Package samer.models
Interface Model.Trainer
-
- All Known Implementing Classes:
- DiffScaler.OffsetTrainer, DiffScaler.ScaleTrainer, DiffScaler.TensionedTrainer, DiffScaler.Trainer, GeneralisedExponential.Trainer, ICA.NewtonTrainer, ICA.ON2DecayWhenActive, ICA.ON2Trainer, ICA.ON3Trainer, ICAWithScaler.DifferentialTrainer, ICAWithScaler.ScalerTrainer, MatrixTrainer, Mixture.Trainer, Scaler.OffsetTrainer, Scaler.ScaleTrainer, Scaler.Trainer, SparseICA.ON2Trainer, SparseICA.ON3Trainer
- Enclosing interface:
- Model
public static interface Model.TrainerThis represents a training algorithm for a Model Trainer is responsible for counting calls to accumulate() between flushes
-
-
Method Summary
All Methods Instance Methods Abstract Methods Modifier and Type Method and Description voidaccumulate()collect statistics for parameter updatevoidaccumulate(double w)weighted accumulatevoiddispose()voidflush()use collected stats to update parameters and resetvoidoneshot()Must be equivalent to reset(); accumulate(); flush(); but can be optimised for non-batched trainingvoidreset()reset accumulators without updating parameters
-
-
-
Method Detail
-
accumulate
void accumulate()
collect statistics for parameter update
-
accumulate
void accumulate(double w)
weighted accumulate
-
flush
void flush()
use collected stats to update parameters and reset
-
oneshot
void oneshot()
Must be equivalent to reset(); accumulate(); flush(); but can be optimised for non-batched training
-
reset
void reset()
reset accumulators without updating parameters
-
dispose
void dispose()
-
-