A Newton-type algorithm for federated learning based on incremental Hessian eigenvector sharing.
Nicolò Dal FabbroSubhrakanti DeyMichele RossiLuca SchenatoPublished in: CoRR (2022)
Keyphrases
- learning algorithm
- incremental learning
- learning speed
- detection algorithm
- preprocessing
- reinforcement learning
- single pass
- dynamic programming
- optimal solution
- computational cost
- significant improvement
- supervised learning
- computational complexity
- matching algorithm
- cost function
- active learning
- learning phase
- simulated annealing
- learning process
- optimization algorithm
- convergence rate
- neural network
- segmentation algorithm
- linear combination
- online learning
- linear programming
- least squares
- lower bound