Summary of Analysing the Behaviour Of Tree-based Neural Networks in Regression Tasks, by Peter Samoaa et al.
Analysing the Behaviour of Tree-Based Neural Networks in Regression Tasks
by Peter Samoaa, Mehrdad Farahani, Antonio Longa, Philipp Leitner, Morteza Haghir Chehreghani
First submitted to arxiv on: 17 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores the application of deep learning models in source code analysis for regression tasks, specifically predicting execution time from source code. The authors extend established tree-based neural network models such as CNNs, Code2Vec, and Transformer-based methods to predict execution time by parsing source code into Abstract Syntax Trees (ASTs). While these models excel in classification tasks, their performance is limited when applied to regression challenges. To address this, the authors propose a novel dual-transformer approach that operates on both source code tokens and AST representations, utilizing cross-attention mechanisms for enhanced interpretability between domains. Additionally, they investigate the adaptation of Graph Neural Networks (GNNs) to this tree-based problem, leveraging the graphical nature of ASTs. The empirical evaluations demonstrate that the proposed dual-transformer model outperforms other tree-based neural networks and GNN-based models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well deep learning models can work with source code for predicting execution time. It takes established models like CNNs, Code2Vec, and Transformer-based methods and tries to use them for this task by turning the source code into a special tree structure called an Abstract Syntax Tree (AST). These models are great for classifying things, but they don’t do so well when it comes to predicting numbers. To fix this, the authors create a new approach that uses two different parts of the model: one for the source code and one for the AST. They also look at how Graph Neural Networks can be used in this problem because they’re good at working with tree-like structures. The results show that their new approach does much better than the other models. |
Keywords
* Artificial intelligence * Classification * Cross attention * Deep learning * Gnn * Neural network * Parsing * Regression * Syntax * Transformer
