Parameters as interacting particles: long time convergence and asymptotic error scaling of neural networks.
Grant M. RotskoffEric Vanden-EijndenPublished in: NeurIPS (2018)
Keyphrases
- neural network
- pattern recognition
- error rate
- scaling factors
- highly non linear
- root mean square error
- fuzzy inference system
- genetic algorithm
- number of iterations required
- initial conditions
- maximum likelihood
- artificial neural networks
- parameter settings
- parameter values
- particle swarm
- asymptotically optimal
- parameter estimation
- input data
- expected values
- initial guess
- worst case