Continual Learning

59 public types in this namespace, organized by kind.

Models & Types (28)

Type Summary
AveragedGEM<T> Implements Averaged Gradient Episodic Memory (A-GEM) for continual learning.
BufferStatistics Statistics about the experience replay buffer.
ContinualEvaluationResult<T> Comprehensive evaluation result across all learned tasks.
ContinualLearningResult<T> Result from training on a single task in continual learning.
DataPoint<T, TInput, TOutput> Represents a single data point with input and output.
EWCTrainer<T, TInput, TOutput> Continual learning trainer using Elastic Weight Consolidation (EWC).
ElasticWeightConsolidation<T, TInput, TOutput> Elastic Weight Consolidation (EWC) strategy for continual learning.
EpochEventArgs<T> Event arguments for epoch completion events.
ExpectedGradientLength<T, TInput, TOutput> Expected Gradient Length (EGL) strategy for continual learning.
ExperienceDataPoint<T, TInput, TOutput> Alias for DataPoint used in experience replay contexts.
ExperienceReplayBuffer<T, TInput, TOutput> A memory buffer for storing examples from previous tasks for experience replay.
ExperienceReplay<T> Implements Experience Replay for continual learning.
GEMTrainer<T, TInput, TOutput> Continual learning trainer using Gradient Episodic Memory (GEM).
GenerativeReplay<T> Implements Generative Replay (also known as Deep Generative Replay) for continual learning.
GradientEpisodicMemory<T, TInput, TOutput> Gradient Episodic Memory (GEM) strategy for continual learning with gradient constraints.
LearningWithoutForgetting<T, TInput, TOutput> Learning without Forgetting (LwF) strategy for continual learning.
LwFTrainer<T, TInput, TOutput> Continual learning trainer using Learning without Forgetting (LwF).
MASTrainer<T, TInput, TOutput> Continual learning trainer using Memory Aware Synapses (MAS).
MemoryAwareSynapses<T, TInput, TOutput> Memory Aware Synapses (MAS) strategy for continual learning.
OnlineEWC<T> Implements Online Elastic Weight Consolidation (Online EWC) for continual learning.
PackNet<T, TInput, TOutput> PackNet strategy for continual learning through network pruning.
ProgressiveNeuralNetworks<T, TInput, TOutput> Progressive Neural Networks (PNN) strategy for continual learning.
SITrainer<T, TInput, TOutput> Continual learning trainer using Synaptic Intelligence (SI).
SynapticIntelligence<T, TInput, TOutput> Synaptic Intelligence (SI) strategy for continual learning.
TaskCompletedEventArgs<T> Event arguments for task completion events.
TaskEvaluationResult<T> Result from evaluating model performance on a single task.
TaskEventArgs Event arguments for task events.
VariationalContinualLearning<T> Implements Variational Continual Learning (VCL) for Bayesian continual learning.

Base Classes (2)

Type Summary
ContinualLearnerBase<T, TInput, TOutput> Base class for continual learning trainers that provides common functionality.
ContinualLearningStrategyBase<T, TInput, TOutput> Abstract base class for continual learning strategies providing common functionality.

Interfaces (8)

Type Summary
IContinualLearnerConfig<T> Configuration interface for continual learning algorithms.
IContinualLearner<T, TInput, TOutput> Interface for continual learning trainers that can learn multiple tasks sequentially.
IContinualLearningStrategy<T, TInput, TOutput> Strategy interface for continual learning algorithms.
IDistillationStrategy<T, TInput, TOutput> Extended strategy interface for knowledge distillation-based strategies.
IGenerativeModel<T> Interface for generative models used with GenerativeReplay.
IGradientCapable<T, TInput, TOutput> Interface for models that support gradient computation.
IGradientConstraintStrategy<T, TInput, TOutput> Extended strategy interface for gradient-based constraint strategies.
IMemoryBasedStrategy<T, TInput, TOutput> Extended strategy interface for strategies that store task examples.

Enums (6)

Type Summary
BufferStrategy<T> Defines the buffer management strategy.
DistillationLossType Types of distillation loss functions.
ImportanceAccumulationMode Mode for accumulating importance across tasks.
MASImportanceMode Mode for computing parameter importance in MAS.
MemorySamplingStrategy Memory sampling strategies for experience replay.
ReplaySamplingStrategy Strategy for sampling during replay (training time).

Options & Configuration (14)

Type Summary
ContinualLearnerConfig<T> Production-ready configuration for continual learning algorithms.
EGLOptions<T> Configuration options for Expected Gradient Length strategy.
EWCOptions<T> Configuration options for Elastic Weight Consolidation.
EWCTrainerOptions<T> Configuration options for the EWC trainer.
GEMOptions<T> Configuration options for Gradient Episodic Memory strategies.
GEMTrainerOptions<T> Configuration options for the GEM trainer.
LwFOptions<T> Configuration options for Learning without Forgetting.
LwFTrainerOptions<T> Configuration options for the LwF trainer.
MASOptions<T> Configuration options for Memory Aware Synapses.
MASTrainerOptions<T> Configuration options for the MAS trainer.
PNNOptions<T> Configuration options for Progressive Neural Networks strategy.
PackNetOptions<T> Configuration options for PackNet strategy.
SIOptions<T> Configuration options for Synaptic Intelligence.
SITrainerOptions<T> Configuration options for the SI trainer.

Helpers & Utilities (1)

Type Summary
DatasetExtensions Extension methods for IDataset to convert to DataPoints.