Multi-task Gaussian Process Research Articles

A multi-task Gaussian process self-attention neural network for real-time prediction of the need for mechanical ventilators in COVID-19 patients

ObjectiveThe Coronavirus Disease 2019 (COVID-19) pandemic has overwhelmed the capacity of healthcare resources and posed a challenge for worldwide hospitals. The ability to distinguish potentially deteriorating patients from the rest helps facilitate reasonable allocation of medical resources, such as ventilators, hospital beds, and human resources. The real-time accurate prediction of a patient's risk scores could also help physicians to provide earlier respiratory support for the patient and reduce the risk of mortality. MethodsWe propose a robust real-time prediction model for the in-hospital COVID-19 patients' probability of requiring mechanical ventilation (MV). The end-to-end neural network model incorporates the Multi-task Gaussian Process to handle the irregular sampling rate in observational data together with a self-attention neural network for the prediction task. ResultsWe evaluate our model on a large database with 9,532 nationwide in-hospital patients with COVID-19. The model demonstrates significant robustness and consistency improvements compared to conventional machine learning models. The proposed prediction model also shows performance improvements in terms of area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) compared to various deep learning models, especially at early times after a patient's hospital admission. ConclusionThe availability of large and real-time clinical data calls for new methods to make the best use of them for real-time patient risk prediction. It is not ideal for simplifying the data for traditional methods or for making unrealistic assumptions that deviate from observation's true dynamics. We demonstrate a pilot effort to harmonize cross-sectional and longitudinal information for mechanical ventilation needing prediction.

Scalable multi-task Gaussian processes with neural embedding of coregionalization

Multi-task regression attempts to exploit the task similarity in order to achieve knowledge transfer across related tasks for improving the quality of prediction and alleviating the demand of big data. The application of Gaussian process (GP) in this scenario yields the non-parametric yet informative Bayesian multi-task regression paradigm. Multi-task GP (MTGP) provides not only the prediction mean but also the associated prediction variance to quantify uncertainty, thus gaining popularity in various scenarios. The linear model of coregionalization (LMC) is a well-known MTGP paradigm which exploits the dependency of tasks through linear combination of several independent and diverse GPs. The LMC however suffers from high model complexity and limited model capability when handling complicated multi-task cases. To this end, we develop the neural embedding of coregionalization that transforms the latent GPs into a high-dimensional latent space to induce rich yet diverse behaviors. Furthermore, we use advanced variational inference as well as sparse approximation to devise a tight and compact evidence lower bound (ELBO) for higher quality of scalable model inference. Extensive numerical experiments have been conducted to verify the higher prediction quality and better generalization of our model, named NSVLMC, on various real-world multi-task datasets and the cross-fluid modeling of unsteady fluidized bed.

Machine prescription for chronic migraine.

Responsive to treatment individually, chronic migraine remains strikingly resistant collectively, incurring the second-highest population burden of disability worldwide. A heterogeneity of responsiveness, requiring prolonged—currently heuristic—individual evaluation of available treatments, may reflect a diversity of causal mechanisms, or the failure to identify the most important, single causal factor. Distinguishing between these possibilities, now possible through the application of complex modelling to large-scale data, is critical to determine the optimal approach to identify new interventions in migraine and making the best use of existing ones. Examining a richly phenotyped cohort of 1446 consecutive unselected patients with chronic migraine, here we use causal multitask Gaussian process models to estimate individual treatment effects across 10 classes of preventatives. Such modelling enables us to quantify the accessibility of heterogeneous responsiveness to high-dimensional modelling, to infer the likely scale of the underlying causal diversity. We calculate the treatment effects in the overall population, and the conditional treatment effects among those modelled to respond and compare the true response rates between these two groups. Identifying a difference in response rates between the groups supports a diversity of causal mechanisms. Moreover, we propose a data-driven machine prescription policy, estimating the time-to-response when sequentially trialling preventatives by individualized treatment effects and comparing it to expert guideline sequences. All model performances are quantified out-of-sample. We identify significantly higher true response rates among individuals modelled to respond, compared with the overall population (mean difference of 0.034; 95% confidence interval 0.003–0.065; P = 0.033), supporting significant heterogeneity of responsiveness and diverse causal mechanisms. The machine prescription policy yields an estimated 35% reduction in time-to-response (3.750 months; 95% confidence interval 3.507–3.993; P < 0.0001) compared with expert guidelines, with no substantive increase in expense per patient. We conclude that the highly distributed mode of causation in chronic migraine necessitates high-dimensional modelling for optimal management. Machine prescription should be considered an essential clinical decision-support tool in the future management of chronic migraine.

Energy-Efficient Ground Traversability Mapping Based on UAV-UGV Collaborative System

With the development of science and technology, robots have been widely used in smart cities. The traversability mapping of environment perception is the prerequisite for robots to perform tasks. To save the energy consumption of traversability mapping for unmanned ground vehicle (UGV), we fusion a wide range of aerial images and a small amount of ground images to provide vision for UGV. Current map fusion methods are usually constrained by homogeneous model of robotic systems and lack of diverse sensors. As a result, they cannot work well in heterogeneous collaborative robotic systems that consist of aerial and ground robots. In this paper, we use heterogeneous robot systems, including UGV and unmanned aerial vehicles (UAV) to build an occupancy grid map that can be used for navigation. To fuse sensor data of different types, we propose a Collaborative Map Fusion algorithm based on Multi-task Gaussian Process Classification (MTGPC) using heterogeneous robotic systems. Besides, probabilistic model is exploited in traversability mapping, so the active perception can be used to build the map efficiently. Our system is tested in real scenes and can achieve an accuracy of more than 70&#x0025;. The map fusion using active perception is better than map fusion using random strategy in terms of speed and accuracy. To our knowledge, this is the first work that can build the occupancy grid map using sparse data points sampled from aerial images and ground lidar map.

Multi-task Gaussian process upper confidence bound for hyperparameter tuning and its application for simulation studies of additive manufacturing

In many scientific and engineering applications, Bayesian Optimization (BO) is a powerful tool for hyperparameter tuning of a machine learning model, materials design and discovery, etc. Multi-task BO is a general method to efficiently optimize multiple different, but correlated, “black-box” functions. The objective of this work is to develop an algorithm for multi-task BO with automatic task selection so that only one task evaluation is needed per query round. Specifically, a new algorithm, namely, Multi-Task Gaussian Process Upper Confidence Bound (MT-GPUCB), is proposed to achieve this objective. The MT-GPUCB is a two-step algorithm, where the first step chooses which query point to evaluate, and the second step automatically selects the most informative task to evaluate. Under the bandit setting, a theoretical analysis is provided to show that our proposed MT-GPUCB is no-regret under some mild conditions. Our proposed algorithm is verified experimentally on a range of synthetic functions. In addition, our algorithm is applied to Additive Manufacturing simulation software, namely, Flow-3D Weld, to determine material property values, ensuring the quality of simulation output. The results clearly show the advantages of our query strategy for both design point and task.

Recursive Gaussian Process over graphs for Integrating Multi-timescale Measurements in Low-Observable Distribution Systems

The transition to a smarter grid is empowered by enhanced sensor deployments and smart metering infrastructure in the distribution system. Measurements from these sensors and meters can be used for many applications, including distribution system state estimation (DSSE). However, these measurements are typically sampled at different rates and could be intermittent due to losses during the aggregation process. These multi time-scale measurements should be reconciled in real-time to perform accurate grid monitoring. This paper tackles this problem by formulating a recursive multi-task Gaussian process (RGP-G) approach that sequentially aggregates sensor measurements. Specifically, we formulate a recursive multi-task GP with and without network connectivity information to reconcile the multi time-scale measurements in distribution systems. The proposed framework is capable of aggregating the multi-time scale measurements batch-wise or in real-time. Following the aggregation of the multi time-scale measurements, the spatial states of the consistent time-series are estimated using matrix completion based DSSE approach. Simulation results on IEEE 37 and IEEE 123 bus test systems illustrate the efficiency of the proposed methods from the standpoint of both multi time-scale data aggregation and DSSE.

Probabilistic Wind Power Forecasting for Newly-Built Wind Farm Based on Multi-Task Gaussian Process Method

Probabilistic Wind Power Forecasting for Newly-Built Wind Farm Based on Multi-Task Gaussian Process Method

Regulation-based probabilistic substance quality index and automated geo-spatial modeling for water quality assessment

Natural environments are recognized as complex heterogeneous structures thus requiring numerous multi-scale observations to yield a comprehensive description. To monitor the current state and identify negative impacts of human activity, fast and precise instruments are in urgent need. This work provides an automated approach to the assessment of spatial variability of water quality using guideline values on the example of 1526 water samples comprising 21 parameters at 448 unique locations across the New Moscow region (Russia). We apply multi-task Gaussian process regression (GPR) to model the measured water properties across the territory, considering not only the spatial but inter-parameter correlations. GPR is enhanced with a Spectral Mixture Kernel to facilitate a hyper-parameter selection and optimization. We use a 5-fold cross-validation scheme along with R^2-score to validate the results and select the best model for simultaneous prediction of water properties across the area. Finally, we develop a novel Probabilistic Substance Quality Index (PSQI) that combines probabilistic model predictions with the regulatory standards on the example of the epidemiological rules and hygienic regulations established in Russia. Moreover, we provide an interactive map of experimental results at 100 m2 resolution. The proposed approach contributes significantly to the development of flexible tools in environment quality monitoring, being scalable to different standard systems, number of observation points, and region of interest. It has a strong potential for adaption to environmental and policy changes and non-unified assessment conditions, and may be integrated into support-decision systems for the rapid estimation of water quality spatial distribution.

A missing sensor measurement data reconstruction framework powered by multi-task Gaussian process regression for dam structural health monitoring systems

The sensor-based structural health monitoring (SHM) systems are widely embedded in the new-constructed and rehabilitated dam. Due to the harsh working environment, poor installation, and sampling error, sensor fault often inevitably occurs. In this paper, rather than using conventional Gaussian process regression(GPR) to reconstruct missing data from multiple sensors independently, we propose a multi-task GPR (mGPR) paradigm for capturing the correlation among various sensors to reconstruct missing data from faulty sensors as a whole. In this framework, for a particular sensor, the missing data is reconstructed by the approach which not only learns other known data from this sensor but also learns the whole known measurements from other sensors. The proposed paradigm is quite beneficial for dam SHM systems since the missing data from the faulty sensor(s) can be efficiently and accurately learned by the whole historical data including both faulty and normal sensors. The usefulness of the proposed paradigm is demonstrated through three measurement items including air temperature, dam displacements, and crack opening displacements collected from two dams in long-term service. We investigate two missing data scenarios with distinct positions in sensors. The experimental results show our proposed mGPR has significantly better performance than conventional multiple GPR for all the tested measurement items, especially in the scenarios that the missing part occurs at the beginning or the end of the dataset. It is also shown the multi-task learning paradigm powered by mGPR is considerable to address missing data reconstruction for dam SHM systems.

MGP-AttTCN: An interpretable machine learning model for the prediction of sepsis.

With a mortality rate of 5.4 million lives worldwide every year and a healthcare cost of more than 16 billion dollars in the USA alone, sepsis is one of the leading causes of hospital mortality and an increasing concern in the ageing western world. Recently, medical and technological advances have helped re-define the illness criteria of this disease, which is otherwise poorly understood by the medical society. Together with the rise of widely accessible Electronic Health Records, the advances in data mining and complex nonlinear algorithms are a promising avenue for the early detection of sepsis. This work contributes to the research effort in the field of automated sepsis detection with an open-access labelling of the medical MIMIC-III data set. Moreover, we propose MGP-AttTCN: a joint multitask Gaussian Process and attention-based deep learning model to early predict the occurrence of sepsis in an interpretable manner. We show that our model outperforms the current state-of-the-art and present evidence that different labelling heuristics lead to discrepancies in task difficulty. For instance, when predicting sepsis five hours prior to onset on our new realistic labels, our proposed model achieves an area under the ROC curve of 0.660 and an area under the PR curve of 0.483, whereas the (less interpretable) previous state-of-the-art model (MGP-TCN) achieves 0.635 AUROC and 0.460 AUPR and the popular commercial InSight model achieves 0.490 AUROC and 0.359 AUPR.

Emotion Analysis on Text Using Multiple Kernel Gaussian...

The ability to discern human emotions is critical for making chatbox behave like humans. Gaussian Process (GP) is a non-parametric Bayesian modeling and can be used to predict the presence of either a single emotion (single-task GP) or multiple emotions (multi-task GP) in natural language text. Employing multiple kernels in GP can enhance the performance of the emotion analysis tasks. The particular choice of kernel functions determines the properties such as smoothness, length scales, sharpness, and amplitude, drawn from the GP prior. Using a specific kernel may be a source of bias and can be avoided by using different kernels together. The default kernel used with GP is a Radial Basis Function (RBF). It is infinitely differentiable; GP with this function has mean square derivatives of all orders and is thus very smooth. The sharpness which occurs in the midst of the smoothness can be detected using the exponential kernel. The multi-layer perceptron kernel has greater generalization for each training example and is good for extrapolation. Our experiments show that, for learning the presence of a single emotion in a natural language sentence (single-task), multiple kernel GP with the sum of RBF and multi-layer perceptron kernels performs better than single kernel GP. Likewise, for learning the presence of several different emotions in a sentence (multi-task), multiple kernel GP with the sum of RBF, exponential and multi-layer perceptron kernels performs better than single kernel GP. Multiple Kernel Gaussian Process also outperforms Convolutional Neural Network (CNN).

Characterizing the spatial and temporal variation of the land surface temperature hotspots in Wuhan from a local scale

ABSTRACT Land Surface Temperature (LST) derived from space-borne Thermal-infrared (TIR) sensors is a key parameter of urban climate studies. Current studies are inefficient to capture the spatial and temporal variations of LST for only one snapshot adopted at one time. Focusing on the characterization of the spatial and temporal of LST variations at local scales, the latent patterns, and morphological characteristics are extracted in this study. Technically, sixteen MODerate-resolution Imaging Spectroradiometer (MODIS) eight-day synthesized LST products (MYD11A2) in 2002, 2007, 2012, and 2017 are employed. First, the non-parametric Multi-Task Gaussian Process Model (MTGP) is used to extract the smooth and continuous Latent LST (LLST) patterns using one LST subset and its temporally adjacent images. Second, the Multi-Scale Shape Index (MSSI) is then applied to quantify the morphological characteristics at the optimal scale. Then, the LLST patterns and MSSI maps are clustered into multiple spatial categories. The specific clusters with the highest LLST and MSSI values are considered as local LLST hotspots. The Hotspots Weighted Mean Center (HSWMC) and standard deviation ellipse are adopted to further investigate the spatiotemporal change of hotspots orientation, direction, and trajectories. Results revealed that Impervious Surfaces (IS) composition is the most significant external forcing of local LST anomalies. The configuration factors (e.g., shape index, aggregation index) also have a noticeable local warming effect. This study represents a latent pattern and morphology-based framework for LST hotspots spatial and temporal variations characterization, catering to the zoning and grading strategies in urban planning.

Enhanced Gaussian process regression-based forecasting model for COVID-19 outbreak and significance of IoT for its detection.

Virus based epidemic is one of the speedy and widely spread infectious disease which can affect the economy of the country as well as it is life-threatening too. So, there is a need to forecast the epidemic lifespan, which can help us in taking preventive measures and remedial action on time. These preventive measures and corrective action may consist of closing schools, closing malls, closing theaters, sealing of borders, suspension of public services, and suspension of traveling. Resuming such restrictions is depends upon the outbreak momentum and its decay rate. The accurate forecasting of the epidemic lifespan is one of the enormously essential and challenging tasks. It is a challenging task because the lack of knowledge about the novel virus-based diseases and its consequences with complicated societal-governmental factors can influence the widespread of this newly born disease. At this stage, any forecasting can play a vital role, and it will be reliable too. As we know, the novel virus-based diseases are in a growing phase, and we also do not have real-time data samples. Thus, the biggest challenge is to find out the machine learning-based best forecasting model, which could offer better forecasting with the limited training samples. In this paper, the Multi-Task Gaussian Process (MTGP) regression model with enhanced predictions of novel coronavirus (COVID-19) outbreak is proposed. The purpose of the proposed MTGP regression model is to predict the COVID-19 outbreak worldwide. It will help the countries in planning their preventive measures to reduce the overall impact of the speedy and widely spread infectious disease. The result of the proposed model has been compared with the other prediction model to find out its suitability and correctness. In subsequent analysis, the significance of IoT based devices in COVID-19 detection and prevention has been discussed.

Modeling and Analysis of Permanent Magnet Spherical Motors by a Multitask Gaussian Process Method and Finite Element Method for Output Torque

Permanent magnet spherical motors (PMSMs) operate on the principle of the dc excitation of stator coils and three freedom of motion in the rotor. Each coil generates the torque in a specific direction, collectively they move the rotor to a direction of motion. Modeling and analysis of the output torque are of critical importance for precise position control applications. The control of these motors requires precise output torques by all coils at a specific rotor position, which is difficult to achieve in the three-dimension space. This article is the first to apply the Gaussian process to establish the relationship of the rotor position and the output torque for PMSMs. Traditional methods are difficult to resolve such a complex three-dimensional problem with a reasonable computational accuracy and time. This article utilizes a data-driven method using only input and output data validated by experiments. The multitask Gaussian process is developed to calculate the total torque produced by multiple coils at the full operational range. The training data and test data are obtained by the finite-element method. The effectiveness of the proposed method is validated and compared with existing data-driven approaches. The results exhibit superior performance of accuracy.

Multi-parametric system for risk stratification in mitral regurgitation: A multi-task Gaussian prediction approach.

We hypothesized that a multi-parametric approach incorporating medical comorbidity information, electrocardiographic P-wave indices, echocardiographic assessment, neutrophil-to-lymphocyte ratio (NLR) and prognostic nutritional index (PNI) calculated from laboratory data can improve risk stratification in mitral regurgitation (MR). Patients diagnosed with mitral regurgitation between 1 March 2005 and 30 October 2018 from a single centre were retrospectively analysed. Outcomes analysed were incident atrial fibrillation (AF), transient ischemic attack (TIA)/stroke and mortality. This study cohort included 706 patients, of whom 171 had normal inter-atrial conduction, 257 had inter-atrial block (IAB) and 266 had AF at baseline. Logistic regression analysis showed that age, hypertension and mean P-wave duration (PWD) were significant predictors of new-onset AF. Low left ventricular ejection fraction (LVEF), abnormal P-wave terminal force in V1 (PTFV1) predicted TIA/stroke. Age, smoking, hypertension, diabetes mellitus, hypercholesterolaemia, ischemic heart disease, secondary mitral regurgitation, urea, creatinine, NLR, PNI, left atrial diameter (LAD), left ventricular end-diastolic dimension, LVEF, pulmonary arterial systolic pressure, IAB, baseline AF and heart failure predicted all-cause mortality. A multi-task Gaussian process learning model demonstrated significant improvement in risk stratification compared to logistic regression and a decision tree method. A multi-parametric approach incorporating multi-modality clinical data improves risk stratification in mitral regurgitation. Multi-task machine learning can significantly improve overall risk stratification performance.

A Respiratory Motion Prediction Method Based on Improved Relevance Vector Machine

Thoracic and abdominal tumor radiotherapy calls for prediction to compensate the impact of respiratory movement on real-time tracking of the target. Amidst this backdrop, this paper proposes a method to improve relevance vector machine, which is able to first forecast the three dimensions of respiratory movement respectively in virtue of offline training. Then the output results will be sent into multi-task Gaussian process model simultaneously to correct prediction error with the correlation between three-dimensional data and dynamically updating the training set, thus eventually realizing 3D real-time prediction of respiratory movement. The experimental results indicate that compared with the traditional relevance vector machine, the reduction range of the root-mean-square error predicted with this method at intervals of 154 ms and 308 ms is 8.8% ~ 15.7%. The prediction accuracy has been significantly improved.

A virtual metrology method with prediction uncertainty based on Gaussian process for chemical mechanical planarization

The prediction of the average material removal rate (MRR) in chemical mechanical planarization (CMP) process has been recognized to be a critical factor of virtual metrology (VM) modeling for advanced process control (APC). This paper proposes a Gaussian process regression (GPR)-based model to dynamically predict MRR in CMP process. The proposed method uses K-nearest neighbor (KNN) to search for reference MRR samples in the historical dataset. Furthermore, a GPR model is trained to fuse the information from reference samples. Finally, the proposed method employs multi-task Gaussian process (MTGP) to predict the final MRR and quantify the prediction uncertainty based on the historical and the reference MRR. Compared with other methods in the recent literature, the proposed method, named KNN-MTGP model, yields better prediction accuracy than ensemble models, and comparable accuracy with deep neural networks (NN). Besides, KNN-MTGP model is capable to demonstrate the behavior of the past MRR changing with time and provide quantified prediction uncertainties. In this paper, the feasibility and advantages of KNN-MTGP model are evaluated based on the dataset of 2016 prognostic and health management (PHM) data challenge.

Power Prediction for Heterogeneous Ground Robots Through Spatial Mapping and Sharing of Terrain Data

Ground robot power consumption often varies significantly throughout an off-road environment, depending on the terrain. Previous traversals over the environment can inform future predictions through spatial mapping of collected power data. However, such predictions are particular to a single robot. We develop a framework based on multi-task Gaussian process regression (MTGP) to share terrain information across multiple robots. The framework allows for the efficient selection of hyperparameters and the prediction of power consumption along paths through nearest neighbor data selection. The scaling of this approach is demonstrated in simulation. Furthermore, experimental testing with two robots shows a significant reduction in power prediction error when an environment, mapped by one robot, informs the power predictions of another robot.

Spatiotemporal Multitask Learning for 3-D Dynamic Field Modeling

3-D dynamic field modeling using data acquired from sensor networks is typically complex due to the data sparsity and missing problem. In this article, we consider the ubiquitous missing data problem in current sensor networks and aim to take complete advantage of the existing sensor data for thermal field modeling. In the common scenario, data from the target network are not always obtainable, but data from other neighboring networks with homogeneous fields are accessible. Thus, a novel method that captures the information acquired from these neighboring networks is proposed. To achieve accurate thermal field estimation using limited sensor observations, we develop a mixed-effect model framework in which the dynamic field is decomposed into a mean profile and local variability. In particular, we establish a spatiotemporal field multitask learning (FML) approach to identify the spatiotemporal correlation by integrating a multitask Gaussian process (MGP) framework into an autoregressive (AR) model using neighboring data sources from homogeneous fields. Our proposed method is verified through a real case study of thermal field estimation during grain storage. Note to Practitioners —The proposed method aims to obtain an accurate estimation of a thermal field when certain sensor data are inaccessible. To better implement this method in practice, three things are noteworthy: First, the mean profile of the thermal field should be extracted using the thermodynamic model, so that the remaining data are able to follow a Gaussian process. Second, the FML approach considers neighboring data sources from homogeneous thermal fields to achieve an accurate estimation of the target thermal field. Thus, the target thermal field and other thermal fields should be under similar external conditions, e.g., environmental surroundings, geographical location, and field size. Third, the proposed method can not only process the data from grid-based sensor networks, but also can be extended to other topological structures of sensor networks for field estimation.

How Do the Multi-Temporal Centroid Trajectories of Urban Heat Island Correspond to Impervious Surface Changes: A Case Study in Wuhan, China.

Conspicuous expansion and intensification of impervious surfaces accompanied by rapid urbanization are widely recognized to have exerted evident impacts on the urban thermal environment. Investigating the spatially and temporally varying relationships between Land Surface Temperature (LST) and impervious surfaces (IS) at multiple scales is of great significance for steering IS expansion and intensification. This study proposes an analytical framework to investigate the spatiotemporal variations of LST and its responses to IS in Wuhan, China at both city scale and sub-region scale. The summer LST patterns in 2002–2017 are extracted by Multi-Task Gaussian Process (MTGP) model from raw 8-day synthesized MODerate-resolution Imaging Spectroradiometer (MODIS) LST data. At the city scale, the weighted center of LST (LSTWC) and impervious surface fraction (ISFWC), multi-temporal trajectories and coupling indicators are utilized to comprehensively examine the spatial and temporal dynamics of LST and IS within Wuhan. At the sub-region scale, urban heat island ratio index (URI), impervious surfaces contribution index (ISCI) and sprawl rate are introduced for further quantifying the relationships of LST and IS. The results reveal that IS and hot thermal landscapes expanded by 407.43 km2 and 255.82 km2 in Wuhan in 2002–2017 at city scale. The trajectories of LSTWCs and ISFWCs are visually coherent and both heading to southeast direction in general. At the sub-region scale, the specific cardinal directions with the highest ISCI variations are examined to be the exact directions of ISFWC trajectories in 2002–2017. The results reveal that the spatiotemporal variations of LST and IS are highly correlated at both city and sub-region scales within Wuhan, thus testifying the significance of steering IS expansion and renewal for controlling urban thermal environment deterioration.