Neural Networks as Artificial Specifications
This work addresses the problem of program specification for developers, but it is incremental as it builds on prior research.
The paper tackled the challenge of training neural networks as artificial specifications for programs by investigating learning modes, aggressiveness levels, and abstraction functions, resulting in promising improvements over earlier experiments with high false positives.
In theory, a neural network can be trained to act as an artificial specification for a program by showing it samples of the programs executions. In practice, the training turns out to be very hard. Programs often operate on discrete domains for which patterns are difficult to discern. Earlier experiments reported too much false positives. This paper revisits an experiment by Vanmali et al. by investigating several aspects that were uninvestigated in the original work: the impact of using different learning modes, aggressiveness levels, and abstraction functions. The results are quite promising.