Reinforcement learning algorithms suffer from the delayed reward problem and usually only perform well when being trained with vast amounts of data. AlignRUDDER overcomes this problem by using a multiple sequence alignment for the initialization performed with ClustalW. In this thesis we work with different datasets and AlignRUDDER to search for optimal align ment hyperparameters for reinforcement learning problems.
Anotace v angličtině
Reinforcement learning algorithms suffer from the delayed reward problem and usually only perform well when being trained with vast amounts of data. AlignRUDDER overcomes this problem by using a multiple sequence alignment for the initialization performed with ClustalW. In this thesis we work with different datasets and AlignRUDDER to search for optimal align ment hyperparameters for reinforcement learning problems.
Reinforcement learning algorithms suffer from the delayed reward problem and usually only perform well when being trained with vast amounts of data. AlignRUDDER overcomes this problem by using a multiple sequence alignment for the initialization performed with ClustalW. In this thesis we work with different datasets and AlignRUDDER to search for optimal align ment hyperparameters for reinforcement learning problems.
Anotace v angličtině
Reinforcement learning algorithms suffer from the delayed reward problem and usually only perform well when being trained with vast amounts of data. AlignRUDDER overcomes this problem by using a multiple sequence alignment for the initialization performed with ClustalW. In this thesis we work with different datasets and AlignRUDDER to search for optimal align ment hyperparameters for reinforcement learning problems.