Performance analysis of a neural network based scheduling algorithm

We analyse the use of artificial neural networks (ANNs) to approximate solving scheduling problems. It is well established that the ANNs main advantage is the small amount of time they take to find an approximate solution, but a question arises: what about the optimality of the obtained solution? A considerable variety of work has been carried out on this subject but, unfortunately, the majority of the studies have focused on the analysis of the classical TSP problem. The obtained results are useful as a reference but can´t be directly extrapolated for real-time systems. We analyse the behaviour of an ANN based scheduling algorithm when scheduling tasks in a real-time system, using the baseline task set from the Hartstone Benchmark which is considered as a typical set for some real-time applications