Recently, the hybrid convolutional neural network hidden Markov model (CNN-HMM) has been introduced for offline handwritten Chinese text recognition (HCTR) and has achieved state-of-the-art ...performance. However, modeling each of the large vocabulary of Chinese characters with a uniform and fixed number of hidden states requires high memory and computational costs and makes the tens of thousands of HMM state classes confusing. Another key issue of CNN-HMM for HCTR is the diversified writing style, which leads to model strain and a significant performance decline for specific writers. To address these issues, we propose a writer-aware CNN based on parsimonious HMM (WCNN-PHMM). First, PHMM is designed using a data-driven state-tying algorithm to greatly reduce the total number of HMM states, which not only yields a compact CNN by state sharing of the same or similar radicals among different Chinese characters but also improves the recognition accuracy due to the more accurate modeling of tied states and the lower confusion among them. Second, WCNN integrates each convolutional layer with one adaptive layer fed by a writer-dependent vector, namely, the writer code, to extract the irrelevant variability in writer information to improve recognition performance. The parameters of writer-adaptive layers are jointly optimized with other network parameters in the training stage, while a multiple-pass decoding strategy is adopted to learn the writer code and generate recognition results. Validated on the ICDAR 2013 competition of CASIA-HWDB database, the more compact WCNN-PHMM of a 7360-class vocabulary can achieve a relative character error rate (CER) reduction of 16.6% over the conventional CNN-HMM without considering language modeling. By adopting a powerful hybrid language model (N-gram language model and recurrent neural network language model), the CER of WCNN-PHMM is reduced to 3.17%. Moreover, the state-tying results of PHMM explicitly show the information sharing among similar characters and the confusion reduction of tied state classes. Finally, we visualize the learned writer codes and demonstrate the strong relationship with the writing styles of different writers. To the best of our knowledge, WCNN-PHMM yields the best results on the ICDAR 2013 competition set, demonstrating its power when enlarging the size of the character vocabulary.
•We propose a guideline to distill the architecture and knowledge of pre-trained standard CNNs simultaneously for fast compression and acceleration.•The effectiveness is first verified on offline ...HCTR. Compared with the baseline CNN, the corresponding compact network can reduce the computational cost by >10×and model size by >8×with negligible accuracy loss.•Furthermore, the proposed method is successfully used to reduce resource consumption of the mainstream backbone networks on CTW and MNIST.
The distillation technique helps transform cumbersome neural networks into compact networks so that models can be deployed on alternative hardware devices. The main advantage of distillation-based approaches include a simple training process, supported by most off-the-shelf deep learning software and no special hardware requirements. In this paper, we propose a guideline for distilling the architecture and knowledge of pretrained standard CNNs. The proposed algorithm is first verified on a large-scale task: offline handwritten Chinese text recognition (HCTR). Compared with the CNN in the state-of-the-art system, the reconstructed compact CNN can reduce the computational cost by >10×and the model size by >8×with negligible accuracy loss. Then, by conducting experiments on two additional classification task datasets: Chinese Text in the Wild (CTW) and MNIST, we demonstrate that the proposed approach can also be successfully applied on mainstream backbone networks.
We propose a projected shell model (PSM) for description of stellar weak-interaction rates between even-even and odd-odd nuclei with extended configuration space where up to six-quasiparticle (qp) ...configurations are included, and the stellar weak-interaction rates for eight rp-process waiting-point (WP) nuclei, 64Ge, 68Se, 72Kr, 76Sr, 80Zr, 84Mo, 88Ru and 92Pd, are calculated and analyzed for the first time within the model. Higher-order qp configurations are found to affect the underlying Gamow-Teller strength distributions and the corresponding stellar weak-interaction rates. Under rp-process environments with high temperatures and densities, on one hand, thermal population of excited states of parent nuclei tends to decrease the stellar β+ decay rates. On the other hand, the possibility of electron capture (EC) tends to provide increasing contribution to the rates with temperature and density. The effective half-lives of WP nuclei under the rp-process peak condition are predicted to be reduced as compared with the terrestrial case, especially for 64Ge and 68Se.
Writing style is an abstract attribute in handwritten text. It plays an important role in recognition systems and is not easy to define explicitly. Considering the effect of writing style, a writer ...adaptation method is proposed to transform a writer-independent recognizer toward a particular writer. This transformation has the potential to significantly increase accuracy. In this paper, under the deep learning framework, we propose a general fast writer adaptation solution. Specifically, without depending on other complex skills, a well designed style extractor network (SEN) trained by identification loss (IDL) is introduced to explicitly extract personalized writer information. The architecture of SEN consists of a stack of convolutional layers followed by a recurrent neural network with gated recurrent units to remove semantic context and retain writer information. Then, the outputs of the GRU are further integrated into a one-dimensional vector that is adopted to represent writing style. Finally, the extracted style information is fed into the writer-independent recognizer to achieve adaptation. Validated on offline handwritten text recognition tasks, the proposed fast sentence-level adaptation achieves remarkable improvements in Chinese and English text recognition tasks. Specifically, in the HETR task, a multi-information fusion network that is equipped with a hybrid attention mechanism and that integrates visual features, context features and writing style is proposed. In addition, under the same condition (only one writer-specific text line used as adaptation data), the proposed solution, without consuming extra time, can significantly outperform the previous multiple-pass decoding method. The code is available at https://github.com/Wukong90/Handwritten-Text-Recognition.
Cervical spondylotic myelopathy (CSM) is a common cause of disability with few treatments. Aberrant mitochondrial dynamics play a crucial role in the pathogenesis of various neurodegenerative ...diseases. Thus, regulation of mitochondrial dynamics may offer therapeutic benefit for the treatment of CSM. Muscone, the active ingredient of an odoriferous animal product, exhibits anti‐inflammatory and neuroprotective effects for which the underlying mechanisms remain obscure. We hypothesized that muscone might ameliorate inflammatory responses and neuronal damage by regulating mitochondrial dynamics. To this end, the effects of muscone on a rat model of chronic cervical cord compression, as well as activated BV2 cells and injured neurons, were assessed. The results showed that muscone intervention improved motor function compared with vehicle‐treated rats. Indeed, muscone attenuated pro‐inflammatory cytokine expression, neuronal‐apoptosis indicators in the lesion area, and activation of the nod‐like receptor family pyrin domain‐containing 3 inflammasome, nuclear transcription factor‐κB, and dynamin‐related protein 1 in Iba1‐ and βIII‐tubulin‐labeled cells. Compared with vehicle‐treated rats, compression sites of muscone‐treated animals exhibited elongated mitochondrial morphologies in individual cell types and reduced reactive oxygen species. In vitro results indicated that muscone suppressed microglial activation and neuronal damage by regulating related‐inflammatory or apoptotic molecules. Moreover, muscone inhibited dynamin‐related protein 1 activation in activated BV2 cells and injured neurons, whereby it rescued mitochondrial fragmentation and reactive oxygen species production, which regulate a wide range of inflammatory and apoptotic molecules. Our findings reveal that muscone attenuates neuroinflammation and neuronal damage in rats with chronic cervical cord compression by regulating mitochondrial fission events, suggesting its promise for CSM therapy.
Cervical spondylotic myelopathy (CSM) is a common cause of disability with few treatments. Muscone, the ingredient from an animal product, exhibits anti‐inflammatory and neuroprotective effects for which the underlying mechanisms remain obscure. In this study, we explored the effects of muscone on a classic CSM rat model, microglial and neuronal cell. Results indicated that muscone could attenuate both inflammatory responses and neuronal damage in a rat CSM model and in vitro. Such effects may be associated with its role on mitochondrial dynamics and downstream signaling, suggesting that mitochondrial dynamics and muscone may be new highlights for the treatment of CSM.
Structural interpenetration in metal–organic frameworks (MOFs) significantly impacts on their properties and functionalities. However, understanding the interpenetration on third-order nonlinear ...optics (NLO) of MOFs have not been reported to date. Herein, we report two 3D porphyrinic MOFs, a 2-fold interpenetrated Zn2(TPyP)(AC)2 (ZnTPyP-1) and a noninterpenetrated Zn3(TPyP)(H2O)2(C2O4)2 (ZnTPyP-2), constructed from 5,10,15,20-tetra(4-pyridyl)porphyrin (TPyP(H2)) and Zn(NO3)2 (AC = acetate, C2O4 = oxalate). ZnTPyP-1 achieves excellent optical limiting (OL) performance with a giant nonlinear absorption coefficient (3.61 × 106 cm/GW) and large third-order susceptibility (7.73 × 10–7 esu), which is much better than ZnTPyP-2 and other reported OL materials. The corresponding MOFs nanosheets are dispersed into a polydimethylsiloxane (PDMS) matrix to form highly transparent and flexible MOFs/PDMS glasses for practical OL application. In addition, the OL response optimized by adjusting the MOFs concentration in the PDMS matrix and the type of metalloporphyrin are discussed in the ZnTPyP-1 system. The theoretical calculation confirmed that the abundant π–π interaction from porphyrinic groups in the interpenetrated framework increased the electron delocalization/transfer and boosted the OL performance. This study opens a new avenue to enhance OL performance by the construction of interpenetrated structures and provides a new approach for the preparation of transparent and flexible MOF composites in nonlinear optical applications.
Summary
The objective of this paper was to investigate the complex of grape seed procyanidins (GSP) with zein hydrolysate (ZH). The interaction was determined using isothermal titration calorimetry ...(ITC) and fluorescence spectroscopy. The particle size, ζ‐potential, scanning electronic microscopy (SEM) and stability test of the ZH‐GSP complex were measured. The results of ITC, particle size and SEM suggested that there was a hydrogen bond‐dominated interaction between GSP and ZH, and the ZH‐GSP complex presented as a spherical shape with particle size of 234 nm, polydispersity index of 0.11 and ζ‐potential of −46.4 mV. The results of the fluorescence spectroscopy showed that the fluorescence quenching effect of GSP on ZH was GSP concentration‐dependent and the quenching process was mainly static quenching. The ZH‐GSP complex had better physical stability in aqueous solution, and their solution was more stable than GSP solution under illumination and high‐temperature treatment at 75°C.
The formation mechanism, morphology and stability of the ZH‐GSP complex were investigated.
Both carbon tax and cap-and-trade systems are widely applied to reduce emission. This article compares the clean innovation effects of carbon tax with cap-and-trade systems by a static optimal model. ...Firstly, both cap-and-trade system and carbon tax stimulates clean innovation and reduce emission. Secondly, cap-and-trade system is more efficient to reduce emission and to promote clean innovation than carbon tax. Finally, firms undertake a loss under carbon tax, while the effects of cap-and-trade system on firms' profits are uncertain, which depends on the carbon cap. In summary, this article supports cap-and-trade system to cope with global climate change, but the regulator should choose the suitable emission cap and carbon trading price to guarantee the efficiency of cap-and-trade system. So, different purposes match with different carbon emission tax policies.
•Both cap-and-trade system and carbon tax stimulate clean innovation.•The effects of cap-and-trade system on firms' profits are uncertain.•Cap-and-trade system has higher efficiency than carbon taxes on global climate change control.•This article further develops the innovative theory under both carbon tax and cap-and-trade systems.
Although the combination of halide perovskite photocatalysts and plasma ensures the effective conversion of CO2, there is still much room to improve its conversion ratio and energy efficiency. The ...traditional experimental trial-and-error method is extremely demanding for researchers in each experimental operation and result analysis, while the experiments greatly consume time and raw materials and require complex equipment. In this paper, for the first time, we modeled the process of CO2 conversion synergistically driven by dielectric barrier discharge (DBD) plasma and a Cs2TeCl6 photocatalyst via machine learning. K-fold cross-validation combined with the coefficient of determination (R2) was used to evaluate the regression algorithms, and the BPANN with the best performance was selected to establish a model for predicting the CO2 conversion ratio and energy efficiency. In order to make the predictions more accurate, genetic algorithms, particle swarm optimization and Bayesian optimization were applied to improve the hyperparameters of the neural network, and the GA-BPANN model achieved an R2 of 0.9713 and 0.9622 on the training and testing sets, respectively, while its practical application was also demonstrated. In addition, the effect of each process parameter on conversion efficiency was quantified by the Spearman correlation coefficients, which could provide insights into the roles of different process parameters in CO2 conversion. This work provides a new approach for boosting CO2 conversion, which could facilitate future experimental design and process optimization to promote carbon dioxide utilization.
PDF
