Top Read Articles

    Published in last 1 year |  In last 2 years |  In last 3 years |  All
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Recognition of classroom learning behaviors based on the fusion of human pose estimation and object detection
    Zejie WANG, Chaomin SHEN, Chun ZHAO, Xinmei LIU, Jie CHEN
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 55-66.   DOI: 10.3969/j.issn.1000-5641.2022.02.007
    Abstract381)   HTML57)    PDF (1026KB)(397)      

    As a result of ongoing advances in artificial intelligence technology, the potential for learning analysis in teaching evaluation and educational data mining is gradually being recognized. In classrooms, artificial intelligence technology can help to enable automated student behavior analysis, so that teachers can effectively and intuitively grasp students’ learning behavior engagement; the technology, moreover, can provide data to support subsequent improvements in learning design and implementation of teaching interventions. The main scope of the research is as follows: Construct a classroom student behavior dataset that provides a basis for subsequent research; Propose a behavior detection method and a set of feasible, high-precision behavior recognition models. Based on the global features of the human posture extracted from the Openpose algorithm and the local features of the interactive objects extracted by the YOLO v3 algorithm, student behavior can be identified and analyzed to help improve recognition accuracy; Improve the model structure, compress and optimize the model, and reduce the consumption of computing power and time. Four behaviors closely related to the state of learning engagement: listening, turning sideways, bowing, and raising hands are recognized. The accuracy of the detection and recognition method on the verification set achieves 95.45%. The recognition speed and accuracy of common behaviors, such as playing with mobile phones and writing, are greatly improved compared to the original model.

    Table and Figures | Reference | Related Articles | Metrics
    Survey of few-shot instance segmentation methods
    Xueming ZHOU, Dingjiang HUANG
    Journal of East China Normal University(Natural Science)    2022, 2022 (5): 136-146.   DOI: 10.3969/j.issn.1000-5641.2022.05.012
    Abstract345)   HTML11)    PDF (968KB)(57)      

    Instance segmentation is an important task in computer vision. In recent years, the development of meta- and few-shot learning has promoted the combination of computer vision learning tasks, which has overcome the bottleneck of detection and classification with regard to objects that are difficult to manually label and those with high labeling costs. Although great progress has been made with few-shot semantic segmentation and object detection, instance segmentation based on few-shot learning has not become a research hotspot until very recently. Beginning with an overview of few-shot instance segmentation, existing approaches are divided into categories of anchor-based and anchor-free algorithms. The architectures and primary technologies behind those approaches are respectively discussed, and common datasets and evaluation indices are described. Additionally, advantages and disadvantages of algorithm performance are analyzed, and future development directions and challenges are presented.

    Table and Figures | Reference | Related Articles | Metrics
    Research progress of microplastics and attached organisms in marine environment
    Daoji LI, Xuri DONG
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 1-7.   DOI: 10.3969/j.issn.1000-5641.2022.03.001
    Abstract334)   HTML885)    PDF (475KB)(208)      

    In recent years, white pollution caused by waste plastics has attracted widespread attention. Microplastics, which are smaller than 5 mm, are widely distributed in the marine environment. The organisms attached to microplastic surfaces include potential pathogenic bacteria that are harmful to marine life and even human health, as well as plastic-degrading bacteria that can reduce their pollution. Microplastics are difficult to degrade, so they can exist in the aquatic environment for a long time, and the microorganisms attached to their surface can also live stably. In addition, microplastics may pass through the food chain to organisms at higher nutritional levels, and may be eaten by fish and affect fish growth. This paper reviews the distribution of microplastics in the ocean and the potential effects of harmful substances contained or attached to the microplastic surface on organisms. The ecological effects of pathogenic microorganisms attached to the surface of microplastics and plastic decomposition microorganisms, as well as the potential of microplastic transmission to high nutritional levels through the food chain were discussed. The ecological risk of microplastic distribution and surface-attached organisms was analyzed. Furtherly, it is still necessary to understand the impact of plastic waste and microplastics on the marine ecosystem, so as to fully understand the ecological effects of marine microplastics and their attachments, and provide a scientific basis for marine plastic pollution control.

    Reference | Related Articles | Metrics
    Optically mediated entanglement between Bose-Einstein condensates
    Shuai GAO, Matthew PREST, Ebubechukwu O. ILO-OKEKE, Manikandan KONDAPPAN, Juan E. ARISTIZABAL-ZULUAGA, Valentin IVANNIKOV, Tim BYRNES
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 93-105.   DOI: 10.3969/j.issn.1000-5641.2022.02.011
    Abstract292)   HTML44)    PDF (1315KB)(102)      

    This paper explores a method for generating optically mediated entanglement between Bose-Einstein condensates (BECs). Using a quantum nondemolition Hamiltonian with BECs placed in a Mach-Zehnder configuration, it is shown that entangled states can be induced by performing measurement on light. In particular, the effects of the entangled state in the presence of decoherence were analyzed. The behavior of the entangled state was found to be sensitive to the atom-light interaction time. The entangled state is relatively stable when the dimensionless interaction time $ \tau \lesssim \frac{1}{\sqrt{N}} $ and relatively fragile when the time is greater.

    Table and Figures | Reference | Related Articles | Metrics
    Infinite dimensional 3-Pre-Lie algebras
    Ruipu BAI, Shan LIU
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 1-8.   DOI: 10.3969/j.issn.1000-5641.2022.02.001
    Abstract251)   HTML298)    PDF (625KB)(123)      

    Constructing 3-Pre-Lie algebras has always been a difficult problem; until now, there have been very few examples of 3-Pre-Lie algebras. In this paper, we use homogenous Rota-Baxter operators of weight zero on the infinite dimensional 3-Lie algebra $A_{\omega}=\langle L_m | m\in {\mathbb{Z}}\rangle$ to construct 3-Pre-Lie algebras $B_k,~0\leqslant k\leqslant 4$ , and we subsequently discuss the structure. It is shown that $B_2$ and $B_4$ are non-isomorphic simple 3-Pre-Lie algebras, $B_1$ is an indecomposable 3-Pre-Lie algebra with infinitely many one-dimensional ideals, and $B_3$ is an indecomposable 3-Pre-Lie algebra with finitely many ideals.

    Reference | Related Articles | Metrics
    Design and implementation of automatic correction for college mathematics assignments
    Qinmin YANG, Zhisong JIANG
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 76-83.   DOI: 10.3969/j.issn.1000-5641.2022.02.009
    Abstract197)   HTML41)    PDF (821KB)(81)      

    By combining information extraction technology, data matching technology, and one-time manual processing, mathematical subjective questions can be transformed into tree-shaped multiple-choice questions. In this study, an automatic correction system for college mathematics assignments was developed by combining modern information and network technology; the system was subsequently trialed in the teaching of entry-level college mathematics courses. The proposed system solves bottlenecks related to automatic grading of subjective mathematics questions, including multiple-choice questions, fill-in-the-blank questions, judgment questions, quiz questions, calculation questions, and proof questions. The system can correct routine exercises for mathematics courses of primary schools, middle schools, and universities so as to achieve more efficient completion. The idea, furthermore, can be applied to various aspects of mathematics teaching, such as previews before class, classroom exercises, reviews after class, preparations for examinations, online examinations, etc. The electronic data collected in the process of automatic correction can subsequently be used for data analysis, teaching guidance, teaching research, and the construction of educational informatization.

    Table and Figures | Reference | Related Articles | Metrics
    Diversity of soil microbes in a vertical forest
    Xianya XU, Stefano BOERI, Yibo XU
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 8-16.   DOI: 10.3969/j.issn.1000-5641.2022.03.002
    Abstract184)   HTML852)    PDF (1182KB)(82)      

    In this study, we analyzed the diversity of microbes across 45 collected soil samples using meta-barcoding. The analysis showed similar α diversity of soil microbes in a vertical forest (VF) building and the surrounding green area, but a high level of differentiation in the microbe community composition and β diversity between these two types of habitats. The results indicated that a VF can accommodate a large number of microbe species and provided evidence for the contribution of VFs to the conservation of urban biodiversity.

    Table and Figures | Reference | Related Articles | Metrics
    The viscosity solution of the discounted Hamilton-Jacobi equation in non-compact space
    Suting CHEN, Xia LI
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 9-15.   DOI: 10.3969/j.issn.1000-5641.2022.02.002
    Abstract182)   HTML47)    PDF (550KB)(78)      

    The discounted Hamilton-Jacobi equation (H-J equation) is a special form of the contact Hamilton-Jacobi equation; hence, study of the discounted H-J equation is important. In this article, we first study an expression of the viscosity solution $u_{\lambda}(x,t)$ for the discounted H-J equation in non-compact space. Then, we explore the convergence of the viscosity solution $u_{\lambda}(x,t)$ for a specific discounted H-J equation with $\lambda >0$ in non-compact space for the initial value in different cases.

    Reference | Related Articles | Metrics
    Modeling and simulation technology of roads for a battlefield environment
    Shucheng LU, Yan GAO, Changbo WANG
    Journal of East China Normal University(Natural Science)    2022, 2022 (4): 79-94.   DOI: 10.3969/j.issn.1000-5641.2022.04.008
    Abstract182)      PDF (3909KB)(57)      

    Battlefield environments are combat spaces that contain geographic elements such as terrain and roads. Road modeling and simulation is an important part of battlefield simulation and plays a key role in complex combat decision-making. Traditional road modeling is unable to handle the complex terrain conditions present in the field; hence, this paper proposes road modeling and simulation method for field environments. In particular, in order to support road modeling and simulation of complex terrain environments, road construction designs oriented to typical battlefield environments are proposed. This method divides the road network into different sub-models according to their characteristics and models them separately, improving the demand for realism in battlefield simulation. Then, the proposed method uses OpenStreetMap geographic information data to drive road network construction. The model offers real-time, high accuracy road information content and complete classification that can meet the needs of military operations and modeling simulations for typical battlefield environments. Secondly, using terrain elevation data, road construction rules, and other auxiliary information, the road height is adjusted to adapt to the complex terrain conditions of the battlefield and possible multi-level road network structures. Lastly, the introduction of a $ {G}^{2} $ continuous Hermite interpolation spline can flexibly represent the center line of the road and improves the reusability of the road model through grid deformation. Experiments show that the proposed simulation method can more reliably restore the real details of a road network to effectively fit complex terrain and improve the reusability of road models. Finally, it provides a feasible analysis angle and modeling method for researching geographic elements in battlefield environments.

    Reference | Related Articles | Metrics
    Influence of habitat fragments on the richness of remnant natural forest species
    Hong LIANG, Liangjun DA, Kun SONG, Chunling ZHOU, Zhiwen GAO, Kai WANG
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 17-26.   DOI: 10.3969/j.issn.1000-5641.2022.03.003
    Abstract181)   HTML849)    PDF (862KB)(90)      

    Based on the classical theory of island biogeography, thirteen remnant forest patches of fragmented urban habitats were chosen as islands, and the Lao Mountain as the mainland. Then, canonical correspondence analysis (CCA) was used to analyze the relationship between species composition and four factors — namely, patch area, distance to species pool, human disturbance intensity, and isolation degree. The lengths of vectors in CCA biplots were used as weights for each influencing factor to calculate the overall sum, named the isolated island index (III); an isolated island index evaluation system of forest patches in fragmented urban habitats was subsequently designed. The results indicated that the relationship between III and richness was significantly negative for native species, but not significant for alien species richness. There was a significant relationship found for bird dispersal and wind dispersal species. The linear function was determined to be the best simulating model.

    Table and Figures | Reference | Related Articles | Metrics
    Analysis of changes in water quality at the Qingcaosha Reservoir water intake over a period of 10 years
    Yiping ZHU
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 50-60.   DOI: 10.3969/j.issn.1000-5641.2022.03.006
    Abstract177)   HTML61)    PDF (2362KB)(39)      

    Based on daily monitoring data of the Qingcaosha Reservoir water intake from 2010 to 2019, this paper analyzes the interannual and seasonal variation trends in the main physical and chemical water quality indicators, and explores the correlation between these indicators. The results show that: ① The dissolved oxygen at the intake of Qingcaosha Reservoir was consistently high, and the pH value was slightly alkaline. ② The concentration of ammonia nitrogen was low at the intake of Qingcaosha Reservoir, the concentration of nitrate-nitrogen was between 1.2 and 2.0 mg/L, the concentration of total phosphorus was between 0.1 and 0.2 mg/L, and the permanganate index was between 2.0 and 4.0 mg/L. These indexes, furthermore, all showed a downward trend after 2015, suggesting that the quality of incoming water subsequently improved. ③ Dissolved oxygen concentration, water temperature, and pH exhibited obvious seasonal variations, while total hardness, permanent hardness, conductivity, and chloride showed consistent variations under the influence of seawater intrusion. There was no significant seasonal variation in the other indicators. ④ The concentration of total phosphorus and the permanganate index increased with turbidity. The concentration of total phosphorus and nitrate nitrogen, moreover, decreased with an increase in water discharge at Datong.

    Table and Figures | Reference | Related Articles | Metrics
    Inversion methods for typhoon wind fields using Sentinel-1
    Peng YU, Xiaojing ZHONG, Xupu GENG
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 125-136.   DOI: 10.3969/j.issn.1000-5641.2022.03.013
    Abstract174)   HTML66)    PDF (23345KB)(46)      

    Under high-speed wind conditions, cross-polarization synthetic aperture radar (SAR) is not affected by signal saturation. Hence, SAR can be used to observe expansive, high-speed wind fields under all-weather, day- and night-time conditions and offers great potential for monitoring typhoons. Sentinel-1, which was launched by the European Space Agency (ESA), is one of the few available SAR satellites in orbit at present that can provide cross-polarization data. Based on Sentinel-1 cross-polarization data, seven different cross-polarization models, including the C-band cross polarization ocean model (C-2PO), C-band cross-polarization coupled-parameters ocean model (C-3PO), and quad-polarization stripmap cross-polarization model (QPS-CP), developed from 2011 to 2021 were used to estimate the typhoon wind fields of Higos and Molave. A denoising method was applied to remove the noise from extra wide (EW) mode SAR images. The results show that the denoising method can effectively reduce the noise and improve the retrieved wind fields. The C-3PO model performs well in monitoring high-speed winds, but does not obtain reliable results for low- to moderate-speed winds compared with the Sentinel-1 Level-2 Ocean (OCN) product. By merging results from the cross-polarization model and the OCN wind product, the combined wind field can effectively reproduce the inner high-speed winds and outer relative low-speed winds. This study is of significant value for forecasting, data assimilation, and research of typhoon disasters.

    Table and Figures | Reference | Related Articles | Metrics
    Calculation of particle motion trajectories in optical force fields
    Hailei XIE, Junjie DU
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 106-113.   DOI: 10.3969/j.issn.1000-5641.2022.02.012
    Abstract173)   HTML35)    PDF (893KB)(85)      

    In this paper, the motion trajectory of micro-nanoparticles is calculated based on the Euler-Richardson algorithm after the optical force exerted on the particles is determined using Mie scattering theory. The Euler-Richardson algorithm has better calculation accuracy and faster convergence speed than the Euler algorithm and the Euler-Kromer algorithm, and thus is an appropriate approach to describe the trajectory of particles. Hence, the motion trajectory of a nanoparticle in a periodic conservative optical force field is calculated based on the Euler-Kromer algorithm; the results confirm consistency with the physical analysis, further verifying the effectiveness and stability of the approach. The calculation method shown in this paper provides a high-efficiency approach to study optical trapping, transport, sorting of colloidal particles, and biological macromolecules as well as the cooling of macroscopic particles in optical micro-manipulation.

    Table and Figures | Reference | Related Articles | Metrics
    Research on user behavior portrait and subject mining in the express logistics field during Coronavirus epidemic
    Jiling LI, Baolin LI, Songru YAN
    Journal of East China Normal University(Natural Science)    2022, 2022 (5): 100-114.   DOI: 10.3969/j.issn.1000-5641.2022.05.009
    Abstract173)   HTML12)    PDF (1321KB)(83)      

    Based on logistics-field blog post data from Weibo from November 2019 to May 2022, the user behaviors of express logistics services in the context of the Coronavirus epidemic are profiled. Using grounded theory and abstract clustering methods, five user behaviors and 22 subject contents are abstracted, and the corresponding user profile is generated. This paper further discusses the subject contents, the subject evolution, and the analysis of group differences. The results show that user satisfaction with logistics services was similar, and the dissatisfaction was diversified with obvious escalation. Variables of transportation efficiency and logistics guarantee were the main factors affecting the evaluation, and the development of the epidemic affected the concerns and attitudes of the subject contents, which had obvious group differences at different degrees.

    Table and Figures | Reference | Related Articles | Metrics
    Temperature sensitivity and controlling factors of nitrogen fixation processes in sediments of the Yangtze River Estuary
    Zhiyong YOU, Bolin LIU, Cheng LIU, Dengzhou GAO
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 101-108.   DOI: 10.3969/j.issn.1000-5641.2022.03.011
    Abstract172)   HTML75)    PDF (790KB)(30)      

    To understand the stability of the estuarine ecosystem and nitrogen balance in the context of global climate change, it is important to investigate the temperature sensitivity of the microbial nitrogen fixation process. Until now, there have been few studies in the literature on the response of the nitrogen fixation process to temperature changes and its influencing factors. We selected six sampling sites around the Yangtze River Estuary (including four sites inside and two sites outside the Yangtze River Estuary) for the scope of the study; in particular, we explored the temperature sensitivity and influencing factors of the nitrogen fixation process on sediments of the Yangtze River Estuary using slurry incubation experiments and the 15N2 isotope tracer technique. The results showed that the in-situ temperature nitrogen fixation rate in the sediments of the Yangtze River Estuary ranged from 0.72 to 2.85 nmol·g–1·h–1. At 5 ~ 10 ℃ and20 ~ 30 ℃, the nitrogen fixation rate was inhibited by an increase in temperature. However, in the range of 10 ~ 20 ℃, the nitrogen fixation rate was significantly promoted with an increase in temperature. The sensitivity of the nitrogen fixation rate to temperature is relatively consistent, although the physical and chemical properties of the sediments vary significantly. Correlation analysis showed that the contents of sulfide, ferrous iron, nitrate, and total organic carbon were the main environmental factors affecting nitrogen fixation.

    Table and Figures | Reference | Related Articles | Metrics
    First-principles calculations investigations of two-dimensional transition metal phosphide MnTn+1(M = V, Cr; T = P, As, and Sb) slices
    Yaqiong ZHANG, Wenhui XIE
    Journal of East China Normal University(Natural Science)    2022, 2022 (2): 84-92.   DOI: 10.3969/j.issn.1000-5641.2022.02.010
    Abstract171)   HTML36)    PDF (1827KB)(60)      

    In this paper, the atomic structure, stability, electronic structure, and magnetism of two-dimensional transition metal phosphide MnTn+1 (M = V, Cr; T = P, As, and Sb) slices were systematically studied using the first-principles calculations based on density functional theory. By calculating the formation energy and phonon spectrum, it was determined that only V4As5, Cr2P3, Cr3P4, Cr4P5, Cr2As3, and Cr3As4 are stable two-dimensional magnetic multilayers. The results show that these stable two-dimensional magnetic materials are antiferromagnetic metals. In addition, the electronic structure and the magnetic coupling mechanism of these materials were further analyzed.

    Table and Figures | Reference | Related Articles | Metrics
    New form of the alternating direction iteration scheme for real positive definite linear systems
    Daosheng ZHENG
    Journal of East China Normal University(Natural Science)    2022, 2022 (4): 1-12.   DOI: 10.3969/j.issn.1000-5641.2022.04.001
    Abstract169)   HTML373)    PDF (789KB)(61)      

    Alternating direction iteration (ADI) scheme is an effective method for solving real positive definite linear systems; in many cases, however, the method requires that all the direction matrices involved are multiplication exchangeable, which severely limits the scope of application. In this paper, new revised alternating direction iteration (RADI) schemes are proposed, that do not stipulate the multiplication exchangeable requirement, thereby expanding the application scope. In parallel, measures to improve the efficiency of RADI schemes are also discussed.

    Reference | Related Articles | Metrics
    Comparison of the fixation effect of formaldehyde and glutaraldehyde on the Ludox–QPS method
    Miaoxun WANG, Yuan XU
    Journal of East China Normal University(Natural Science)    2022, 2022 (3): 82-89.   DOI: 10.3969/j.issn.1000-5641.2022.03.009
    Abstract168)   HTML66)    PDF (4170KB)(59)      

    Density gradient centrifugation in combination with quantitative protargol stain (Ludox–QPS) is an important research method for the separation, extraction, and staining of microbenthos. Glutaraldehyde solution is the most widely used fixative for this method; however, it is a hazardous chemical and transportation by civil aviation is forbidden. Hence, biogeography studies using glutaraldehyde solution on a large scale are greatly limited. In contrast, samples fixed with formaldehyde solution can be transported by air consignment under certain conditions. To test whether the results of a study on microbenthic communities fixed with formaldehyde and glutaraldehyde are comparable, we collected microbenthic samples from three natural habitats (bare mudflat, Scirpus mariqueter meadow, sandy beach) and fixed them with either formaldehyde (final concentration is 1%) or glutaraldehyde (final concentration is 2%). Then, we tested the significance of differences in species composition, diversity, and biomass of the communities using an ANOSIM two-way crossed analysis (fixative type and habitat type). The results of this study demonstrate that although significant differences exist in the community structures among the three habitats, the species composition, diversity indices (number of species, number of individuals, evenness, Shannon-Wiener diversity index), and biomass of the microbenthic communities between the two fixative types have no significant differences. NMDS analysis also shows similar results (stress = 0.11). In conclusion, the data indicates that the microbenthic communities fixed by the two fixatives are comparable. Therefore, formaldehyde is also suitable for ecological research of microbenthos as a fixative.

    Table and Figures | Reference | Related Articles | Metrics
    Generalized number operators defined in the space of a discrete time normal martingale functional
    Yulan ZHOU, Huafang KONG, Xiuqiang CHENG, Rui XUE, Jia CHEN
    Journal of East China Normal University(Natural Science)    2022, 2022 (4): 13-25.   DOI: 10.3969/j.issn.1000-5641.2022.04.002
    Abstract168)   HTML373)    PDF (826KB)(63)      

    A family of linear operators $\{N_{h};h\in\mathcal{P_{+}}(\mathbb{N})\}$ in $L^{2}(M)$ are defined. Firstly, we prove that $N_{h}$ is a positive, densely defined, self-adjoint closed linear operator. In general, $N_{h}$ is not bounded, hence, we explore the sufficient and necessary conditions such that $N_{h}$ is bounded. Secondly, we consider the dependence of $N_{h}$ on $h$ : $N_{h}$ is strictly increasing with respect to $h$ , and the operator-valued mapping $N_{h}$ is an isometry from $l^{1}_{+}(\mathbb{N})$ to the subspace of bounded generalized number operators on $L^{2}(M)$ , where $l^{1}_{+}(\mathbb{N})$ is the space of the summable function on $\mathbb{N}$ . We consider the conditions such that $\{N_{h_{n}};n\geqslant1\}$ is strongly and uniformly convergent. If $\{h_{n};n\geqslant1\}$ is convergent monotonically to $h$ , the domain of $\{N_{h_{n}};n\geqslant1\}$ and $N_{h}$ have some interesting properties, we show, furthermore, that a convergent family of $\{N_{h_{n}};n\geqslant1\}$ can be obtained. We prove that $\{N_{h};h\in\mathcal{P_{+}}(\mathbb{N})\}$ is commutative observable on $\mathcal{S}_{0}(M)$ .

    Reference | Related Articles | Metrics
    Neural architecture search algorithms based on a recursive structure
    Jizhou LI, Xin LIN
    Journal of East China Normal University(Natural Science)    2022, 2022 (4): 31-42.   DOI: 10.3969/j.issn.1000-5641.2022.04.004
    Abstract168)   HTML38)    PDF (772KB)(85)      

    Neural architecture search algorithms aim to find more efficient neural network structures in a huge neural network structure space using computer heuristic search instead of manual search. Previous studies have addressed the problem of inefficient and time-consuming search for early neural network structures by introducing various constraints on the search space. While constraints on the search space can improve and stabilize the performance of the model, they ignore potentially efficient model structures. Hence, in this study, we constructed a recursive model search space that focuses more on the macroscopic structure of neural networks. We proposed a neural architecture search algorithm that explores this search space through a step-by-step incremental search approach. Experiments showed that the algorithm can efficiently perform neural architecture search tasks in complex search spaces, but still fell slightly short of the latest constrained search space-based neural architecture search algorithms.

    Table and Figures | Reference | Related Articles | Metrics