Accurate prediction of taxi demands between urban regions can provide decision support information for taxi guidance and scheduling as well as passenger travel recommendation， so as to optimize the relation between taxi supply and demand. However， most of the existing models only focus on modeling and predicting the taxi demands within a region， do not consider enough the spatial-temporal correlation between regions， and pay less attention to the more fine-grained demand prediction between regions. To solve the above problems， a prediction model for taxi demands between urban regions — Origin-Destination fusion with Spatial-Temporal Network （ODSTN） model was proposed. In this model， complex spatial-temporal correlations between regions was captured from spatial dimensions of the regions and region pairs respectively and three temporal dimensions of recent， daily and weekly periods by using graph convolution and attention mechanism， and a new path perception fusion mechanism was designed to combine the multi-angle features and finally realize the taxi demand prediction between urban regions. Experiments were carried out on two real taxi order datasets in Chengdu and Manhattan. The results show that the Mean Absolute Error （MAE）， Root Mean Square Error （RMSE） and Mean Absolute Percentage Error （MAPE） of ODSTN model are 0.897 1， 3.527 4， 50.655 6% and 0.589 6， 1.163 8， 61.079 4%， respectively， indicating that ODSTN model has high accuracy in taxi demand prediction tasks.

In Hematoxylin-Eosin （HE）-stained pathological images， the uneven distribution of cell staining and the diversity of various tissue morphologies bring great challenges to automated segmentation. Traditional convolutions cannot capture the correlation features between pixels in a large neighborhood， making it difficult to further improve the segmentation performance. Therefore， a Multi-Channel Segmentation Network with gated axial self-attention （MCSegNet） model was proposed to achieve accurate segmentation of nuclei in pathological images. In the proposed model， a dual-encoder and decoder structure was adopted， in which the axial self-attention encoding channel was used to capture global features， while the convolutional encoding channel based on residual structure was used to obtain local fine features. The feature representation was enhanced by feature fusion at the end of the encoding channel， providing a good information base for the decoder. And in the decoder， segmentation results were gradually generated by cascading multiple upsampling modules. In addition， the improved hybrid loss function was used to alleviate the common problem of sample imbalance in pathological images effectively. Experimental results on MoNuSeg2020 public dataset show that the improved segmentation method is 2.66 percentage points and 2.77 percentage points higher than U-Net in terms of F1-score and Intersection over Union （IoU） indicators， respectively， and effectively improves the pathological image segmentation effect and the reliability of clinical diagnosis.

In order to study the problem about data combinations occurring in real life, different kinds of data information were combined together, leading to a structure called associated-combinatorial structure. Actually, the structure was constituted by a data set, an associated relation and a partition. The aim was to use the structure to set up a method of data combination. To this end, the associated-combinatorial structure was transformed into a granulation structure by granulating the associated relation. In this process, data combinations were completed in accordance with the data classifications. Moreover, because an associated-combinatorial structure or a granulation structure could be represented by the associated matrix, the transformation from a structure to another structure was characterized by algebraic calculations determined by matrix transformations. Therefore, the research not only involved theoretical analysis for the data combination, but also established the data processing method connected with matrix transformations. Accordingly, a computer program with linear complexity was formulated according to the data combinations method. The experimental result proves that the program is accurate and fast.

Aiming at the nonlinear modeling of Power Amplifier (PA), an improved Radial Basis Function Neural Networks (RBFNN) model was proposed. Firstly, time-delay of cross terms and output feedback were added in the input. Parameters (weigths and centers) of the proposed model were extracted using the Orthogonal Least Square (OLS) algorithm. Then Doherty PA was trained and validated successfully by 15MHz three-carrier Wideband Code Division Multiple Access (WCDMA) signal, and the Normalized Mean Square Error (NMSE) can reach -45dB. Finally, the inverse class F power amplifier was used to test the universality of the model. The simulation results show that the model can more truly fit characteristics of power amplifier.

The original New Edge-Directed Interpolation (NEDI) algorithm is of high complexity, difficult for hardware implementation, and the interpolated images may suffer from blurring edges around edge area. To achieve a better subjective quality, an improved NEDI algorithm was proposed in this paper. In the new algorithm, a circular window was adopted, and the interpolation coefficient calculation was calculated only once, which could be reused in interpolating the center-pixels, thus the errors introduced by iterative computation were avoided and the interpolation time was saved. As to non-center pixels, six original neighbors were involved to estimate local covariance characteristics at high resolution. In comparison with the results of bi-cubic interpolation and the traditional NEDI, the experimental results indicate that proposed algorithm can eliminate the sawtooth of the interpolated picture in large-scale, and decrease the computational complexity.