Search

The development of reusable artificial intelligence (AI) models for wider use and rigorous validation by the community promises to unlock new opportunities in multi-messenger astrophysics. Here we develop a workflow that connects the Data and Learning Hub for Science, a repository for publishing AI models, with the Hardware Accelerated Learning (HAL) cluster, using funcX as a universal distributed computing service. Using this workflow, an ensemble of four openly available AI models can be run on HAL to process an entire month's worth (August 2017) of advanced Laser Interferometer Gravitational-Wave Observatory data in just seven minutes, identifying all four all four binary black hole mergers previously identified in this dataset and reporting no misclassifications. This approach combines advances in AI, distributed computing, and scientific data infrastructure to open new pathways to conduct reproducible, accelerated, data-driven discovery., 17 pages, 5 figures; v2: 12 pages, 6 figures. Accepted to Nature Astronomy. See also the Behind the Paper blog in Nature Astronomy "https://astronomycommunity.nature.com/posts/from-disruption-to-sustained-innovation-artificial-intelligence-for-gravitational-wave-astrophysics"

Seismograms are convolution results between seismic sources and the media that seismic waves propagate through, and, therefore, the primary observations for studying seismic source parameters and the Earth interior. The routine earthquake location and travel-time tomography rely on accurate seismic phase picks (e.g., P and S arrivals). As data increase, reliable automated seismic phase-picking methods are needed to analyze data and provide timely earthquake information. However, most traditional autopickers suffer from low signal-to-noise ratio and usually require additional efforts to tune hyperparameters for each case. In this study, we proposed a deep-learning approach that adapted soft attention gates (AGs) and recurrent-residual convolution units (RRCUs) into the backbone U-Net for seismic phase picking. The attention mechanism was implemented to suppress responses from waveforms irrelevant to seismic phases, and the cooperating RRCUs further enhanced temporal connections of seismograms at multiple scales. We used numerous earthquake recordings in Taiwan with diverse focal mechanisms, wide depth, and magnitude distributions, to train and test our model. Setting the picking errors within 0.1 s and predicted probability over 0.5, the AG with recurrent-residual convolution unit (ARRU) phase picker achieved the F1 score of 98.62% for P arrivals and 95.16% for S arrivals, and picking rates were 96.72% for P waves and 90.07% for S waves. The ARRU phase picker also shown a great generalization capability, when handling unseen data. When applied the model trained with Taiwan data to the southern California data, the ARRU phase picker shown no cognitive downgrade. Comparing with manual picks, the arrival times determined by the ARRU phase picker shown a higher consistency, which had been evaluated by a set of repeating earthquakes. The arrival picks with less human error could benefit studies, such as earthquake location and seismic tomography.

A three-phase PANI/BaFe12O19@Hal heterostructure was designed and fabricated in this paper as efficient lightweight electromagnetic wave absorbing material through the combination of citrate assisted sol–gel self-propagating combustion and in-situ oxidative polymerization of aniline. In addition, the effects of the weight ratio of different PANI to BF@Hal on the microwave absorption properties of the materials were studied. The results show that when the weight ratio of PANI is 40%, the material has the best microwave absorption performance. The frequency bandwidth below −5 dB reached 9.60 GHz and the minimum absorption peak at 11.92 GHz was −14.77 dB. The combination of the PANI and BF@Hal nanosheets take advantage of the interfacial polarization, natural resonance, dielectric polarization and trapping of EM waves by internal reflection in PANI/BaFe12O19@Hal. Taking advantage of the unique microscopic morphology and interface characteristics, halloysite was introduced to improve the microwave absorption performance and enrich the absorbing mechanism of the composite materials. This work may provide a reliable candidate for the synthesis of electromagnetic attenuation materials with fairly good performances.

The template-matching algorithm (TMA) has become an important tool for detecting small and/or unconventional earthquakes, and newly detected seismic events have improved our understanding of earthquake physics, regional tectonics, and geological hazards. Standard TMA-based detection methods do not take into account the fact that the template waveforms themselves are contaminated with noise. In this study, we propose a weighted template-matching algorithm (WTMA), in which the normalized cross-correlation time series are weighted according to the signal-to-noise ratios of the corresponding template waveforms before they are being stacked for event picking. We present results from an extensive collection of numerical experiments to demonstrate that WTMA is capable of improving the detection rate of small to tiny (e.g., ML

We present the details of a GPU capable exchange correlation (XC) scheme integrated into the open source QUantum Interaction Computational Kernel (QUICK) program. Our implementation features an octree based numerical grid point partitioning scheme, GPU enabled grid pruning and basis/primitive function prescreening and fully GPU capable XC energy and gradient algorithms. Benchmarking against the CPU version demonstrated that the GPU implementation is capable of delivering an impres- sive performance while retaining excellent accuracy. For small to medium size protein/organic molecular systems, the realized speedups in double precision XC energy and gradient computation on a NVIDIA V100 GPU were 60 to 80-fold and 140 to 780- fold respectively as compared to the serial CPU implementation. The acceleration gained in density functional theory calculations from a single V100 GPU significantly exceeds that of a modern CPU with 40 cores running in parallel.

Foreshocks and/or aftershocks play critical roles in improving our understanding of the processes of faulting, such as nucleation of earthquakes, earthquake triggering, and postseismic deformation. A rapid and accurate earthquake detection and location algorithm can provide timely information of seismic activities, thereby benefitting our understanding of physical mechanisms of faulting and seismic hazard assessment. We have developed a graphic processing unit (GPU)-accelerated automatic microseismic monitoring algorithm (GAMMA) for accurate and near real-time detection and location of earthquakes. GAMMA utilizes methods based on backprojection to automatically detect potential earthquakes, and then the waveforms of qualified earthquakes are selected as templates when searching for small earthquakes in continuous recordings using the template-matching algorithm. The use of GPUs has substantially accelerated the calculations and has made GAMMA capable of (near-)real-time earthquake monitoring. We have successfully applied GAMMA to the 2019 Ridgecrest earthquake sequence in southern California. The number of earthquakes detected by GAMMA is more than 21 times that documented in the regional catalog. The more complete catalog determined by GAMMA may provide crucial information for improving our understanding of the physical mechanisms of faulting and also supply useful constraints for a variety of types of studies, including dynamic rupture simulations and crustal deformation modeling.

Seismic tomography, when calibrated with borehole data, can be a highly effective tool for investigating the deep structure of the critical zone. In this study, we have successfully applied seismic adjoint tomography to obtain a fully three-dimensional shear-wave velocity model under a granitic ridge in the Blair-Wallis watershed, southeastern Wyoming. The seismic data used in our tomography are ambient-noise Green's functions obtained from a minimally invasive, “large-N” seismic survey using a square array of 400 autonomous geophones. Lateral variations of our 3D shear-wave velocity model show strong correlations with surface topography. Depth variations of our velocity model may give clues about the depth interval within which chemical weathering plays a significant role. With the calibration of borehole casing depths and saprolite refusal depths, we selected the isosurface of shear-wave velocity 491 m/s as the interface between highly chemically weathered saprolite and fractured bedrocks, which we use to represent the weathering front in this study. Large-scale spatial variations of the weathering front are consistent with groundwater table, which provides observational support to a recently proposed hypothesis that subsurface weathering is mainly driven by bedrock drainage of chemically equilibrated water. Small-scale spatial variations of the weathering front calls for more sophisticated mechanisms that couple the effect of top-down infiltration of reactive meteoric water with the influence of pre-existing fractures.

We report a new multi-GPU capable ab initio Hartree-Fock/density functional theory implementation integrated into the open source QUantum Interaction Computational Kernel (QUICK) program. Details on the load balancing algorithms for electron repulsion integrals and exchange correlation quadrature across multiple GPUs are described. Benchmarking studies carried out on up to four GPU nodes, each containing four NVIDIA V100-SXM2 type GPUs demonstrate that our implementation is capable of achieving excellent load balancing and high parallel efficiency. For representative medium to large size protein/organic molecular systems, the observed parallel efficiencies remained above 82% for the Kohn-Sham matrix formation and above 90% for nuclear gradient calculations. The accelerations on NVIDIA A100, P100, and K80 platforms also have realized parallel efficiencies higher than 68% in all tested cases, paving the way for large-scale ab initio electronic structure calculations with QUICK.

Finding new ways to use artificial intelligence (AI) to accelerate the analysis of gravitational wave data, and ensuring the developed models are easily reusable promises to unlock new opportunities in multi-messenger astrophysics (MMA), and to enable wider use, rigorous validation, and sharing of developed models by the community. In this work, we demonstrate how connecting recently deployed DOE and NSF-sponsored cyberinfrastructure allows for new ways to publish models, and to subsequently deploy these models into applications using computing platforms ranging from laptops to high performance computing clusters. We develop a workflow that connects the \texttt{Data and Learning Hub for Science} (DLHub), a repository for publishing machine learning models, with the Hardware Accelerated Learning (HAL) deep learning computing cluster, using funcX as a universal distributed computing service. We then use this workflow to search for binary black hole gravitational wave signals in open source advanced LIGO data. We find that using this workflow, an ensemble of four openly available deep learning models can be run on HAL and process the entire month of August 2017 of advanced LIGO data in just seven minutes, identifying all four binary black hole mergers previously identified in this dataset, and reporting no misclassifications. This approach, which combines advances in AI, distributed computing, and scientific data infrastructure opens new pathways to conduct reproducible, accelerated, data-driven gravitational wave detection.

A three-phase PANI/BaFe

Objective: To evaluate the diagnostic value of fluorescence in situ hybridization (FISH) in bladder cancer. Methods: We enrolled healthy volunteers and patients who were clinically suspected to have bladder cancer and conducted FISH tests and cytology examinations from August 2007 to December 2008. Receiver operating characteristic (ROC) curve analysis was performed and the area under curve (AUC) values were calculated for both the FISH and urine cytology tests. Results: A cohort of 988 healthy volunteers was enrolled to establish a reference range for the normal population. A total of 4807 patients with hematuria were prospectively, randomly enrolled for the simultaneous analysis of urine cytology, FISH testing, and a final diagnosis as determined by the pathologic findings of a biopsy or a surgically-excised specimen. Overall, the sensitivity of FISH in detecting transitional-cell carcinoma was 82.7%, while that of cytology was 33.4% (p

Background Aurora A kinase is frequently overexpressed in a variety of tumor types, including the prostate. However, the function of Aurora A in autophagy in prostate cancer has not been investigated. Here, we aimed to study the functioning mechanism and autophagy associated signaling pathways of Aurora A in prostate cancer. Methods To investigate the biological function of Aurora A, down-regulation of Aurora A was performed followed by functional testing assays. Immunohistochemistry was used to detect the expression of Aurora A in human prostate cancer specimens. CCK8, Transwell, flow cytometric analysis and measurement of tumor formation in nude mice were performed to test the effects of Aurora A down-regulation in vivo and in vitro. Signaling pathway analysis was performed by using Western blot. Autophagy activity was measured by monitoring the expression levels of LC3-II. Results Aurora A overexpression was significantly higher in human prostate cancer specimens than in BPH. Furthermore, Aurora A knockdown inhibited the proliferation of prostate cancer cells by suppressing the Akt pathway, indicating that Akt is a novel Aurora A substrate in prostate cancer. Additionally, Aurora A down-regulation prompts autophagy in prostate cancer cells. Most importantly, Aurora A ablation almost fully abrogates tumorigenesis in nude mice, suggesting that Aurora A is a key oncogenic effector in prostate cancer. Conclusions Taken together, our data suggest that Aurora-A plays an important role in the suppression of autophagy by inhibiting the phosphorylation of Akt, which in turn prevents autophagy-induced apoptosis in prostate cancer.

A barium ferrite/amorphous carbon nanotubes composite (BaF/ACNTs) was synthesized via solvothermal method after ACNTs were prepared by using halloysite as the template. The fabricated composite exhibited hexagonal BaF nanosheets surrounded by ACNTs with partially graphitized structures. The effect of weight ratio of BaF/ACNTs on the microwave absorption properties was studied. When the ratio was 40/60, the material possessed the best microwave absorption performance. In the frequency range of 2–18 GHz, the frequency bandwidth below −10 dB reached 3.92 GHz, and the maximum absorption peak at 7.92 GHz was −43.82 dB. The mechanism of the improvement of microwave adsorption performance was explored. The conductive ability of ACNTs was used, then, the unique hollow tubular micromorphology of the composite material led to multiple reflections of electromagnetic waves and promotes the loss of electromagnetic energy. This work may provide a reliable path for the synthesis of electromagnetic attenuation materials with fairly good performances.

Previously glaciated landscapes are often covered with glacial sediments with poorly sorted particles. The thickness of glacial sediments can reach hundreds of meters. These sediments, which have experienced high intensity erosion through glacial transportation, are susceptible to chemical weathering processes because of the increased reactive mineral surfaces induced via erosion. Infiltrated meteoric water will first attack the glacial sediments, becoming (partially) saturated with weathering products, before interacting with the local bedrock underneath. As a consequence, the glacial deposits will hinder the weathering processes of the underlying local bedrock. In this study, we test this hypothesis by imaging the weathered layer of a previously glaciated terrain in the Medicine Bow Mountains, Wyoming using full-3D seismic ambient-noise tomography. To the best of our knowledge, our study provides the first full-3D seismic velocity model of the Critical Zone structure under a previously glaciated landscape. Our 3D seismic velocity model reveals that the saprolite weathered from local bedrock is thinner where the overlying glacial deposit layer is thicker. This anti-correlation suggests that the glacial deposits may locally protect the underlying bedrock from chemical weathering by depleting reactive chemicals in the meteoric water and (partially) saturate the meteoric water with weathering products. Our study may provide a potential counter mechanism against glaciation-induced acceleration of chemical weathering of the continents, which regulates atmospheric concentration of carbon dioxide and the global climate.

Recent advances in full-3D seismic tomography (F3DT), in which the starting model and the inverted model perturbations are 3D in space and the sensitivity (Frechet) kernels are computed using purely numerical solutions of 3D (visco) elastodynamic equations, have opened up new possibilities for imaging the structures of the near surface where mechanical properties of materials play important roles in various geotechnical engineering applications. In this study, we have successfully applied F3DT to image the near-surface structure under a granitic ridge in the Blair Wallis watershed, southeastern Wyoming. The seismic data used in our tomography consist of both 2D active-source seismic refraction data and 3D ambient-noise Green’s functions obtained from a minimally invasive “large-N” seismic survey using a square array of 400 autonomous geophones. Lateral variations of our 3D velocity model show strong correlations with surface topography. Depth variations of our velocity model may give clues about the depth interval within which chemical weathering plays a significant role. With the calibration of borehole casing depths and saprolite refusal depths, our F3DT technique can be a highly effective tool for investigating the structure of the near surface.

Barium carbonate (BaCO3) materials with the controllable morphology of nanoparticles were selectively loaded into the lumen halloysite nanotubes (abbreviated as Hal) by a urease assisted catalytic implementation strategy. The Hal mineral was pre-treated through leaching by hydrochloric acid (abbreviated as A-Hal), resulting in increased defect sites and zeta potential. The negatively charged urease was loaded inside the positively charged A-Hal lumen, and then through the decomposition of urea catalyzed by urease to produce carbonate ions and ammonia. When Ba2+ diffused in, BaCO3 particles were selectively synthesized in the lumen of A-Hal, the pore channels of A-Hal effectively controlled the growth and aggregation of BaCO3 nanocrystals and their geometrical morphology. The obtained BaCO3/A-Hal-T was characterized by transmission electron microscopy, Fourier transformation infrared spectroscopy and X-ray diffraction, differential scanning calorimetry-thermogravimetry (DSC-TG). The BaCO3/A-Hal-T may provide a candidate for potential applications.

In this paper, a cost-effective strategy for fabricating silicon-carbon composites was designed to further improve the electrochemical performance and commercialization prospects of Si anodes for lithium-ion batteries (LIBs). Silicon-carbon fibers (CFs) were prepared by loading Si nanoparticles (SiNPs) on interconnected carbon fibers via an electrospinning technique (SiNPs@CFs). The Si nanoparticles were obtained by the reduction reaction of natural clay minerals. As a flexible anode for LIBs, the SiNPs@CFs anode demonstrated a reversible capacity of 1238.1 mAh·g−1 and a capacity retention of 77% after 300 cycles (in contrast to the second cycle) at a current density of 0.5 A·g−1. With a higher current density of 5.0 A·g−1, the electrode showed a specific capacity of 528.3 mAh·g−1 after 1000 cycles and exhibited a superior rate capability compared to Si nanoparticles. The excellent electrochemical properties were attributed to the construction of flexible electrodes and the composite comprising carbon fibers, which lessened the volume expansion and improved the conductivity of the system.

We report in this article the systematical acquisition of physico-chemical parameters for two newly discovered halloysite (Hal) minerals from Shiyan and Tongling in China. As the comparative reference, the data from Hal in Linfen, Chenxi, and the salt lake in Australia (samples were abbreviated as Hal-AU, Hal-SY, Hal-LF, Hal-CX and Hal-TL, respectively) were also investigated using X-ray diffraction (XRD), scanning electronic microscopy (SEM), transmission electron microscopy (TEM), Fourier transformation infrared spectroscopy (FTIR), differential scanning calorimetry-thermogravimetry (DSC-TG), X-ray fluorescence, surface zeta potential measurements and N2 adsorption-desorption isotherms. The newly found minerals were probably formed in hydrothermal leaching and sedimentary circumstances. The Hal-SY contains 7 Å-halloysite and dickite, while Hal-TL contains 10 Å-halloysite with some alunite (similar with Hal-CX). Other impurities found in the samples include quartz, gibbsite, iron oxide and anatase. All of them showed tubular morphology with diameter in the range of 30–90 nm and a length of 300–2500 nm, while the Hal-SY has the largest inner diameter to about 150 nm. Specific surface areas varied from 26.0~59.0 m2·g−1. In addition, maximum CEC (cation exchange capacity) of the newly found Hal was about 40 cmol/kg, while that of Hal-AU was relatively low (8 cmol/kg) due to the sedimentary nature of Salt Lake circumstances. The surface charge was predominantly negative over most of the relevant pH range (>2.0). It can be concluded that the different morphology and impurity content of halloysite will greatly affect the surface area, pore volume, and cationic exchange capacity (CEC) of the minerals.

Ursolic acid has various pharmacological activities, and can reduce blood fat as well as having antihepatic, antitumoral, anti-inflammatory and antiviral properties. However, the pro-apoptotic mechanism by which ursolic acid influences human prostate cancer requires additional study. The aim of the present study was to assess whether ursolic acid activates the apoptosis of prostate cancer and to investigate the mechanism by which the Rho-associated protein kinase 1 (ROCK1)/phosphatase and tensin homolog (PTEN) signaling pathway performs a role in ursolic acid-mediated cofilin-1 to induce apoptosis in human prostate cancer. Firstly, the present study determined the pro-apoptotic mechanism by which ursolic acid influences the cell proliferation and apoptosis of human prostate LNCaP cancer cells. Caspase-3/9 activities and ROCK1, PTEN, Cofilin-1 and cytochrome c protein expression levels were also analyzed. In the present study, it is reported that the pro-apoptotic mechanism of ursolic acid potently suppressed the cell proliferation of human prostate LNCaP cancer cells. The present study revealed that the mediation of ROCK1/PTEN-cofilin-1/cytochrome c protein expression activates caspase-3/9 activities which subsequently induced the apoptosis of human prostate cancer cells. In conclusion, these findings demonstrated that ursolic acid activates the apoptosis of prostate cancer via ROCK/PTEN mediated cofilin-1/cytochrome c which mediated caspase-3/9 activities.

We have successfully ported an arbitrary high-order discontinuous Galerkin method for solving the three-dimensional isotropic elastic wave equation on unstructured tetrahedral meshes to multiple Graphic Processing Units (GPUs) using the Compute Unified Device Architecture (CUDA) of NVIDIA and Message Passing Interface (MPI) and obtained a speedup factor of about 28.3 for the single-precision version of our codes and a speedup factor of about 14.9 for the double-precision version. The GPU used in the comparisons is NVIDIA Tesla C2070 Fermi, and the CPU used is Intel Xeon W5660. To effectively overlap inter-process communication with computation, we separate the elements on each subdomain into inner and outer elements and complete the computation on outer elements and fill the MPI buffer first. While the MPI messages travel across the network, the GPU performs computation on inner elements, and all other calculations that do not use information of outer elements from neighboring subdomains. A significant portion of the speedup also comes from a customized matrix–matrix multiplication kernel, which is used extensively throughout our program. Preliminary performance analysis on our parallel GPU codes shows favorable strong and weak scalabilities.

There is increasing evidence suggesting that establishment of sister chromatid cohesion N-acetyltransferase 1 (ESCO1) was involved in tumorigenesis. However, its role in bladder cancer remains unclear. In this study, we aimed to study the clinical correlation and biological significance of ESCO1 in bladder cancer. Our results showed that ESCO1 was significantly over-expressed in bladder cancer tissues compared with that in adjacent normal tissues. And, increased ESCO1 expression was significantly associated with higher grade (P

This study investigates the efficacy of benign prostatic hyperplasia (BPH) treatment by prostate transurethral enucleation using a 2-μm laser. A total of 107 patients with BPH were treated by prostate transurethral enucleation using a RevoLix 2-μm laser surgery system. Bleeding volume, operation time, catheterization time, voiding situation, maximum urinary flow rate, and hospital stay were observed. The mean operation time was 74 min ± 12 min (range 45 to 150 min), the mean follow-up period was 2 to 6 months, the mean catheter time was 5 days, and the mean peak urinary flow rate increased from 6.3 ± 0.6 to 17.5 ± 1.5 mL/s. The International Prostate Symptom Score and quality of life significantly declined (p

SOX genes play an important role in a number of developmental processes. The transcription factor SOX11 is one of the members of the SOX family emerging as important transcriptional regulators. The aim of this study was to investigate the role of SOX11 in prostate cancer (PCa) and its expression pattern and clinical significance. The gene expression of SOX11 in human PCa tissues compared with benign prostate hyperplasia (BPH) tissues was detected using real-time quantitative reverse transcriptase-polymerase chain reaction (QRT-PCR) analysis and immunohositochemistry. SOX11 overexpression cell model was used to examine the role of SOX11 in cell growth and metastasis in vitro. The results showed that the positive rate of SOX11 staining was 16.67 % (10/60) in cases of prostatic carcinoma and 81.67 % (49/60) in cases of BPH, and the difference of SOX11 expression between PCa and BPH was statistically significant (P < 0.001). SOX11 mRNA level was lowly expressed in PCa cell lines compared to RWPE-1. SOX11 overexpression suppresses PCa cell migration and invasion. In conclusion, our findings demonstrate that SOX11 could suppress cell proliferation, migration, and invasion of PCa in vitro.

In Vehicular ad hoc network (VANET), vehicular users (VUs) are capable of connecting to different access networks for interacting both safety related information and user application related information. In the overlapped area of heterogeneous networks, VUs are allowed to perform vertical handoff between various access networks. As the performance of vertical handoff plays an important role in guaranteeing user quality of service (QoS) and achieving system performance enhancement, it should be examined and evaluated thoroughly. In this paper, vertical handoff performance of VUs in a VANET scenario consisting of heterogeneously integrated universal mobile telecommunications system (UMTS) and wireless local area network (WLAN) is evaluated based on network simulator 2 (NS2). Multi-mode nodes, which are capable of accessing both networks and a media independent handover (MIH)-based vertical handoff scheme are implemented in NS2. The impacts of user velocity, the strength of traffic load and the type of handoff mechanisms on the communication performance of VUs, i.e., throughput, handoff latency and packet loss are examined.

To describe our experience in the diagnosis and treatment of 23 patients with ureteral endometriosis.We performed a retrospective analysis of 23 cases of ureteral endometriosis with histopathological results from 2002 to 2011.In patients with ureteral endometriosis, 23 cases were diagnosed by ultrasound, 21 by intravenous urography, 11 by retrograde urography, 16 by computed tomography, and 8 with magnetic resonance imaging. All cases were treated by operative treatment. The treatments included ureterolysis in 3 cases, partial ureteral resection and ureteroneocystostomy in 6 cases, partial ureteral resection and end-to-end ureteral anastomosis in 12 cases, and endoscopic resection of ureteral endometriosis lesion in 2 cases. All of the pathologic exam­ination results were endometriosis.Our findings suggest that surgery is an effective treatment option in most patients with ureteral endometriosis exhibiting mild or moderate to severe hydronephrosis. The type of technique depends on the location and depth of the lesion.

Next generation wireless network is expected to integrate multiple wireless technologies and provide mobile users with communication connectivity of different QoS requirements. However, the heterogeneity of access networks and the variety of user application requirements pose new challenges on access network selection scheme, i.e., selecting the optimal network for multi-interface mobile terminal (MT) in the case that multiple networks are available. In this paper, the theory of neural network (NN) is applied in designing network selection scheme for heterogeneous networks and an NN-based access network selection algorithm is proposed, which models the network characteristics based on network and user sample data and then evaluates the performance of candidate networks based on current network and service features, the candidate network with the best service performance is then chosen as the destination network Simulation results demonstrate that the proposed algorithm offers better QoS performance comparing to the traditional algorithm.