Improving the Neural GPU Architecture for Algorithm Learning.
Karlis FreivaldsRenars LiepinsPublished in: CoRR (2017)
Keyphrases
- learning algorithm
- detection algorithm
- network architecture
- learning rules
- computational cost
- search space
- dynamic programming
- times faster
- incremental learning
- learning phase
- online learning
- experimental evaluation
- convergence rate
- significant improvement
- k means
- preprocessing
- computational complexity
- machine learning
- learned models
- learning speed
- segmentation algorithm
- expectation maximization
- worst case
- np hard
- optimal solution
- objective function
- neural network
- matching algorithm
- active learning
- prior knowledge
- parallel implementation
- neural model
- bayesian networks
- gpu accelerated