mL-BFGS: A Momentum-based L-BFGS for Distributed Large-Scale Neural Network Optimization.
Yue NiuZalan FabianSunwoo LeeMahdi SoltanolkotabiSalman AvestimehrPublished in: CoRR (2023)
Keyphrases
- neural network
- stochastic gradient descent
- convergence rate
- learning rate
- global convergence
- limited memory
- superlinear convergence
- quasi newton
- quasi newton method
- line search
- distributed systems
- distributed stream processing
- least squares
- maximum likelihood
- optimization algorithm
- convergence speed
- cooperative
- fuzzy logic
- image reconstruction from projections
- data intensive
- distributed environment
- optimization methods
- small scale
- high scalability
- matrix factorization
- neural nets
- fault diagnosis
- artificial neural networks
- neural network is trained
- fuzzy neural network
- multi agent
- objective function
- highly non linear
- activation function
- learning algorithm
- real world
- cost function
- global optimization
- optimization procedure
- optimization process
- peer to peer
- influence diagrams
- step size
- neural network model