Rethinking development and major research plans of Industrial Internet in China.

This paper gives a definition of the Industrial Internet and expounds on the academic connotation of the future Industrial Internet. From this foundation, we outline the development and current status of the Industrial Internet in China and globally. Moreover, we detail the avant-garde paradigms encompassed within the National Natural Science Foundation of China (NSFC)'s "Future Industrial Internet Fundamental Theory and Key Technologies" research plan and its corresponding management strategies. This research initiative endeavors to enhance interdisciplinary collaborations, aiming for a synergistic alignment of industry, academia, research, and practical implementations. The primary focus of the research plan is on the pivotal scientific challenges inherent to the future industrial internet. It is poised to traverse the "first mile", encompassing foundational research and pioneering innovations specific to the industrial internet, and seamlessly bridges to the "last mile", ensuring the effective commercialization of scientific and technological breakthroughs into tangible industrial market applications. The anticipated contributions from this initiative are projected to solidify both the theoretical and practical scaffolding essential for the cultivation of a globally competitive industrial internet infrastructure in China.

Comorbidity of Myasthenia gravis and Graves' disease as immune reconstitution-associated autoimmune disease in HIV infection: A case report and literature review.

BACKGROUND: Comorbidity of Myasthenia gravis (MG) and Graves' disease (GD) in treated HIV-infected individuals has rarely been described and little study has been done on the link between HIV-related immune reconstitution and autoimmune diseases occurring post antiretroviral therapy. CASE PRESENTATION: Here we report on a 33-year-old Chinese man with HIV infection who had been virologically suppressed since 2018. The patient was diagnosed with GD and was treated in 2020. Early in 2022, he developed fluctuating weakness and fatigue involving the bilateral extraocular muscles and limbs. With a positive neostigmine test, he was considered to have MG, but showed a poor response to oral medication. After multiple failed medication attempts, a thymectomy was finally performed to resolve his symptoms. The consecutive onset of immunological events may have partially resulted from immune reconstitution after viral control. CONCLUSIONS: This is a rare case of HIV-related immune reconstitution-associated autoimmune disease (IRAD) with comorbidity of MG and GD which was reported initially. Cooperation with multidisciplinary teams is essential to avoid misdiagnosis and to promote the overall health of HIV-infected patients.

Long-term efficacy and predictors of pembrolizumab-based regimens in patients with advanced esophageal cancer in the real world.

BACKGROUND: Pembrolizumab combined with chemotherapy has been proven effective as first-line therapy in patients with advanced esophageal cancer. Few trials have assessed the safety and efficacy of this treatment in patients with locally advanced disease. AIM: To analyze long-term outcomes of pembrolizumab in locally advanced or metastatic esophageal squamous cell carcinoma (ESCC) in the real world. METHODS: Patients with advanced ESCC admitted to our center from October 2019 to October 2021 were enrolled in this study. Clinical staging of the patients was based on the 8th edition of the American Joint Committee on Cancer TNM staging system. The patients received different treatments based on clinical stage. In brief, patients with locally advanced and resectable ESCC received neoadjuvant therapy combined with surgery. For those who were not candidates for resection, radical concurrent chemoradiotherapy plus pembrolizumab was more preferable. Patients with metastatic ESCC or who were unsuitable for radiotherapy underwent chemotherapy in combination with pembrolizumab. Long-term survival outcomes such as overall survival (OS), progression-free survival, disease-free survival, long-term adverse effects (AEs), immune maintenance therapy and predictors of immune checkpoint inhibitors (ICIs) efficacy were evaluated. RESULTS: A total of 55 patients with advanced ESCC were enrolled in this retrospective, observational study. The median age was 61 years (range 44-74), with 47.3% (26/55) of the patients in stage IV and 45.5% of the patients had the tumor (25/55) located in the middle third of the esophagus. The median OS in all patients was not reached. The 12-mo OS rate among all patients was 78.8% and the 18-mo OS rate was 72.7%. 9 patients died due to tumor progression and 7 patients died due to treatment-related complications. The therapeutic effect evaluated at the interim evaluation was significantly reflected in the long-term outcome. Patients with complete response or partial response in all patients (P = 0.005) and in the chemoradiotherapy plus pembrolizumab group (P = 0.007) obtained a better prognosis than non-responders. A total of 20 patients (20/55, 36%) received immune maintenance therapy. Baseline peripheral blood biomarkers of the neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and neutrophil-to-(leukocyte-neutrophil) ratio did not predict the efficacy of ICIs. CONCLUSION: Pembrolizumab combined with chemotherapy or radiotherapy resulted in favorable long-term survival in patients with locally advanced or metastatic ESCC, with safe and manageable long-term AEs.

3D printed microstructured ultra-sensitive pressure sensors based on microgel-reinforced double network hydrogels for biomechanical applications.

Hydrogel-based wearable flexible pressure sensors have great promise in human health and motion monitoring. However, it remains a great challenge to significantly improve the toughness, sensitivity and stability of hydrogel sensors. Here, we demonstrate the fabrication of hierarchically structured hydrogel sensors by 3D printing microgel-reinforced double network (MRDN) hydrogels to achieve both very high sensitivity and mechanical toughness. Polyelectrolyte microgels are used as building blocks, which are interpenetrated with a second network, to construct super tough hydrogels. The obtained hydrogels show a tensile strength of 1.61 MPa, and a fracture toughness of 5.08 MJ m-3 with high water content. The MRDN hydrogel precursors exhibit reversible gel-sol transitions, and serve as ideal inks for 3D printing microstructured sensor arrays with high fidelity and precision. The microstructured hydrogel sensors show an ultra-high sensitivity of 0.925 kPa-1, more than 50 times that of plain hydrogel sensors. The hydrogel sensors are assembled as an array onto a shoe-pad to monitor foot biomechanics during gaiting. Moreover, a sensor array with a well-arranged spatial distribution of sensor pixels with different microstructures and sensitivities is fabricated to track the trajectory of a crawling tortoise. Such hydrogel sensors have promising application in flexible wearable electronic devices.

Machine Learning Approach Based on a Range-Corrected Deep Potential Model for Efficient Vibrational Frequency Computation.

As an ensemble average result, vibrational spectrum simulation can be time-consuming with high accuracy methods. We present a machine learning approach based on the range-corrected deep potential (DPRc) model to improve the computing efficiency. The DPRc method divides the system into "probe region" and "solvent region"; "solvent-solvent" interactions are not counted in the neural network. We applied the approach to two systems: formic acid C═O stretching and MeCN C≡N stretching vibrational frequency shifts in water. All data sets were prepared using the quantum vibration perturbation approach. Effects of different region divisions, one-body correction, cut range, and training data size were tested. The model with a single-molecule "probe region" showed stable accuracy; it ran roughly 10 times faster than regular deep potential and reduced the training time by about four. The approach is efficient, easy to apply, and extendable to calculating various spectra.

The relationship between mushroom consumption and cognitive performance among middle-aged and older adults: a cross-sectional study.

Objectives: Bioactive compounds in mushrooms may protect the brain from neurodegeneration by inhibiting the production of amyloid-ß and playing an antioxidant role. This study aimed at examining the associations of mushroom consumption with cognitive function and mild cognitive impairment (MCI) among middle-aged and older adults in China. Design: A cross-sectional study. Setting and participants: This study was conducted in seven cities in China and included 2203 middle-aged and older adults. Methods: Data on mushroom consumption were collected using a semi-quantitative food frequency questionnaire. Cognitive function was evaluated by the Auditory Verbal Learning Test (AVLT), Verbal Fluency Test (VFT), Digit Symbol Substitution Test (DSST), and Trail Making Test-B (TMT-B). The composite z score was used to reflect global cognition. MCI was determined according to the Petersen criteria. Multiple linear regression and logistic regression were used to examine the relationship between mushroom consumption and cognitive performance. Results: This study included 2203 participants aged 55 years and above (mean age = 63.43 years). After controlling demographic characteristics, lifestyle factors, other dietary factors, and history of chronic disease, higher mushroom consumption was associated with better global cognition. Compared to the lowest quartile (Q1, 0-4.00 g day-1), the ßs (95% confidence intervals, 95% CIs) were 0.10 (0.03, 0.18) for Q2 (4.01-10.42 g day-1), 0.13 (0.06, 0.20) for Q3 (10.43-20.84 g day-1), and 0.13 (0.06, 0.20) for Q4 (>20.84 g day-1). The higher mushroom consumption was positively related to better performance in DSST and TMT-B (P-values < 0.05). A 10 g day-1 increment in mushroom consumption was related to 12% lower odds of MCI (odds ratio = 0.88, 95% CI: 0.80-0.97). Conclusions: Higher mushroom consumption was positively related to better cognitive function and associated with lower odds of MCI. Further studies are needed to replicate our findings in other populations and determine the underlying mechanisms.

Sonocatalytic Optimization of Titanium-Based Therapeutic Nanomedicine.

Recent considerable technological advances in ultrasound-based treatment modality provides a magnificent prospect for scientific communities to conquer the related diseases, which is featured with remarkable tissue penetration, non-invasive and non-thermal characteristics. As one of the critical elements that influences treatment outcomes, titanium (Ti)-based sonosensitizers with distinct physicochemical properties and exceptional sonodynamic efficiency have been applied extensively in the field of nanomedical applications. To date, a myriad of methodologies has been designed to manipulate the sonodynamic performance of titanium-involved nanomedicine and further enhance the productivity of reactive oxygen species for disease treatments. In this comprehensive review, the sonocatalytic optimization of diversified Ti-based nanoplatforms, including defect engineering, plasmon resonance modulation, heterojunction, modulating tumor microenvironment, as well as the development of synergistic therapeutic modalities is mainly focused. The state-of-the-art Ti-based nanoplatforms ranging from preparation process to the extensive medical applications are summarized and highlighted, with the goal of elaborating on future research prospects and providing a perspective on the bench-to-beside translation of these sonocatalytic optimization tactics. Furthermore, to spur further technological advancements in nanomedicine, the difficulties currently faced and the direction of sonocatalytic optimization of Ti-based therapeutic nanomedicine are proposed and outlooked.

Association of plant-based diet index with sleep quality in middle-aged and older adults: The Healthy Dance Study.

OBJECTIVES: To investigate the association between plant-based diet indices and sleep quality in Chinese middle-aged and older adults. METHODS: The study included 2424 participants aged 45 years and older. Dietary data were collected using a semi-quantitative food frequency questionnaire, and sleep quality was assessed by the Pittsburgh Sleep Quality Index scale. Plant-based diet was categorized based on 3 indices (score range, 17-85) covering 17 food groups: the overall plant-based diet index, healthful plant-based diet index, and unhealthful plant-based diet index. The associations between these plant-based diet indices and sleep quality were examined using logistic and linear regression analyses. RESULTS: After controlling for sociodemographic, lifestyle, and multiple disease-related factors, participants in the highest quartile of the healthful plant-based diet index had 0.55 higher odds of better sleep quality (95% CI: 0.42, 0.72; Ptrend< .001). In contrast, participants in the highest quartile of the unhealthful plant-based diet index had 2.03 higher odds of poor sleep quality (95% CI: 1.51, 2.72; Ptrend< .001). In addition, plant-based diet index and healthful plant-based diet index were inversely associated with Pittsburgh Sleep Quality Index scores, while unhealthful plant-based diet index and Pittsburgh Sleep Quality Index scores were positively associated. CONCLUSIONS: We found unhealthy plant-based diets are significantly associated with poor sleep quality. Adherence to overall plant-based diets, especially healthy plant-based diets, was positively associated with optimal sleep quality.

Erasable, Rewritable, and Reprogrammable Dual Information Encryption Based on Photoluminescent Supramolecular Host-Guest Recognition and Hydrogel Shape Memory.

Information encryption technologies are very important for security, health, commodity, and communications, etc. Novel information encryption mechanisms and materials are desired to achieve multimode and reprogrammable encryption. Here, a supramolecular strategy is demonstrated to achieve multimodal, erasable, reprogrammable, and reusable information encryption by reversibly modulating fluorescence. A butyl-naphthalimide with flexible ethylenediamine functionalized ß-cyclodextrin (N-CD) is utilized as a fluorescent responsive ink for printing or patterning information on polymer brushes with dangling adamantane group grafted on responsive hydrogels. The photoluminescent naphthalimide moiety is bonded to ß-CD and entrapped in the cavity. Its fluorescence is highly weakened in ß-CD cavity and recovers after being expelled from the cavity by a competing guest molecule to emit bright green photoluminescence under UV. Experiments and theoretical calculations suggest π-π stacking and ICT as the primary mechanism for the naphthalimides assembly and fluorescence, which can be quenched through insertion of conjugated molecules and recover by removing the insert. Such reversible quenching and recovering are used to achieve repeated writing, erasing, and re-writing of information. Supramolecular recognition and hydrogel shape memory are further combined to achieve reversible dual-encryption. This study provides a novel strategy to develop smart materials with improved information security for broad applications.

The Prevalence and Agreement of Sarcopenic Obesity Using Different Definitions and Its Association with Mild Cognitive Impairment.

BACKGROUND: The consistent definition of sarcopenic obesity (SO) is limited, its association with mild cognitive impairment (MCI) has not been clarified. OBJECTIVE: This study aimed to evaluate the prevalence and agreement of SO using different definitions and the association between SO and MCI. METHODS: SO was diagnosed by the co-existence of sarcopenia defined by the Asia Working Group for Sarcopenia (AWGS) and obesity by body mass index (BMI), visceral fat area (VFA), waist circumference (WC), or body fat percentage (BF%). Cohen's kappa was used to assess the agreement between the different definitions. The association between SO and MCI was assessed using multivariable logistic regression. RESULTS: Among 2,451 participants, the prevalence of SO ranged from 1.7% to 8.0% under different definitions. SO defined by AWGS and BMI (AWGS+BMI) showed fair agreements with the other three criteria (κ ranged from 0.334 to 0.359). The other criteria showed good agreements with each other. The κ statistics were 0.882 for AWGS+VFA and AWGS+BF%, 0.852 for AWGS+VFA and AWGS+WC, and 0.804 for AWGS+BF% and AWGS+WC, respectively. When using different diagnoses of SO, compared with the health group, the adjusted ORs of MCI for SO were 1.96 (95% CI: 1.29-2.99, SO: AWGS+WC), 1.75 (95% CI: 1.14-2.68, SO: AWGS+VFA), 1.94 (95% CI: 1.29-2.93, SO: AWGS+BF%), and 1.45 (95% CI: 0.67-3.12, SO: AWGS+BMI), respectively. CONCLUSION: Using different obesity indicators combined with AWGS to diagnose SO, BMI had lower prevalence and agreement compared with other three indicators. SO was associated with MCI under different methods (WC, VFA, or BF%).

InOR-Net: Incremental 3-D Object Recognition Network for Point Cloud Representation.

3-D object recognition has successfully become an appealing research topic in the real world. However, most existing recognition models unreasonably assume that the categories of 3-D objects cannot change over time in the real world. This unrealistic assumption may result in significant performance degradation for them to learn new classes of 3-D objects consecutively due to the catastrophic forgetting on old learned classes. Moreover, they cannot explore which 3-D geometric characteristics are essential to alleviate the catastrophic forgetting on old classes of 3-D objects. To tackle the above challenges, we develop a novel Incremental 3-D Object Recognition Network (i.e., InOR-Net), which could recognize new classes of 3-D objects continuously by overcoming the catastrophic forgetting on old classes. Specifically, category-guided geometric reasoning is proposed to reason local geometric structures with distinctive 3-D characteristics of each class by leveraging intrinsic category information. We then propose a novel critic-induced geometric attention mechanism to distinguish which 3-D geometric characteristics within each class are beneficial to overcome the catastrophic forgetting on old classes of 3-D objects while preventing the negative influence of useless 3-D characteristics. In addition, a dual adaptive fairness compensations' strategy is designed to overcome the forgetting brought by class imbalance by compensating biased weights and predictions of the classifier. Comparison experiments verify the state-of-the-art performance of the proposed InOR-Net model on several public point cloud datasets.

Open-Ended Online Learning for Autonomous Visual Perception.

The visual perception systems aim to autonomously collect consecutive visual data and perceive the relevant information online like human beings. In comparison with the classical static visual systems focusing on fixed tasks (e.g., face recognition for visual surveillance), the real-world visual systems (e.g., the robot visual system) often need to handle unpredicted tasks and dynamically changed environments, which need to imitate human-like intelligence with open-ended online learning ability. Therefore, we provide a comprehensive analysis of open-ended online learning problems for autonomous visual perception in this survey. Based on "what to online learn" among visual perception scenarios, we classify the open-ended online learning methods into five categories: instance incremental learning to handle data attributes changing, feature evolution learning for incremental and decremental features with the feature dimension changed dynamically, class incremental learning and task incremental learning aiming at online adding new coming classes/tasks, and parallel and distributed learning for large-scale data to reveal the computational and storage advantages. We discuss the characteristic of each method and introduce several representative works as well. Finally, we introduce some representative visual perception applications to show the enhanced performance when using various open-ended online learning models, followed by a discussion of several future directions.

A Comprehensive Study of 3-D Vision-Based Robot Manipulation.

Robot manipulation, for example, pick-and-place manipulation, is broadly used for intelligent manufacturing with industrial robots, ocean engineering with underwater robots, service robots, or even healthcare with medical robots. Most traditional robot manipulations adopt 2-D vision systems with plane hypotheses and can only generate 3-DOF (degrees of freedom) pose accordingly. To mimic human intelligence and endow the robot with more flexible working capabilities, 3-D vision-based robot manipulation has been studied. However, this task is still challenging in the open world especially for general object recognition and pose estimation with occlusion in cluttered backgrounds and human-like flexible manipulation. In this article, we propose a comprehensive analysis of recent progress about the 3-D vision for robot manipulation, including 3-D data acquisition and representation, robot-vision calibration, 3-D object detection/recognition, 6-DOF pose estimation, grasping estimation, and motion planning. We then present some public datasets, evaluation criteria, comparisons, and challenges. Finally, the related application domains of robot manipulation are given, and some future directions and open problems are studied as well.

The Accuracy of Semi-Empirical Quantum Chemistry Methods on Soot Formation Simulation.

Soot molecules are hazardous compounds threatening human health. Computational chemistry provides efficient tools for studying them. However, accurate quantum chemistry calculation is costly for the simulation of large-size soot molecules and high-throughput calculations. Semi-empirical (SE) quantum chemistry methods are optional choices for balancing computational costs. In this work, we validated the performances of several widely used SE methods in the description of soot formation. Our benchmark study focuses on, but is not limited to, the validation of the performances of SE methods on reactive and non-reactive MD trajectory calculations. We also examined the accuracy of SE methods of predicting soot precursor structures and energy profiles along intrinsic reaction coordinate(s) (IRC). Finally, we discussed the spin density predicted by SE methods. The SE methods validated include AM1, PM6, PM7, GFN2-xTB, DFTB2, with or without spin-polarization, and DFTB3. We found that the shape of MD trajectory profiles, the relative energy, and molecular structures predicted by SE methods are qualitatively correct. We suggest that SE methods can be used in massive reaction soot formation event sampling and primary reaction mechanism generation. Yet, they cannot be used to provide quantitatively accurate data, such as thermodynamic and reaction kinetics ones.

Hydrogel-Tissue Interface Interactions for Implantable Flexible Bioelectronics.

Hydrogels have emerged as multifunctional interface materials between implantable bioelectronic devices and biotissues. The soft and wet materials with low and alterable mechanical properties can match the mechanical, chemical, electrical, and biological properties of biotissues and thus diminish the mechanical and electrical mismatch. Interactions at the hydrogel-biotissue and hydrogel-device interfaces have attracted broad research interest. Great efforts have been devoted to establishing instant, strong, and conformal adhesion at the interface by chemical bonding, electrostatic interaction, hydrogen bonding, supramolecular recognition, hydrophobic association, and even topological entanglements at the interfaces. This Perspective provides a brief account of representative progress on the hydrogel-tissue adhesive that forms seamless and conformal interface adhesion and applications in implantable devices for physiological, cardiac, and neuronal signal collection and electrical stimulation. Major challenges such as wet adhesion and the stability of the adhesive hydrogel-tissue interface are identified for examination in future investigations.

Poor sleep quality is associated with fatigue and depression in early Parkinson's disease: A longitudinal study in the PALS cohort.

Background: Sleep disorders are common in Parkinson's disease (PD). However, the longitudinal relationship between sleep quality and the other non-motor symptoms of PD has not been well characterized, especially in early PD. Objective: To explore the value of baseline sleep quality in predicting the progression of other non-motor symptoms in early PD. Methods: 109 early PD patients were recruited to the study. Patients were stratified into good and poor sleepers using the Pittsburgh Sleep Quality Index (PSQI). Assessments performed at baseline and 1 year follow-up included the Epworth Sleepiness Scale, Fatigue Severity Scale, Non-Motor Symptom Scale, Geriatric Depression Scale, Hospital Anxiety and Depression Scale, Apathy Scale, Montreal Cognitive Assessment and detailed neuropsychological assessments. Multivariable linear regression was performed at baseline to investigate differences in clinical scores between poor and good sleepers, while multivariable regression models were used to investigate associations between sleep quality and progression of test scores at 1 year follow-up. Results: 59 poor sleepers and 50 good sleepers were identified. At baseline, poor sleepers had greater HADS anxiety scores (p = 0.013) [2.99 (95% CI 2.26, 3.73)] than good sleepers [1.59 (95% CI 0.75, 2.42)]. After 1 year, poor sleepers had greater fatigue (FSS scores +3.60 as compared to -2.93 in good sleepers, p = 0.007) and depression (GDS scores +0.42 as compared to -0.70, p = 0.006). Conclusion: This study shows a longitudinal association between sleep quality, fatigue, and depression in early PD patients, independent of medication effect and disease severity, this may support the hypothesis that a common serotonergic pathway is implicated in these non-motor symptoms.

Genotypic diversity and antifungal susceptibility of Scedosporium species from clinical settings in China.

BACKGROUND: Scedosporium species have drawn significant interest as inhabitants of polluted soil and water and as cause of high mortality in near-drowning patients. So far, most cases have been reported from Europe and Australia, while knowledge on their prevalence and genotypic diversity from Asia is scant. OBJECTIVES: To increase the knowledge of the genetic diversity and in vitro antifungal susceptibility of Scedosporium species involved in human infections from China. METHODS: Here, we applied the ISHAM-MLST consensus scheme for molecular typing of Scedosporium species and revealed both high species diversity and high genotypic diversity among 45 Chinese clinical Scedosporium isolates. RESULTS: Among the five species, Scedosporium boydii (n = 22) was the most common, followed by S. apiospermum (n = 18), S. aurantiacum (n = 4) and S. dehoogii (n = 1). S. aurantiacum was reported for the first time from clinical samples in China. The predominant sequence types (STs) were ST17 in S. apiospermum, ST4 in S. boydii and ST92 in S. aurantiacum, including four novel STs (ST40, ST41, ST42 and ST43) in S. apiospermum. Based on the CLSI-M38 A2 criterion, voriconazole was the only antifungal compound with low MIC values (MIC90 ≤ 1 µg/ml) for all Scedosporium isolates in our study. CONCLUSIONS: The genetic diversity of clinical isolates of Scedosporium species from China is extremely high, with S. boydii being predominant and S. aurantiacum being firstly reported here. VOR was the only antifungal compound with low MIC values for all Scedosporium isolates in our study, which should be recommended as the firstline antifungal treatment against scedosporiosis in China.

Dual Aligned Siamese Dense Regression Tracker.

Anchor or anchor-free based Siamese trackers have achieved the astonishing advancement. However, their parallel regression and classification branches lack the tracked target information link and interaction, and the corresponding independent optimization maybe lead to task-misalignment, such as the reliable classification prediction with imprecisely localization and vice versa. To address this problem, we develop a general Siamese dense regression tracker (SDRT) with both task and feature alignments. It consists of two cooperative and mutual-guidance core branches: dense local regression with RepPoint representation, the global and local multi-classifier fusion with aligned features. They complement and boost each other to constrain the results with well-localized followed to also be well-classified. Specifically, a dense local regression with RepPoint representation, directly estimates and averages multiple dense local bounding box offsets for accurate localization. And then, the refined bounding boxes can be used to learn the global and local affine alignment features for reliable multi-classifier fusion. The classified scores in turn guide the assigned positive bounding boxes for the regression task. The mutual guidance operations can bridge the connection between classification and regression substantially, since the assigned labels of one task depend on the prediction quality of the other task. The proposed tracking module is general, and it can boost both the anchor or anchor-free based Siamese trackers to some extent. The extensive tracking comparisons on six tracking benchmarks verify its favorable and competitive performance over states-of-the-arts tracking modules.

Representative Task Self-Selection for Flexible Clustered Lifelong Learning.

Consider the lifelong machine learning paradigm whose objective is to learn a sequence of tasks depending on previous experiences, e.g., knowledge library or deep network weights. However, the knowledge libraries or deep networks for most recent lifelong learning models are of prescribed size and can degenerate the performance for both learned tasks and coming ones when facing with a new task environment (cluster). To address this challenge, we propose a novel incremental clustered lifelong learning framework with two knowledge libraries: feature learning library and model knowledge library, called Flexible Clustered Lifelong Learning (FCL3). Specifically, the feature learning library modeled by an autoencoder architecture maintains a set of representation common across all the observed tasks, and the model knowledge library can be self-selected by identifying and adding new representative models (clusters). When a new task arrives, our FCL3 model firstly transfers knowledge from these libraries to encode the new task, i.e., effectively and selectively soft-assigning this new task to multiple representative models over feature learning library. Then: 1) the new task with a higher outlier probability will be judged as a new representative, and used to redefine both feature learning library and representative models over time; or 2) the new task with lower outlier probability will only refine the feature learning library. For model optimization, we cast this lifelong learning problem as an alternating direction minimization problem as a new task comes. Finally, we evaluate the proposed framework by analyzing several multitask data sets, and the experimental results demonstrate that our FCL3 model can achieve better performance than most lifelong learning frameworks, even batch clustered multitask learning models.