Most of the existing deep learning based face hallucination algorithms only use a single network partition to reconstruct high-resolution output images without considering the structural information in the face images, resulting in the lack of sufficient details in the reconstruction of vital organs on the face. Therefore, a face hallucination algorithm based on combined learning was proposed to tackle this problem. In the algorithm, the regions of interest were reconstructed independently by utilizing the advantages of different deep learning models, thus the data distribution of each face region was different to each other in the process of network training, and different sub-networks were able to obtain more accurate prior information. Firstly, for the face image, the superpixel segmentation algorithm was used to generate the facial component parts and facial background image. Secondly, the facial component image patches were independently reconstructed by the Component-Generative Adversarial Network (C-GAN) and the facial background reconstruction network was used to generate the facial background image. Thirdly, the facial component fusion network was used to adaptively fuse the facial component image patches reconstructed by two different models. Finally, the generated facial component image patches were merged into the facial background image to reconstruct the final face image. The experimental results on FEI dataset show that the Peak Signal to Noise Ratio (PSNR) of the proposed algorithm is respectively 1.23 dB and 1.11 dB higher than that of the face image hallucination algorithms Learning to hallucinate face images via Component Generation and Enhancement (LCGE) and Enhanced Discriminative Generative Adversarial Network (EDGAN). The proposed algorithm can perform combined learning of the advantages of different deep learning models to learn and reconstruct more accurate face images as well as expand the sources of image reconstruction prior information.

In the traditional Convolutional Neural Network (CNN), the information cannot be transmitted to each other between the neurons of the same layer, the feature information at the same layer cannot be fully utilized, making the lack of the representation of the characteristics of the sentence system. As the result, the feature learning ability of model is limited and the text classification effect is influenced. Aiming at the problem, a model based on joint network CNN-BiGRU and attention mechanism was proposed. In the model, the CNN-BiGRU joint network was used for feature learning. Firstly, deep-level phrase features were extracted by CNN. Then, the Bidirectional Gated Recurrent Unit (BiGRU) was used for the serialized information learning to obtain the characteristics of the sentence system and strengthen the association of CNN pooling layer features. Finally, the effective feature filtering was completed by adding attention mechanism to the hidden state weighted calculation. Comparative experiments show that the method achieves 91.93% F1 value and effectively improves the accuracy of text classification with small time cost and good application ability.

Chinese addresses on the Internet are always non-normalized, which cannot be used directly in location-based services. To solve the problem, an algorithm to extract administrative divisions from non-normalized Chinese addresses was proposed. Firstly, preprocessing "road" feature word grouping for original data; using administrative division dictionary and moving window maximum matching algorithm, extract all possible administrative region data sets from Chinese address. Then, using the Chinese administrative divisions between the elements of the hierarchical relationship between the characteristics, the administrative set conditional set operation rule was established and the acquired data set was aggregated. using the administrative division of matching, a set of administrative division set rules were established to calculate the credibility of the administrative division. Finally, the credibility of the maximum amount of information the most complete Chinese address of the administrative divisions were obtained. By using the extracted from the Internet about 250000 Chinese address data whether the use of "road" feature word packet processing and whether to carry on the credibility calculation process was verified for the availability of the algorithm, and with the current address matching technology for comparison, the accuracy rate of 93.51%.

The very low-resolution image itself contains less discriminant information and is prone to be interfered by noise, which reduces the recognition rate of the existing face recognition algorithm. In order to solve this problem, a very low resolution face recognition algorithm via Super-Resolution (SR) based on Extreme Learning Machine (ELM) was proposed. Firstly, the sparse expression dictionary of Low-Resolution (LR) and High-Resolution (HR) images were learned from sample base, and the HR image could be reconstructed due to the manifold consistency of LR and HR expression coefficients. Secondly, the ELM model was built on the HR reconstructed images, the connection weight of feedforward neural networks was obtained by training. Lastly, the ELM was used to predict the category attribute of the very low-resolution image. Compared with traditional face recognition algorithm based on Collaborative Representation Classification (CRC), the experimental results show that the recognition rate of the proposed algorithm increases by 2% upon the reconstructed HR images. At the same time, it greatly shortens the recognition time. The simulation results show that the proposed algorithm can effectively solve face recognition problem caused by limited discriminant information in very low-resolution image and it has better recognition ability.