloring a certain geometric graph defined by line segments. Surprisingly, the top three teams used the same technique, called conflict optimization. This technique has been introduced in the 2021 edition of the challenge, to solve a coordinated motion planning problem. In this paper, we present the technique in the more general framework of binary constraint satisfaction problems (binary CSP). Then, the top three teams describe their different implementations of the same underlying strategy. We evaluate the performance of those implementations to vertex color not only geometric graphs, but also other types of graphs., Comment: To appear at ACM Journal of Experimental Algorithmics

Transparency and accountability have become major concerns for black-box machine learning (ML) models. Proper explanations for the model behavior increase model transparency and help researchers develop more accountable models. Graph neural networks (GNN) have recently shown superior performance in many graph ML problems than traditional methods, and explaining them has attracted increased interest. However, GNN explanation for link prediction (LP) is lacking in the literature. LP is an essential GNN task and corresponds to web applications like recommendation and sponsored search on web. Given existing GNN explanation methods only address node/graph-level tasks, we propose Path-based GNN Explanation for heterogeneous Link prediction (PaGE-Link) that generates explanations with connection interpretability, enjoys model scalability, and handles graph heterogeneity. Qualitatively, PaGE-Link can generate explanations as paths connecting a node pair, which naturally captures connections between the two nodes and easily transfer to human-interpretable explanations. Quantitatively, explanations generated by PaGE-Link improve AUC for recommendation on citation and user-item graphs by 9 - 35% and are chosen as better by 78.79% of responses in human evaluation.

Graph Neural Networks (GNNs) are currently dominating in modeling graph-structure data, while their high reliance on graph structure for inference significantly impedes them from widespread applications. By contrast, Graph-regularized MLPs (GR-MLPs) implicitly inject the graph structure information into model weights, while their performance can hardly match that of GNNs in most tasks. This motivates us to study the causes of the limited performance of GR-MLPs. In this paper, we first demonstrate that node embeddings learned from conventional GR-MLPs suffer from dimensional collapse, a phenomenon in which the largest a few eigenvalues dominate the embedding space, through empirical observations and theoretical analysis. As a result, the expressive power of the learned node representations is constrained. We further propose OrthoReg, a novel GR-MLP model to mitigate the dimensional collapse issue. Through a soft regularization loss on the correlation matrix of node embeddings, OrthoReg explicitly encourages orthogonal node representations and thus can naturally avoid dimensionally collapsed representations. Experiments on traditional transductive semi-supervised classification tasks and inductive node classification for cold-start scenarios demonstrate its effectiveness and superiority.

$\newcommand{\eps}{\varepsilon}$We present an auction algorithm using multiplicative instead of constant weight updates to compute a $(1-\eps)$-approximate maximum weight matching (MWM) in a bipartite graph with $n$ vertices and $m$ edges in time $O(m\eps^{-1})$, beating the running time of the fastest known approximation algorithm of Duan and Pettie [JACM '14] that runs in $O(m\eps^{-1}\log \eps^{-1})$. Our algorithm is very simple and it can be extended to give a dynamic data structure that maintains a $(1-\eps)$-approximate maximum weight matching under (1) one-sided vertex deletions (with incident edges) and (2) one-sided vertex insertions (with incident edges sorted by weight) to the other side. The total time used is $O(m\eps^{-1})$, where $m$ is the sum of the number of initially existing and inserted edges., Comment: Appeared in IPCO 2023. The newest version of the paper improves the runtime by a log(1/eps) factor. The first version claimed result that the dynamic data structure supported arbitrary edge deletion has been corrected to one-sided vertex deletion and other side vertex insertion

Relational graph neural networks (RGNNs) are graph neural networks with dedicated structures for modeling the different types of nodes and edges in heterogeneous graphs. While RGNNs have been increasingly adopted in many real-world applications due to their versatility and accuracy, they pose performance and system design challenges: inherent memory-intensive computation patterns, the gap between the programming interface and kernel APIs, and heavy programming effort in optimizing kernels caused by their coupling with data layout and heterogeneity. To systematically address these challenges, we propose Hector, a novel two-level intermediate representation and its code generator framework, that (a) captures the key properties of RGNN models, and opportunities to reduce memory accesses in inter-operator scheduling and materialization, (b) generates code with flexible data access scheme to eliminate redundant data copies, (c) decouples model semantics, data layout, and operators-specific optimization from each other to reduce programming effort. By building on one general matrix multiply (GEMM) template and a node/edge traversal template, Hector achieves up to 9.9x speed-up in inference and 43.7x speed-up in training compared with the state-of-the-art public systems on select models, i.e., RGCN, RGAT and HGT, when running heterogeneous graphs provided by Deep Graph Library (DGL) and Open Graph Benchmark (OGB). In addition, Hector does not trigger any out-of-memory (OOM) exception in these tests. We also propose the linear operator reorder and compact materialization to further accelerate the system by up to 3.8x. As an indicator of programming effort reduction, Hector takes in 51 lines of code expressing the three models and generates a total of 8K lines of CUDA and C++ code., Comment: Accepted by ASPLOS

In this work we establish rigorously a measurement uncertainty relation (MUR) for three unbiased qubit observables, which was previously shown to hold true under some presumptions. The triplet MUR states that the uncertainty, which is quantified by the total statistic distance between the target observables and the jointly implemented observables, is lower bounded by an incompatibility measure that reflects the joint measurement conditions. We derive a necessary and sufficient condition for the triplet MUR to be saturated and the corresponding optimal measurement. To facilitate experimental tests of MURs we propose a straightforward implementation of the optimal joint measurements. The exact values of incompatibility measure are analytically calculated for some symmetric triplets when the corresponding triplet MURs are not saturated. We anticipate that our work may enrich the understanding of quantum incompatibility in terms of MURs and inspire further applications in quantum information science. This work presents a complete theory relevant to a parallel work [Y.-L. Mao, et al., Testing Heisenberg's measurement uncertainty relation of three observables, arXiv:2211.09389] on experimental tests., Comment: arXiv admin note: substantial text overlap with arXiv:2211.09389

Heisenberg's measurement uncertainty relations (MUR) of two quantum observables are essential for contemporary researches in quantum foundations and quantum information science. Going beyond, here we report the first experimental test of MURs for three quantum observables. Following the proposal of Bush, Lahti, and Werner [Phys. Rev. A 89, 012129 (2014)], we first establish rigorously MURs for triplets of unbiased qubit observables as combined approximation errors lower-bounded by an incompatibility measure. We then develop a convex programming protocol to numerically find the exact value of the incompatibility measure and the corresponding optimal measurements. Furthermore, we propose a novel implementation of optimal joint measurements and experimentally test our MURs using a single-photon qubit. Lastly, we discuss to analytically calculate the exact value of incompatibility measure for some symmetric triplets. We anticipate that this work may stimulate broad interests associated with the Heisenberg's uncertainty relation of multiple observables, enriching our understanding of quantum mechanics and inspiring innovative applications in quantum information science.

We consider using the preconditioned-Krylov subspace method to solve the system of linear equations with a three-by-three block structure. By making use of the three-by-three block structure, eight inexact block factorization preconditioners, which can be put into a same theoretical analysis frame, are proposed based on a kind of inexact factorization. By generalizing Bendixson Theorem and developing a unified technique of spectral equivalence, the bounds of the real and imaginary parts of eigenvalues of the preconditioned matrices are obtained. The comparison to eleven existed exact and inexact preconditioners shows that three of the proposed preconditioners can lead to high-speed and effective preconditioned-GMRES in most tests.

The unpredictable process of state collapse caused by quantum measurements makes the generation of quantum randomness possible. In this paper, we explore the quantitive connection between the randomness generation and the state collapse and provide a randomness verification protocol under the assumptions: (I) independence between the source and the measurement devices and (II) the L\"{u}ders' rule for collapsing state. Without involving heavy mathematical machinery, the amount of genereted quantum randomness can be directly estimated with the disturbance effect originating from the state collapse. In the protocol, we can employ general measurements that are not fully trusted. Equipped with trusted projection measurements, we can further optimize the randomness generation performance. Our protocol also shows a high efficiency and yields a higher randomness generation rate than the one based on uncertainty relation. We expect our results to provide new insights for understanding and generating quantum randomness

We revisit the classic problem of simplex range searching and related problems in computational geometry. We present a collection of new results which improve previous bounds by multiple logarithmic factors that were caused by the use of multi-level data structures. Highlights include the following: $\bullet$ For a set of $n$ points in a constant dimension $d$, we give data structures with $O(n^d)$ (or slightly better) space that can answer simplex range counting queries in optimal $O(\log n)$ time and simplex range reporting queries in optimal $O(\log n + k)$ time, where $k$ denotes the output size. For semigroup range searching, we obtain $O(\log n)$ query time with $O(n^d\mathop{\rm polylog}n)$ space. Previous data structures with similar space bounds by Matou\v{s}ek from nearly three decades ago had $O(\log^{d+1}n)$ or $O(\log^{d+1}n + k)$ query time. $\bullet$ For a set of $n$ simplices in a constant dimension $d$, we give data structures with $O(n)$ space that can answer stabbing counting queries (counting the number of simplices containing a query point) in $O(n^{1-1/d})$ time, and stabbing reporting queries in $O(n^{1-1/d}+k)$ time. Previous data structures had extra $\log^d n$ factors in space and query time. $\bullet$ For a set of $n$ (possibly intersecting) line segments in 2D, we give a data structure with $O(n)$ space that can answer ray shooting queries in $O(\sqrt{n})$ time. This improves Wang's recent data structure [SoCG'20] with $O(n\log n)$ space and $O(\sqrt{n}\log n)$ query time., Comment: Updated abstract metadata formatting. Accepted in SODA'23

Graph Neural Networks (GNNs) are powerful deep learning methods for Non-Euclidean data. Popular GNNs are message-passing algorithms (MPNNs) that aggregate and combine signals in a local graph neighborhood. However, shallow MPNNs tend to miss long-range signals and perform poorly on some heterophilous graphs, while deep MPNNs can suffer from issues like over-smoothing or over-squashing. To mitigate such issues, existing works typically borrow normalization techniques from training neural networks on Euclidean data or modify the graph structures. Yet these approaches are not well-understood theoretically and could increase the overall computational complexity. In this work, we draw inspirations from spectral graph embedding and propose $\texttt{PowerEmbed}$ -- a simple layer-wise normalization technique to boost MPNNs. We show $\texttt{PowerEmbed}$ can provably express the top-$k$ leading eigenvectors of the graph operator, which prevents over-smoothing and is agnostic to the graph topology; meanwhile, it produces a list of representations ranging from local features to global signals, which avoids over-squashing. We apply $\texttt{PowerEmbed}$ in a wide range of simulated and real graphs and demonstrate its competitive performance, particularly for heterophilous graphs., Comment: Accepted for publication at the NeurIPS 2022 GLFrontiers Workshop

$\newcommand{\Arr}{\mathcal{A}} \newcommand{\numS}{k} \newcommand{\ArrX}[1]{\Arr(#1)} \newcommand{\eps}{\varepsilon} \newcommand{\opt}{\mathsf{o}}$ For point sets $P_1, \ldots, P_\numS$, a set of lines $L$ is halving if any face of the arrangement $\ArrX{L}$ contains at most $|P_i|/2$ points of $P_i$, for all $i$. We study the problem of computing a halving set of lines of minimal size. Surprisingly, we show a polynomial time algorithm that outputs a halving set of size $O(\opt^{3/2})$, where $\opt$ is the size of the optimal solution. Our solution relies on solving a new variant of the weak $\eps$-net problem for corridors, which we believe to be of independent interest. We also study other variants of this problem, including an alternative setting, where one needs to introduce a set of guards (i.e., points), such that no convex set avoiding the guards contains more than half the points of each point set.

This paper describes a new method for representing embedding tables of graph neural networks (GNNs) more compactly via tensor-train (TT) decomposition. We consider the scenario where (a) the graph data that lack node features, thereby requiring the learning of embeddings during training; and (b) we wish to exploit GPU platforms, where smaller tables are needed to reduce host-to-GPU communication even for large-memory GPUs. The use of TT enables a compact parameterization of the embedding, rendering it small enough to fit entirely on modern GPUs even for massive graphs. When combined with judicious schemes for initialization and hierarchical graph partitioning, this approach can reduce the size of node embedding vectors by 1,659 times to 81,362 times on large publicly available benchmark datasets, achieving comparable or better accuracy and significant speedups on multi-GPU systems. In some cases, our model without explicit node features on input can even match the accuracy of models that use node features., Comment: To appear in the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (SIGKDD 22)

Can we combine heterogenous graph structure with text to learn high-quality semantic and behavioural representations? Graph neural networks (GNN)s encode numerical node attributes and graph structure to achieve impressive performance in a variety of supervised learning tasks. Current GNN approaches are challenged by textual features, which typically need to be encoded to a numerical vector before provided to the GNN that may incur some information loss. In this paper, we put forth an efficient and effective framework termed language model GNN (LM-GNN) to jointly train large-scale language models and graph neural networks. The effectiveness in our framework is achieved by applying stage-wise fine-tuning of the BERT model first with heterogenous graph information and then with a GNN model. Several system and design optimizations are proposed to enable scalable and efficient training. LM-GNN accommodates node and edge classification as well as link prediction tasks. We evaluate the LM-GNN framework in different datasets performance and showcase the effectiveness of the proposed approach. LM-GNN provides competitive results in an Amazon query-purchase-product application.

In the field of optics, bound state in the continuum (BIC) has been researched in many photonic crystals and periodic structures due to a strong resonance and an ultrahigh Q factor. Some designs of narrowband transmission filters, lasers, and sensors were proposed based on excellent optical properties of BIC. In this paper, we consider symmetrical rectangular grating structure firstly, then cut off the corner of one of the gratings, the Fano peak of quasi-BIC can be observed in the spectrum. After that, we further change the tilt parameter of the other grating, which minimizes the Fano line width. In the momentum space, the process of structural change corresponds to topological charges split from q=1 into two half charges q=1/2.We analyze guided mode resonance (GMR) excitation of the grating structure, and discuss the dispersion relations in the waveguide layer with the position of BIC in energy bands. In addition, the reflectance spectrum is found to exhibit asymmetric line-shapes with different values of the asymmetry parameters, M1 and M2. BIC is transformed into quasi-BIC as the symmetry of the structure is broken. This work demonstrates a double trapezoid structure with strong resonance properties, which has significant implications for exploring the phenomenon of BIC., Comment: 7 pages, 6 figures

Many real world graphs contain time domain information. Temporal Graph Neural Networks capture temporal information as well as structural and contextual information in the generated dynamic node embeddings. Researchers have shown that these embeddings achieve state-of-the-art performance in many different tasks. In this work, we propose TGL, a unified framework for large-scale offline Temporal Graph Neural Network training where users can compose various Temporal Graph Neural Networks with simple configuration files. TGL comprises five main components, a temporal sampler, a mailbox, a node memory module, a memory updater, and a message passing engine. We design a Temporal-CSR data structure and a parallel sampler to efficiently sample temporal neighbors to formtraining mini-batches. We propose a novel random chunk scheduling technique that mitigates the problem of obsolete node memory when training with a large batch size. To address the limitations of current TGNNs only being evaluated on small-scale datasets, we introduce two large-scale real-world datasets with 0.2 and 1.3 billion temporal edges. We evaluate the performance of TGL on four small-scale datasets with a single GPU and the two large datasets with multiple GPUs for both link prediction and node classification tasks. We compare TGL with the open-sourced code of five methods and show that TGL achieves similar or better accuracy with an average of 13x speedup. Our temporal parallel sampler achieves an average of 173x speedup on a multi-core CPU compared with the baselines. On a 4-GPU machine, TGL can train one epoch of more than one billion temporal edges within 1-10 hours. To the best of our knowledge, this is the first work that proposes a general framework for large-scale Temporal Graph Neural Networks training on multiple GPUs., Comment: VLDB'22

The all-photonic quantum repeater (APQR) is a promising repeater scheme to realize long-distance quantum communication. For a practical APQR, an indispensable requirement is the robustness of the repeater graph state (RGS) against photon loss. We propose a new loss-tolerant scheme by applying the generalized Shor code to RGS, which can be experimentally demonstrated with current technology. Experimentally, we first prepare and verify the nine-qubit Shor code. Then, by applying the generalized Shor code to APQR and preparing a simplified encoded RGS with the structure of $1\times2$ based on the Shor code state, the effectiveness of our loss-tolerant scheme and the loss tolerance of the encoded RGS are respectively verified. Our results make an essential step toward a practical APQR and enrich the research of quantum error correction code., Comment: 8 pages, 5 figures

It has recently been discovered that the nonlocality of an entangled qubit pair can be recycled for several Bell experiments. Here, we go beyond standard Bell scenarios and investigate the recycling of nonlocal resources in a quantum network. We realise a photonic quantum 3-branch star network in which three sources of entangled pairs independently connect three outer parties with a central node. After measuring, each outer party respectively relays their system to an independent secondary measuring party. We experimentally demonstrate that the outer parties can perform unsharp measurements that are strong enough to violate a network Bell inequality with the central party, but weak enough to maintain sufficient entanglement in the network to allow the three secondary parties to do the same. Moreover, the violations are strong enough to exclude any model based on standard projective measurements on the EPR pairs emitted in the network. Our experiment brings together the research program of recycling quantum resources with that of Bell nonlocality in networks.