Traditional machine learning in plant phenotyping research requires the assistance of professional data scientists and domain experts to adjust the structure and hy-perparameters tuning of neural network models with much human intervention, making the model training and deployment ineffective. In this paper, the automated machine learning method is researched to construct a multi-task learning model for Arabidopsis thaliana genotype classification, leaf number, and leaf area regression tasks. The experimental results show that the genotype classification task’s accuracy and recall achieved 98.78%, precision reached 98.83%, and classification F₁ value reached 98.79%, as well as the R² of leaf number regression task and leaf area regression task reached 0.9925 and 0.9997 respectively. The experimental results demonstrated that the multi-task automated machine learning model can combine the benefits of multi-task learning and automated machine learning, which achieved more bias information from related tasks and improved the overall classification and prediction effect. Additionally, the model can be created automatically and has a high degree of generalization for better phenotype reasoning. In addition, the trained model and system can be deployed on cloud platforms for convenient application.

Classification of rice disease is one significant research topics in rice phenotyping. Recognition of rice diseases such as Bacterialblight, Blast, Brownspot, Leaf smut, and Tungro are a critical research field in rice phenotyping. However, accurately identifying these diseases is a challenging issue due to their high phenotypic similarity. To address this challenge, we propose a rice disease phenotype identification framework which utilizing the transfer learning and SENet with attention mechanism on the cloud platform. The pre-trained parameters are transferred to the SENet network for parameters optimization. To capture distinctive features of rice diseases, the attention mechanism is applied for feature extracting. Experiment test and comparative analysis are conducted on the real rice disease datasets. The experimental results show that the accuracy of our method reaches 0.9573. Furthermore, we implemented a rice disease phenotype recognition platform based microservices architecture and deployed it on the cloud, which can provide rice disease phenotype recognition task as a service for easy usage.

Artificial intelligence for science is a methodology that integrates artificial intelligence into scientific research to improve the precision and efficiency of data analysis and experimental processes. Specifically in potato late blight severity grading, due to the demand for both accuracy and cost-effective deployment, traditional methods are limited by subjective evaluation and timeconsuming manual measurement. In this paper, a lightweight grading model based on an enhanced YOLOv8-UNet3Plus network is proposed to enable objective and accurate potato late blight severity grading. In detail, the YOLOv8 network is optimized by integrating Spatial and Channel Reconstruction Convolution module, Bi-directional Feature Pyramid Network and Powerful-IoU loss, the UNet3Plus network is optimized by embedding Ghost convolution and Multi-Scale Local Response Attention. Experiments on real-world potato late blight datasets demonstrate that our model achieves an precision of 95.73% for leaf localization and an mean Intersection over Union of 82.65% for infected region segmentation with reduced parameters and computational cost. This AI4Science-based model provides an effective solution for potato late blight severity grading.

This paper proposes 3D-MedGAN, MLU-Net and Info-Max-Net models for overcoming the lack of labeled data and extracting the multi-level feature of images in medical image segmentation. 3D-MedGAN is aimed at dealing with the lack of labeled data in medical images. It uses a generative adversarial network to simulate data and then draws newly generated samples from the distribution learned by the model. Training the segmentation model by mixing generated samples with real samples can effectively improve the effect of the segmentation model. MLU-Net uses multiple layers of different levels of convolutional angles to extract feature information from multiple angles in medical images. By adopting the attention mechanism to fuse the multi-level feature information, MLU-Net is able to improve the feature expressions and segmentation effect. Info-Max-Net is aimed at handling the noise problem in medical images. When the information in the images is complex, it is difficult to extract features. Using mutual information to measure the dependency between the image and the extracted features can effectively reduce noise in the image and improve the effect of segmentation. At the same time, for solving the problem that the high dimension of the image makes it difficult to measure mutual information, this paper uses a lower bound BL-estimator to measure the mutual information between the optimized image and the extracted features. Therefore, the model can maintain a high convergence speed as it approaches the true value of mutual information. Considering that the quality of images generated by 3D-MedGAN are not as good as the original images, we combine the 3D-MedGAN, MLU-Net, and Info-Max-Net to improve the sensitivity and the power of feature extraction of the hybird model. The effectiveness of our model is verified through experiments of the 3D-MedGAN, MLU-Net, Info-Max-Net, and the hybrid model over the LIVER100 dataset.

Recommendation systems are now widely used on the Internet. In recommendation systems, user preferences are predicted by the interaction of users with products, such as clicks or purchases. Usually, the heterogeneous information network is used to capture heterogeneous semantic information in data, which can be used to solve the sparsity problem and the cold-start problem. In a more complex heterogeneous information network, the types of nodes and edges are very large, so there are lots of types of metagraphs in a complex heterogeneous information network. At the same time, machine learning tasks on heterogeneous information networks have a large number of parameters and neural network architectures that need to be set artificially. The main goal is to find the optimal hyperparameter settings and neural network architectures for the performance of a task in the set of hyperparameter space. To address this problem, we propose a metapath search method for heterogeneous information networks based on a network architecture search, which can search for metapaths that are more suitable for different heterogeneous information networks and recommendation tasks. We conducted experiments on Amazon and Yelp datasets and compared the architecture settings obtained from an automatic search with manually set structures to verify the effectiveness of the algorithm.