Search

Abstract Fundamental machine learning theory shows that different samples contribute unequally to both the learning and testing processes. Recent studies on deep neural networks (DNNs) suggest that such sample differences are rooted in the distribution of intrinsic pattern information, namely sample regularity. Motivated by recent discoveries in network memorization and generalization, we propose a pair of sample regularity measures with a formulation-consistent representation for both processes. Specifically, the cumulative binary training/generalizing loss (CBTL/CBGL), the cumulative number of correct classifications of the training/test sample within the training phase, is proposed to quantify the stability in the memorization-generalization process, while forgetting/mal-generalizing events (ForEvents/MgEvents), i.e., the misclassification of previously learned or generalized samples, are utilized to represent the uncertainty of sample regularity with respect to optimization dynamics. The effectiveness and robustness of the proposed approaches for mini-batch stochastic gradient descent (SGD) optimization are validated through sample-wise analyses. Further training/test sample selection applications show that the proposed measures, which share the unified computing procedure, could benefit both tasks.

Background and objectives: The inhibitor MLN4924 of Neural Precursor Cell-Expressed Developmentally Down-Regulated 8 (NEDD8) Activating Enzyme 1 (NAE1) has been found to suppress the growth of nasopharyngeal carcinoma (NPC). However, its effect on NPC's radiotherapy sensitivity remains unclear. Methods: By integrating single-cell RNA sequencing and bulk RNA sequencing, we predict the impact of NAE1 on the cell cycle, cell death, and its relationship with radiotherapy sensitivity and prognosis in NPC. The effect of inhibiting NAE1 on NPC cell behavior and radiation sensitivity is explored through MLN4924 intervention in vitro and in vivo. We construct a prognosis prediction model based on NAE1 using machine learning methods and validate the efficacy of NAE1 and the model in clinical cohorts. Results: NPC patients with high NAE1 expression have better prognosis and higher expression in the radiotherapy-sensitive group. Inhibiting NAE1 with MLN4924 causes cell cycle arrest in NPC cells, preventing them from entering the G2/M phase, thereby inhibiting proliferation but not affecting migration and metastasis. However, in vitro and in vivo experiments demonstrate that inhibiting NAE1 with MLN4924 leads to increased resistance of NPC to radiation. Conclusions: Targeting NAE1 for NPC treatment may have dual effects, inhibiting NPC proliferation while also increasing radiation resistance.

Abstract High-quality biofilm carriers are crucial for the formation of biofilm, but problems such as slow biofilm growth on the carrier surface have been troubling a large number of researchers. The addition of a carrier changes the flow state in the reactor, which in turn affects the microbial attachment and the quantity of microorganisms. Also, aerobic microorganisms need to use dissolved oxygen in the water to remove water pollutants. In this paper, a novel recirculating flow carrier with a hollow cylinder structure is proposed, with a certain number of hollow inverted circular plates placed at equal distances inside. In this paper, the hydraulic residence time, aeration volume, and the spacing of the inflow plates of the recirculating flow biofilm carrier, which are three important factors affecting the hydraulic characteristics of the reactor, are first investigated. At the same time, it was compared with the common combined carrier to find the optimal operating conditions for the hydraulic characteristics. Secondly, a reactor start-up study was carried out to confirm that the new recirculating flow biofilm carrier could accelerate the biofilm growth by changing the hydraulic characteristics. The results showed that under the same conditions, the hydraulic properties of the reactor were better with the addition of the recirculating flow carrier, with an effective volume ratio of 98% and a significant reduction in short flows and dead zones. The stabilized removal of COD, NH3-N, and TN in the reactor with the addition of the recirculating flow carrier reached about 94%, 99%, and 91% respectively, at the beginning of the 15th day, which effectively proved the feasibility of the recirculating flow carrier.

From the perspective of artificial intelligence evaluation, the need to discover and explain the potential shortness of the evaluated intelligent algorithms/systems as well as the need to evaluate the intelligence level of such testees are of equal importance. In this paper, we propose a possible solution to these challenges: Explainable Evaluation for visual intelligence. Specifically, we focus on the problem setting where the internal mechanisms of AI algorithms are sophisticated, heterogeneous or unreachable. In this case, a latent attribute dictionary learning method with constrained by mapping consistency is proposed to explain the performance variation patterns of visual perception intelligence under different test samples. By jointly iteratively solving the learning of latent concept representation for test samples and the regression of latent concept-generalization performance, the mapping relationship between deep representation, semantic attribute annotation, and generalization performance of test samples is established to predict the degree of influence of semantic attributes on visual perception generalization performance. The optimal solution of proposed method could be reached via an alternating optimization process. Through quantitative experiments, we find that global mapping consistency constraints can make the learned latent concept representation strictly consistent with deep representation, thereby improving the accuracy of semantic attribute-perception performance correlation calculation.

Green storage plays a key role in modern logistics and is committed to minimizing the environmental impact. To promote the transformation of traditional storage to green storage, research on the capacity allocation of wind-solar-storage microgrids for green storage is proposed. Firstly, this paper proposes a microgrid capacity configuration model, and secondly takes the shortest payback period as the objective function, and uses the improved sparrow search algorithm (ISSA) for optimization. Firstly, the Logistic-Tent compound chaotic mapping method is added to the population initialization of the sparrow search algorithm (SSA). Secondly, the adaptive t-distribution mutation is used to improve the discoverer, and the overall optimization ability of the algorithm is improved. Finally, the hybrid decreasing strategy is adopted in the process of vigilance position update. The ISSA can improve the search efficiency of the algorithm, avoid premature convergence and enhance the robustness of the algorithm, which is helpful to better apply to the optimal configuration of wind-solar-storage microgrid capacity in green storage. By analyzing the optimal capacity allocation results of two typical days, the system can better adapt to the dynamic storage requirements and improve the flexibility and sustainability of the supply chain.

A method was developed for the determination of amantadine sulfonamides, quinolones and chloramphenicols in cooked meat. The samples were extracted with acetonitrile-water mixtures (V:V=8:2), purified with PRiME HLB solid phase extraction column. 36 veterinary drugs were determined by ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) with an electrospray ion source (ESI±) in positive and negative ion switching scanning mode. The results showed that the 36 veterinary drugs showed a good linear relationship in the concentration range of 0.5~20.0 ng/mL (R2>0.999). The limits of detection were 0.10 μg/kg and the limits of quantitation were 0.33 μg/kg. The average recovery was 70.8%~106.9% in the range of high, medium and low concentrations (0.5, 1.0 and 5.0 μg/kg). The relative standard deviations (RSD, n=6) was 0.2%~6.2%. The method is simple, sensitive and accurate, and could be used for the determination of 36 veterinary drug residues in cooked meat.

In this paper, selenium polysaccharide (sKGP) extract was obtained by extraction, separation and purification of selenium-enriched creeson grape, then its physicalchemical properties, structural characteristics and antioxidant activity were studied. The selenium polysaccharide was extracted by ultrasonic-assisted enzymatic, separated and purified by diethylaminoethyl cross-linked dextran gel chromatography column. Its active functional groups and structures were identified by UV, LC-MS, fourier IR and SEM, and the antioxidant activity indexes were also analyzed. The results showed that the content of selenium in sKGP-3 was (0.642±0.038) mg/kg, and the purified sKGP-3 contained almost no nucleic acid and protein, the sKGP-3 was composed of rhamnose, xylose, arabinose, mannose, glucose and fructose, the molar ratio was 0.172:0.743:0.718:0.082:4.540:0.093. The sKGP-3 had the structure of Se=O, Se-H and typical characteristic peaks of polysaccharides, and the shape of the sKGP-3 was mainly lamellar and filamentous, belonging to amorphous state. The evaluation of antioxidant activity in vitro showed that the sKGP had higher antioxidant activity than the polysaccharide (KGP) of the same concentration before purification, the total antioxidant capacity of sKGP was 8.499 μmol/mL, and the scavenging capacity of DPPH, hydroxyl radicals and ABTS+ radicals were 69.7%, 85.8% and 92.5%, respectively.

The subtransient reactance Xd′′and transient reactance Xd′of large-capacity alternators are crucial parameters to determine the fault current magnitude and design the protection system for the machine. The traditional calculation method of Xd′′and Xd′will bring errors in the graphical decomposition process which depends on the subjective judgement of different individuals, and the analytical method is also inherently imprecise. In this paper, the superconductor simulation method (SSM) incorporated with field-circuit coupled finite element model is presented. The SSM has high accuracy and is capable of considering the saturation effect, skin effect and arbitrary structure of the machine. The unsaturated and saturated values of Xd′′and Xd′of a pumped storage machine are calculated with the proposed SSM. The method of establishing the field-circuit coupled model, governing equations and the superconductor simulation method are introduced in detail. The short-circuit test is conducted on the real machine and the calculated results of both Xd′′and Xd′are verified by the measured values.

Abstract Purpose To evaluate the clinical utility of surgical reconstruction of finger pulp defects using a plantar flap derived from the toes, with vascular anastomosis of the toe-finger artery and the plantar-palmar vein of the finger. Methods Between April 2018 and November 2020, 29 patients with finger pulp defects underwent treatment via the transplantation of pulp tissue from the second toe, with the plantar vein of the toe and the palmar vein of the finger being anastomosed during this procedure. In addition, an anastomosis of the toe and finger artery and nerve was conducted, with a flap size of 1.0 cm * 0.8 cm–2.3 cm * 4.0 cm being used for such repair. Donor tissue sites were closed without introducing deformities or other complications. Results In all patients in the present study, flap tissues survived and did not exhibit evidence of vascular crisis over a mean 16.8-month follow-up period (range 8–24 months). After successful skin flap grafting, they exhibited good elasticity and a soft texture. At three months post-surgery, some patients reported partial recovery of touch sensation in the transplanted tissue, while pain recovery was evident in some patients at 4–6 months post-surgery. No deformities or other complications were observed at the donor site, and the ability of patients to walk normally was not impaired. Conclusion The anastomosis of toe plantar flaps with the palmar vein can facilitate the repair of finger pulp injuries without the need to dissect the dorsal vein of the toe, allowing for the suturing of donor tissue sites without causing any deformities or other complications. This surgical approach can easily be conducted with satisfactory clinical outcomes.

High-speed railway catenaries are vital components in railway traction power supply systems. To ensure stable contact between the pantograph and the catenary, droppers are positioned between the messenger wire and contact line. The failure of one or more droppers will affect the power supply of the catenary and the operation of the railway. In this paper, we modify the You Only Look Once version five (YOLOv5) model in several ways and propose a method for improving the identification of dropper status and the detection of small defects. Firstly, to focus on small target features, the selective kernel attention module is added to the backbone. Secondly, the feature graphs of different scales extracted from the backbone network are fed into the bidirectional feature pyramid network for multiscale feature fusion. Thirdly, the YOLO head is replaced by a decoupled head to improve the convergence speed and detection accuracy of the model. The experimental results show that the proposed model achieves a mean average precision of 92.9% on the dropper dataset, an increase of 3.8% over the results using YOLOv5s. The detection accuracy of small dropper defects reaches 79.2%, representing an increase of 10.8% compared with YOLOv5s and demonstrating that our model is better at detecting small defects.

Background This study aimed to explore the prognostic value of particulate matter with a diameter of ≤2.5 μm (PM2.5)-related microRNA-206 combined with interleukin (IL)-4, IL-13 and interferon-γ (INF-γ) in asthma induced pulmonary arterial hypertension (PAH). Methods Fifty SPF BALB/c mice were divided into 5 groups: control group, asthma + PAH group, low-toxic asthma + PAH group, moderately-exposed asthma + PAH group, highly-exposed asthma + PAH group. Differences of microRNA-206, IL-4, IL-13, and INF-γ expression in lung tissue and plasma were detected. A total of 98 patients with asthma induced PAH and 98 healthy persons were collected. Patients were followed up for 12 months. Results Based on microarray analyses, we found that microRNA-206 may be involved in asthma induced PAH stimulated by PM2.5. Compared with healthy people, plasma microRNA-206, IL-4, IL-13, and INF-γ levels in asthma induced PAH patients were significantly higher (P 0.473 was significantly reduced (χ2 = 17.377, P

Purpose – The operating wagon records were produced from distinct railway information systems, which resulted in the wagon routing record with the same oriental destination (OD) was different. This phenomenon has brought considerable difficulties to the railway wagon flow forecast. Some were because of poor data quality, which misled the actual prediction, while others were because of the existence of another actual wagon routings. This paper aims at finding all the wagon routing locus patterns from the history records, and thus puts forward an intelligent recognition method for the actual routing locus pattern of railway wagon flow based on SST algorithm. Design/methodology/approach – Based on the big data of railway wagon flow records, the routing metadata model is constructed, and the historical data and real-time data are fused to improve the reliability of the path forecast results in the work of railway wagon flow forecast. Based on the division of spatial characteristics and the reduction of dimension in the distributary station, the improved Simhash algorithm is used to calculate the routing fingerprint. Combined with Squared Error Adjacency Matrix Clustering algorithm and Tarjan algorithm, the fingerprint similarity is calculated, the spatial characteristics are clustering and identified, the routing locus mode is formed and then the intelligent recognition of the actual wagon flow routing locus is realized. Findings – This paper puts forward a more realistic method of railway wagon routing pattern recognition algorithm. The problem of traditional railway wagon routing planning is converted into the routing locus pattern recognition problem, and the wagon routing pattern of all OD streams is excavated from the historical data results. The analysis is carried out from three aspects: routing metadata, routing locus fingerprint and routing locus pattern. Then, the intelligent recognition SST-based algorithm of railway wagon routing locus pattern is proposed, which combines the history data and instant data to improve the reliability of the wagon routing selection result. Finally, railway wagon routing locus could be found out accurately, and the case study tests the validity of the algorithm. Practical implications – Before the forecasting work of railway wagon flow, it needs to know how many kinds of wagon routing locus exist in a certain OD. Mining all the OD routing locus patterns from the railway wagon operating records is helpful to forecast the future routing combined with the wagon characteristics. The work of this paper is the basis of the railway wagon routing forecast. Originality/value – As the basis of the railway wagon routing forecast, this research not only improves the accuracy and efficiency for the railway wagon routing forecast but also provides the further support of decision-making for the railway freight transportation organization.

It is inevitable that defects happen to key components of the long-running high-speed trains. Thus as an effective inspection approach for defects, image detection becomes significantly important for operation and maintenance in the railway industry. However, a massive number of images collected by inspection devices challenge traditional methods based on manual effort. To address this issue, this paper proposed an automatic detection method, termed as multi-stage pipeline for defect detection (MPDD). MPDD includes two stages, component detection stage improves RPN anchor mechanism and way of feature fusion to promote detection performance, defect classification stage combines super-resolution strategy with CNN to improve defect classification performance. Experiments on high-speed train defect dataset shown that MPDD can reach the highest mAP of 0.792. The mAP on NEU surface defect database reached to 0.765 at the speed of 203ms per image.