Intelligent deep learning-based models have made significant progress for automated source code semantics embedding, and current research works mainly leverage natural language-based methods and graph-based methods. However, natural language-based methods do not capture the rich semantic structural information of source code, and graph-based methods do not utilize rich distant information of source code due to the high cost of message-passing steps. In this article, we propose a novel interpretable model, called graph tensor convolution neural network (GTCN), to generate accurate code embedding, which is capable of comprehensively capturing the distant information of code sequences and rich code semantics structural information. First, we propose to utilize a high-dimensional tensor to integrate various heterogeneous code graphs with node sequence features, such as control flow, data flow. Second, inspired by the current advantages of graph-based deep learning and efficient tensor computations, we propose a novel interpretable graph tensor convolution neural network for learning accurate code semantic embedding from the code graph tensor. Finally, we evaluate three popular applications on the GTCN model: variable misuse detection, source code prediction, and vulnerability detection. Compared with current state-of-the-art methods, our model achieves higher scores with respect to the top-1 accuracy while costing less training time.

Abstract Using multi‐task learning to extract code features can effectively increase the information of the features. However, the existing multi‐task learning methods mainly have two limitations: (1) They cannot extract enough code‐related information or only extract similar semantic features. Similar multi‐task makes the information in the features increased insufficiently. However, the high difference multi‐task is challenging to converge. (2) They cannot train multi‐task on heterogeneous datasets. In standard multi‐task training, we need to label all tasks for all data, which consumes enormous labor. To solve the above limitations, we select two high difference tasks, the cross‐language code completion task and variable misuse task, to extract expressive semantic code features. We propose an attention‐based feature fusion module to merge information among high difference tasks, avoiding the convergence dilemma of standard multi‐task learning. We propose a federated learning framework, extracting semantic information and using the feature fusion module to integrate multi‐task information among single labeled datasets. We experiment on C# and Python datasets for the code completion and variable misuse tasks. The results show that the performance of fusion features by FedMTFF improved by up to 22.6% and 15.1% compared to single tasks. We use FedMTFF to perform four cross‐language multi‐task features fusion, exceeding the current best baseline by 24.1%.