Deploying Large Language Models (LLMs) on resource-constrained devices remains challenging due to limited memory, lack of GPUs, and the complexity of existing runtimes. In this paper, we introduce TranSQL + , a template-based code generator that translates LLM computation graphs into pure SQL queries for execution in relational databases. Without relying on external libraries, TranSQL + , leverages mature database features-such as vectorized execution and out-of-core processing-for efficient inference. We further propose a row-to-column (ROW2COL) optimization that improves join efficiency in matrix operations. Evaluated on Llama3-8B and DeepSeekMoE models, TranSQL + achieves up to 20× lower prefill latency and 4× higher decoding speed compared to DeepSpeed Inference and Llama.cpp in low-memory and CPU-only configurations. Our results highlight relational databases as a practical environment for LLMs on low-resource hardware.

Wastewater infrastructures are vital in urban cities, but aging sanitary sewer systems face issues like cracked pipes and damaged manholes, leading to infiltration and inflow problems. Climate change contributes to more frequent heavy precipitation, potentially causing significant sewer overflows. These overflows can endanger public health by carrying disease-causing microorganisms, pathogens, chemicals, and pollutants. To improve the maintenance of wastewater systems, which often face challenges of sparse sensor coverage in the field, we develop a graph dataset and a spatio-temporal hydraulic model - HydroNet - tailored to urban wastewater systems. The trained HydroNet is able to predict temporal water depth, which can help identify potential infiltration, leading to an automated urban wastewater system with minimal time and labor costs.

Impervious surfaces increase surface runoff, leading to elevated flood risks and nonpoint source pollution. Predicting impervious surface ratios is essential for various urban management practices, ranging from drainage infrastructure design and water quality assessment to utility fee evaluation and flood risk mitigation. Traditionally, the information on impervious surface ratios is often estimated by city managers and engineers based on empirical values and assumptions. Recent studies have highlighted aerial image classification using machine learning and deep learning models, but such approaches are computationally intensive. We propose a graph neural network (GNN)-driven method, named GraphParcelNet, for advancing the quantification of parcel-level impervious surface ratios at the city scale. To our best knowledge, we are the first to transform land parcel datasets, consisting of vector data in geometric shapes (polygons), into a graph model that considers the spatial relationships between parcels. By utilizing a GNN-based approach, GraphParcelNet enhances the representation of spatial dependencies, resulting in more accurate and reliable predictions over traditional methods. Our experimental results demonstrate that GraphParcelNet outperforms previous methods, providing an accurate measurement for impervious surface ratios.

Accurate short-term precipitation prediction at a high spatial resolution is crucial for effective urban water management, flooding warning, and mitigation. However, conventional numerical weather models usually face the challenge of systematic errors and spatiotemporal biases due to an inadequate understanding of many processes and unrealistic parameterizations. In recent years, deep learning techniques have gained popularity as a tool in precipitation forecasting and risk pre-warning. To support deep learning for precipitation forecasting and flooding warning, this paper introduces a large-scale multimodal Geo dataset. This dataset incorporates spatially connected features and real-world climate data, enabling the prediction of extreme precipitations. The dataset comprises Multi-Radar/Multi-Sensor System (MRMS), High-Resolution Rapid Refresh (HRRR), Geostationary Satellite Server (GOES) data, and local hydrological data from the United States Geological Survey (USGS), providing a diverse array of information sources. The compiling of multi-source data within the proposed multimodal Geo scope can improve prediction accuracy over uni-modal data and shows high accuracy in predicting heavy rain when integrating Transformer, which offers the opportunity for more efficient urban water management and improved disaster response strategies. By providing a comprehensive view of environmental conditions, this dataset enables a deeper understanding of precipitation patterns, facilitating effective mitigation efforts.

Query optimizers in Relational Database Management Systems (RDBMS) rely on the approximations of data distributions of attributes, which are used for cardinality-based decisions like selectivity of a predicate or finding the right join order of a multiway join in a SQL query. The commercial databases try to reach best possible approximations of data distribution using compressed histograms. The state of the art in measurement of quality of a histogram is the q-error. However, one of the shortcomings of the measurement of quality of the histograms using q-error is that it pivots the entire quality on the worst case or the infinity norm of the multiplicative errors of the estimations. This may not be symbolic of the workload or the overall range of errors that would be representative of the system. In this paper, we propose a dynamic programming algorithm to achieve an optimal histogram structure by keeping the overall multiplicative errors low. We propose a one pass linear regression on the step function and apply dynamic programming methods to achieve an optimal histogram structure that optimizes the pair (slope β and intercept α). The algorithm achieves lower slope and intercept of the resultant regression on the sorted order of multiplicative errors, thus reducing the overall estimation errors for the system. We refer to this pair as the Q-Regression of a histogram. We then introduce the metric, QRegrArea as a new means to better quantify the cumulative distribution of errors for a given histogram. We provide the experimental validation of the proposed methods against the state-of the-art models in the literature and industry.

A skyline query searches the data points that are not dominated by others in the dataset. It is widely adopted for many applications which require multi-criteria decision making. However, skyline query processing is considerably time-consuming for a high-dimensional large scale dataset. Parallel computing techniques are therefore needed to address this challenge, among which$MapReduce$is one of the most popular frameworks to process big data. A great number of efficient$MapReduce$skyline algorithms have been proposed in the literature and most of their designs focus on partitioning and pruning the given dataset. However, there are still opportunities for further parallelism. In this study, we propose two parallel skyline processing algorithms using a novel$LShape$partitioning strategy and an effective$Propagation$$Filtering$method. These two algorithms are$2Phase$$LShape$and$1Phase$$LShape$, used for multiple reducers and single reducer, respectively. By extensive experiments, we verify that our algorithms outperformed the state-of-the-art approaches, especially for high-dimensional large scale datasets.

GeoSpatial Location is often treated as the golden thread of data analysis while heterogeneous data is connected through the locality. Heterogeneous data integration via location is challenging and very essential in data analysis. For example, a rental price of a residential property is determined by interior factors (e.g., area of the property and interior design) and also external factors related to locality. The key challenge is incorporating data from various sources connected by locality and modeling the joint decision-making process. We propose to use different embeddings to collectively learn the representations of different data jointly while preserving the rental price prediction. The benefit of our heterogeneous data integration design is to model the interactions among all the factors contributing to our predefined task and achieve different levels of abstraction using multi-level spatial representations.

Deep learning has assisted modern life in various ways. One example is that accurate economic prediction helps people better allocate and distribute their resources. In the U.S., home prices have been accelerating during the COVID-19 pandemic and climbed 13.3% in March 2021 from the previous year. Real estate market prediction is critical for home buyers and investors to make wise decisions. In some circumstances, accurate predictions on home prices are more important than usual in helping decision-makers to reduce financial mistakes.

In this study, we propose two parallel skyline processing algorithms using a novel LShape partitioning strategy and an effective Propagation Filtering method. By extensive experiments, we verify that our algorithms outperformed the state-of-the-art approaches, especially for high-dimensional large scale datasets.

With the rapid progress of global urbanization and function division among different geographical regions, it is of urgent need to develop methods that can find regions of desired future function distributions in applications. For example, a company tends to open a new branch in a region where the growth trend of industrial sectors fits its strategic goals, or is similar to that of an existing company location; while a job hunter tends to search regions where his/her expertise aligns with the industrial growth trend providing sufficient job opportunities to sustain future employment and job-hopping.

Relational database management systems (RDBMSs) are powerful because they are able to optimize and execute queries against relational databases. However, when it comes to NLIDB (natural language interface for databases), the entire system is often custom-made for a particular database. Overcoming the complexity and expressiveness of natural languages so that a single NLI can support a variety of databases is an unsolved problem. In this work, we show that it is possible to separate data specific components from latent semantic structures in expressing relational queries in a natural language. With the separation, transferring an NLI from one database to another becomes possible. We develop a neural network classifier to detect data specific components and an adversarial mechanism to locate them in a natural language question. We then introduce a general purpose transfer-learnable NLI that focuses on the latent semantic structure. We devise a deep sequence model that translates the latent semantic structure to an SQL query. Experiments show that our approach outperforms previous NLI methods on the WikiSQL [49] dataset, and the model we learned can be applied to other benchmark datasets without retraining.

A natural language interface (NLI) to databases is an interface that translates a natural language question to a structured query that is executable by database management systems (DBMS). However, an NLI that is trained in the general domain is hard to apply in the spatial domain due to the idiosyncrasy and expressiveness of the spatial questions. Inspired by the machine comprehension model, we propose a spatial comprehension model that is able to recognize the meaning of spatial entities based on the semantics of the context. The spatial semantics learned from the spatial comprehension model is then injected to the natural language question to ease the burden of capturing the spatial-specific semantics. With our spatial comprehension model and information injection, our NLI for the spatial domain, named SpatialNLI, is able to capture the semantic structure of the question and translate it to the corresponding syntax of an executable query accurately. We also experimentally ascertain that SpatialNLI outperforms state-of-the-art methods.

Due to the inadequacy of GPS signals in indoor spaces, Indoor Positioning Services (IPSs) have drawn great attention. The popular smartphone localization technique relies on a centralized server to achieve localization, allowing the server to acquire a user's location in fine granularity. To ensure the privacy of IPS users, we propose an Encrypted Indoor Positioning Service (EIPS) model that protects users' privacy from the centralized server and maintains localization accuracy simultaneously. Our EIPS model enables users to encrypt and decrypt their query through an Encryption and Decryption Server (EDS) bi-directionally in a commutative way, so the users' locations remain private to both EIPS and EDS. We also propose Query Split, Artificial Dimensions and Columns to prevent Known Plaintext Attack (KPA). Our analytical and experimental evaluations show that our model is resilient to a variety of privacy attacks without loss of efficiency and accuracy.

This research proposes an advanced approach that improves the efficiency of skyline query processing by significantly reducing the computational cost on object comparisons, i.e., dominance tests between objects. Our solutions are based on two novel concepts. The skyline query over Minimum Bounding Rectangles (MBRs) receives a set of MBRs and returns the MBRs that are not dominated by other MBRs. In the dominance test for MBRs, the detailed attribute values of objects in the MBRs are not accessed. Moreover, the dependent group of MBRs reduces the search space for dominance tests. Objects in an MBR are only compared with the ones in the corresponding dependent groups of the MBR rather than with the entire dataset. Our solutions apply the two concepts to the R-tree in order to use its hierarchical structure in which every node is a natural abstraction of an MBR. Specifically, given the R-tree index of an input dataset, we first eliminate unqualified objects by utilizing the skyline query over MBRs (i.e., intermediate nodes in the R-tree). Subsequently, we generate dependent groups for the skyline MBRs. Two dependent group generation methods that rely on either the sorting technique or the R-tree index are developed. Further, we apply an existing skyline algorithm to every dependent group, and the results of the original skyline query are the union of skyline objects in the dependent groups. In addition, we also analyze the cardinality of the two new concepts based on a probabilistic model, which enables us to analyze the computational complexity of the proposed solutions. Our experimental results show that the proposed solutions are clearly more efficient than the state-of-the-art approaches.

This paper presents an exact solution and a heuristic solution to a UAV-assisted parcel delivery problem, in which UAVs can only be operated in Visual-Line-Of-Sight (VLOS) areas. In our proposed problem, we assume that trucks travel on road networks, and UAVs move in Euclidean spaces and can launch at any locations on roads. We first demonstrate the overview of our exact solution that iterates all permutations of destinations for an optimal delivery route. Given a specific delivery order, an intuitive approach needs to check all possible locations on roads in the VLOS areas and find a globally optimal location for every destination if UAVs are used for delivery. To avoid high computational cost of searching the optimal location at runtime, we propose an advanced index-based alternative, which computes optimal delivery routes in a pre-processing stage. Due to the nature of NP-hard problems, we also propose a heuristic approach that utilizes delivery groups for the proposed problem of practical size. All proposed solutions are evaluated through extensive experiments.