GPUs support highly parallel workloads, and CUDA streams with Hyper-Q enable concurrent execution. This work proposes a model and scheduler to predict efficient stream pairs without exhaustive testing, using program features to estimate optimal pairs.

This paper proposes a GPU-based evolutionary approach for classification rules with custom crossover and mutation operators. A parallel fitness function uses coverage matrices and reduction vectors to enable efficient CUDA execution of genetic operations.

This study develops a hybrid ANN–genetic algorithm model for thyroid disease classification using clinical data. GA optimizes ANN weights and biases via evolutionary updates, using binary cross-entropy as fitness, improving optimization and generalisation