Parallel imaging techniques can help solving problems of radiofrequency energy deposition and image inhomogeneity， reducing scan time， lowering motion artifacts， and accelerating data acquisition in ultra-high field Magnetic Resonance Imaging （MRI）. To enhance feature extraction ability to MRI complex-valued data and reduce wrap-around artifacts caused by under-sampling in parallel imaging， a Residual Complex convolution scan-specific Robust Artificial-neural-networks for K-space Interpolation （RCRAKI） was proposed. In the algorithm， the raw under-sampled MRI scan data was taken as input， and the advantages of both linear and nonlinear reconstruction methods were combined with a residual structure. In the residual connection part， convolution was used to create a linear reconstruction baseline， while multiple layers of complex convolution were utilized in the main path to compensate for baseline defects， ultimately reconstructing Magnetic Resonance （MR） images with fewer artifacts. Experiments were conducted on data acquired from a 7T ultra-high field MR device developed by the Institute of Energy of Hefei Comprehensive National Science Center， and RCRAKI was compared with residual scan-specific Robust Artificial-neural-networks for K-space Interpolation （rRAKI） under a sampling rate of 40 Automatic Calibration Signals （ACSs） and 8 speedup ratio for mouse imaging quality across different anatomical planes. Experimental results show that in sagittal plane， the proposed algorithm has the Normalized Root Mean Squared Error （NRMSE） decreased by 59.74%， the Structural SIMilarity （SSIM） increased by 0.45%， and the Peak Signal-to-Noise Ratio （PSNR） increased by 13.04%； in axial plane， the proposed algorithm has the NRMSE decreased by 7.97%， the SSIM improved slightly （by 0.005%）， and the PSNR increased by 1.09%； in coronal plane， the proposed algorithm has the NRMSE decreased by 35.03%， the PSNR increased by 5.60%， and the SSIM increased by 0.98%. It can be seen that RCRAKI performs well on all the different anatomical planes of MRI data， can reduce the influence of noise amplification at high speedup ratio， and reconstruct MR images with clearer details.

Text sentiment transfer is to change text’s sentiment attribute while preserving its content. Due to the lack of parallel corpora， most of the existing unsupervised methods for text sentiment transfer construct latent representations of sentiment and content through text reconstruction and classification losses， and then realize sentiment transfer. However， this weakly supervised training strategy results in significant model performance degradation under prompt learning paradigms. To address this issue， an unsupervised text sentiment transfer method based on generation prompt was proposed. Firstly， textual content prompts were generated by using a prompt generator. Secondly， the target sentiment prompts were fused as the ultimate prompt. Finally， a two-stage training strategy was formulated to provide smooth training gradients for the model training， thereby solving the problem of model performance degradation. Experimental results on the public dataset for sentiment transfer — Yelp show that the proposed method significantly outperforms the generation based method UnpairedRL in text preservation， sentiment transfer score， and BLEU （BiLingual Evaluation Understudy）， and the improvements are 39.1%， 62.3%， and 14.5%， respectively.

Aiming at overfitting problem caused by memory behavior of Deep Neural Networks （DNNs） on image data with noisy labels， a meta label correction method based on predictions from shallow neural networks was proposed. In this method， with the use of weakly supervised training method， a label reweighting network was set to reweight noise data， meta learning method was employed to facilitate dynamic learning of the model to noise data， and the prediction output from both deep and shallow networks was used as the pseudo labels to train the model. At the same time， the knowledge distillation algorithm was applied to allow the deep network to guide the training of the shallow networks. In this way， the memory behavior of the model was alleviated effectively and the robustness of the model was enhanced. Experiments conducted on CIFAR10/100 and Clothing1M datasets demonstrate the superiority of the proposed method over Meta Label Correction （MLC） method. Particularly， on CIFAR10 dataset with symmetrical noise ratios of 60% and 80%， the accuracy improvements are 3.49 and 1.56 percentage points respectively. Furthermore， in ablation experiments on CIFAR100 dataset with asymmetric noise ratio of 40%， at most 5.32 percentage points accuracy improvement is achieved by the proposed method over models trained without predicted labels， confirming the feasibility and effectiveness of the proposed method.