对于图形神经网络的全球解释监督.
Negar Etemadyrad1, Yuyang Gao2, Sai Manoj Pudukotai Dinakarrao1
1Department of Electrical and Computer Engineering, George Mason University, Fairfax, VA, United States.
Frontiers in big data
|July 16, 2024
概括
全球GNN解释监督 (GGNES) 通过生成准确的全局解释,提高模型理解和性能来提高图形神经网络 (GNN) 的解释性.
科学领域:
- 人工智能的人工智能
- 机器学习 机器学习
- 图形神经网络 图形神经网络
背景情况:
- 图形神经网络 (GNN) 在图形结构数据上的预测任务中越来越受欢迎.
- 虽然GNN的可解释性至关重要,但目前的方法主要集中在生成解释,而不是它们的准确性或全球适用性.
- 现有的GNN解释监督 (GNES) 框架提高了本地解释的合理性,但缺乏全球范围.
研究的目的:
- 为了解决在GNN中局部解释的局限性.
- 开发一种方法,为GNN预测生成准确而忠实的全球解释.
- 增强GNES框架的解释能力,使GNN得到全球的理解.
主要方法:
- 拟议的全球GNN解释监督 (GGNES) 技术.
- 使用训练有素的GNN来创建本地解释.
- 将本地解释集成到基于全球逻辑的GNN解释器中,用于全球解释学习.
- 雇佣了GNN的代培训和解释者,以获得合理的全球解释.
主要成果:
- GGNES有效地提高了全球GNN解释的质量.
- 拟议的方法保持或甚至提高了骨干GNN模型的预测性能.
- 实验结果证明了GGNES在生成合理和准确的全球解释方面的有效性.
结论:
- 通过专注于全球解释,GGNES在GNN可解释性方面取得了重大进展.
- 代训练方法确保了解释的合理性和模型性能.
- 对于需要对GNN行为的全球理解的应用程序,GGNES提供了一个强大的解决方案.
相关概念视频
Vector Algebra: Graphical Method
12.0K
Vectors can be multiplied by scalars, added to other vectors, or subtracted from other vectors. The vector sum of two (or more) vectors is called the resultant vector or, for short, the resultant.
We use the laws of geometry to construct resultant vectors, followed by trigonometry to find vector magnitudes and directions. For a geometric construction of the sum of two vectors in a plane, we follow the parallelogram rule. Suppose two vectors are at arbitrary positions. Translate either one of...
We use the laws of geometry to construct resultant vectors, followed by trigonometry to find vector magnitudes and directions. For a geometric construction of the sum of two vectors in a plane, we follow the parallelogram rule. Suppose two vectors are at arbitrary positions. Translate either one of...
12.0K
End Point Prediction: Gran Plot
313
A Gran plot is used to predict the equivalence volume or endpoint of a potentiometric or acid-base titration without reaching the endpoint. Typically, titration data is collected as a function of the titrant's volume up to a point less than the equivalence volume and then transformed into a linear format. The straight line is extended to the x-axis, indicating the necessary titrant volume to achieve the equivalence point.
For potentiometric titration, the Gran plot is created by plotting...
For potentiometric titration, the Gran plot is created by plotting...
313
SFG Algebra
115
In Signal Flow Graph (SFG) algebra, the value a node represents is determined by the sum of all signals entering that node. This summed value is then transmitted through every branch leaving the node, making the SFG a powerful tool for visualizing and analyzing control systems.
Each node in an SFG corresponds to a variable, and the interactions between nodes are represented by branches with associated gains. When multiple branches lead into a node, the value at that node is the sum of the...
Each node in an SFG corresponds to a variable, and the interactions between nodes are represented by branches with associated gains. When multiple branches lead into a node, the value at that node is the sum of the...
115
Multi-input and Multi-variable systems
105
Cruise control systems in cars are designed as multi-input systems to maintain a driver's desired speed while compensating for external disturbances such as changes in terrain. The block diagram for a cruise control system typically includes two main inputs: the desired speed set by the driver and any external disturbances, such as the incline of the road. By adjusting the engine throttle, the system maintains the vehicle's speed as close to the desired value as possible.
In the absence...
In the absence...
105
Network Function of a Circuit
280
Frequency response analysis in electrical circuits provides vital insights into a circuit's behavior as the frequency of the input signal changes. The transfer function, a mathematical tool, is instrumental in understanding this behavior. It defines the relationship between phasor output and input and comes in four types: voltage gain, current gain, transfer impedance, and transfer admittance. The critical components of the transfer function are the poles and zeros.
280
Superposition Theorem
578
The superposition principle is a fundamental concept stating that in a linear circuit, the voltage across (or current through) an element can be determined by summing the individual contributions of each independent source acting in isolation. When dealing with linear circuits containing multiple independent sources, this principle serves as a valuable tool for analysis. To apply the superposition principle effectively, one should focus on a single independent source at a time while...
578


