全自动开放式人机区分图灵测试(CAPTCHA)是基于人工智能领域开放性问题而设计的网络安全技术,CAPTCHA识别是该研究领域的重要分支.长短时记忆(Long Short Term Memory,LSTM)型递归神经网络(Recurrent Neural Network,RNN)已被成功应用于...全自动开放式人机区分图灵测试(CAPTCHA)是基于人工智能领域开放性问题而设计的网络安全技术,CAPTCHA识别是该研究领域的重要分支.长短时记忆(Long Short Term Memory,LSTM)型递归神经网络(Recurrent Neural Network,RNN)已被成功应用于CAPTCHA识别,LSTM型RNN实质上是一维RNN,而文本型CAPTCHA为二维图像.提出使用二维RNN对CAPTCHA进行识别.二维RNN能够很好的将特征提取同识别相结合,同时具有较好的上下文保持特性,从而更适合文本型CAPTCHA识别.同时为了进一步提高识别的可靠性,提出一种基于支持向量机(Support vector machine,SVM)的拒识策略,实验结果表明二维RNN较一维RNN能够获得更好的识别率,并且新的拒识策略较其他拒识策略取得更好的拒识效果.展开更多
Automatic partition of Chinese sentence group is very important to the statistical machine translation system based on discourse. This paper presents an approach to this issue： first, each sentence in a discourse is ...Automatic partition of Chinese sentence group is very important to the statistical machine translation system based on discourse. This paper presents an approach to this issue： first, each sentence in a discourse is expressed as a feature vector; second, a special hierarchical clustering algorithm is applied to present a discourse as a sentence group tree. In this paper, local reoccurrence measure is proposed to the selection of key phras and the evaluation of the weight of key phrases. Experimental results show our approach promising.展开更多
Different types of vegetation occupy different geomorphology and water gradient environments in the San- jiang Plain, indicating that the soil moisture dynamics and water balance patterns of the different vegetation c...Different types of vegetation occupy different geomorphology and water gradient environments in the San- jiang Plain, indicating that the soil moisture dynamics and water balance patterns of the different vegetation communi- ties might differ from each other. In this paper, a lowland system, perpendicular to the Nongjiang River in the Honghe National Nature Reserve （HNNR）, was selected as the study area. The area was occupied by the non-wetland plant forest and the typical wetland plant meadow. The Microsoft Windows-based finite element analysis software package for simulating water, heat, and solute transport in variably saturated porous media （HYDRUS）, which can quantita- tively simulate water, heat, and/or solute movement in variably-saturated porous media, was used to simulate soil moisture dynamics in the root zone （20-40 cm） of those two plant communities during the growing season in 2005. The simulation results for soil moisture were in a good agreement with measured data, with the coefficient of determi- nation （R2） of 0.44-0.69 and root mean square error （RMSE） ranging between 0.0291 cm3/cm3 and 0.0457 cm3/cm3, and index of agreement （d） being from I）.612 to 0.968. During the study period, the volumetric soil moisture content of meadow increased with the depth and its coefficient of variation decreased with the depth （from 20 cm to 40 cm）, while under the forest the soil moisture content at different depths varied irregularly. The calculated result of water budget showed that the water budget deficit of the meadow was higher than that of the forest, suggesting that the meadow is more likely to suffer from water stress than the forest. The quantitative simulation by HYDRUS in this study did not take surface runoff and plant growth processes into account. Improved root water uptake and surface runoff models will be needed for higher accuracy in further researches.展开更多
文摘全自动开放式人机区分图灵测试(CAPTCHA)是基于人工智能领域开放性问题而设计的网络安全技术,CAPTCHA识别是该研究领域的重要分支.长短时记忆(Long Short Term Memory,LSTM)型递归神经网络(Recurrent Neural Network,RNN)已被成功应用于CAPTCHA识别,LSTM型RNN实质上是一维RNN,而文本型CAPTCHA为二维图像.提出使用二维RNN对CAPTCHA进行识别.二维RNN能够很好的将特征提取同识别相结合,同时具有较好的上下文保持特性,从而更适合文本型CAPTCHA识别.同时为了进一步提高识别的可靠性,提出一种基于支持向量机(Support vector machine,SVM)的拒识策略,实验结果表明二维RNN较一维RNN能够获得更好的识别率,并且新的拒识策略较其他拒识策略取得更好的拒识效果.
基金National High Technology Research and Development Program of China ( No.2006AA01Z139)Young NaturalScience Foundation of Fujian Province of China ( No.2008F3105)+1 种基金Natural Science Foundation of Fujian Province of China ( No.2006J0043)Fund of Key Research Project of Fujian Province of China (No.2006H0038)
文摘Automatic partition of Chinese sentence group is very important to the statistical machine translation system based on discourse. This paper presents an approach to this issue： first, each sentence in a discourse is expressed as a feature vector; second, a special hierarchical clustering algorithm is applied to present a discourse as a sentence group tree. In this paper, local reoccurrence measure is proposed to the selection of key phras and the evaluation of the weight of key phrases. Experimental results show our approach promising.
基金Under the auspices of National Natural Science Foundation of China (No. 40871241,41001050)Direction Project of Chinese Academy of Science (KZCX2-YW-Q06-03)
文摘Different types of vegetation occupy different geomorphology and water gradient environments in the San- jiang Plain, indicating that the soil moisture dynamics and water balance patterns of the different vegetation communi- ties might differ from each other. In this paper, a lowland system, perpendicular to the Nongjiang River in the Honghe National Nature Reserve （HNNR）, was selected as the study area. The area was occupied by the non-wetland plant forest and the typical wetland plant meadow. The Microsoft Windows-based finite element analysis software package for simulating water, heat, and solute transport in variably saturated porous media （HYDRUS）, which can quantita- tively simulate water, heat, and/or solute movement in variably-saturated porous media, was used to simulate soil moisture dynamics in the root zone （20-40 cm） of those two plant communities during the growing season in 2005. The simulation results for soil moisture were in a good agreement with measured data, with the coefficient of determi- nation （R2） of 0.44-0.69 and root mean square error （RMSE） ranging between 0.0291 cm3/cm3 and 0.0457 cm3/cm3, and index of agreement （d） being from I）.612 to 0.968. During the study period, the volumetric soil moisture content of meadow increased with the depth and its coefficient of variation decreased with the depth （from 20 cm to 40 cm）, while under the forest the soil moisture content at different depths varied irregularly. The calculated result of water budget showed that the water budget deficit of the meadow was higher than that of the forest, suggesting that the meadow is more likely to suffer from water stress than the forest. The quantitative simulation by HYDRUS in this study did not take surface runoff and plant growth processes into account. Improved root water uptake and surface runoff models will be needed for higher accuracy in further researches.
