How kjaved0/mrcl works
This repository is a research codebase for Online-aware Meta-learning (OML), with training and evaluation workflows for sequential classification and regression, plus pretraining and SR-NN comparison runs. The principal workflow is to configure an experiment, obtain datasets and sampled tasks, train a learner/meta-learner, and evaluate or visualize learned representations. The graph includes major README capabilities and indexed modules; edges are limited to README statements, visible calls, or SOURCE INDEX references, so unverified runtime details are omitted.
Main entry points: learner.py Neural learner · oml_omniglot.py OML training · evaluate_omniglot.py Omniglot evaluation · evaluate_imagenet.py ImageNet evaluation
Components 26
Experiment workflows
- OML training
oml_omniglot.py - Regression training
oml_regression.py - MAML baseline
maml-rep_omniglot.py - Omniglot pretraining
pretraining_omniglot.py - ImageNet pretraining
pretraining_imagenet.py - Omniglot SR-NN
srnn_omniglot.py - ImageNet SR-NN
srnn_imagenet.py - Omniglot evaluation
evaluate_omniglot.py - ImageNet evaluation
evaluate_imagenet.py - Regression evaluation
evaluate_regression.py - Representation visualization
visualize_representations.py - Classification options
configs/classification/class_parser.py - Evaluation options
configs/classification/class_parser_eval.py - Regression options
configs/regression/reg_parser.py
Data and tasks
- Dataset selection
datasets/datasetfactory.py - Omniglot dataset
datasets/omniglot.py - Mini-ImageNet dataset
datasets/miniimagenet.py - Task sampling
datasets/task_sampler.py
Learning models
- Model configuration
model/modelfactory.py - Neural learner
model/learner.py - Meta-learning
model/meta_learner.py - Online classification
model/meta_classification.py - OML layer operations
oml/nn.py
Experiment support
- Experiment tracking
experiment/experiment.py - Colored logging
utils/colorer.py
Other components
- Researcher
Connections 38
- Researcher → OML training: runs training
- Researcher → Omniglot evaluation: evaluates models
- Researcher → Representation visualization: visualizes models
- OML training → Classification options: uses options
oml_omniglot.py - OML training → Experiment tracking: uses tracking
oml_omniglot.py - OML training → Dataset selection: selects data
oml_omniglot.py - OML training → Task sampling: samples tasks
oml_omniglot.py - OML training → Model configuration: builds model config
oml_omniglot.py - OML training → Meta-learning: trains meta-learner
oml_omniglot.py - Regression training → Regression options: uses options
oml_regression.py - Regression training → Task sampling: samples tasks
oml_regression.py - Regression training → Meta-learning: trains meta-learner
oml_regression.py - MAML baseline → Meta-learning: uses meta-learner
maml-rep_omniglot.py - Omniglot pretraining → Neural learner: trains learner
- ImageNet pretraining → Neural learner: trains learner
pretraining_imagenet.py - Omniglot SR-NN → Neural learner: uses learner
srnn_omniglot.py - ImageNet SR-NN → Neural learner: uses learner
srnn_imagenet.py - Omniglot evaluation → Evaluation options: uses options
evaluate_omniglot.py - Omniglot evaluation → Dataset selection: selects data
evaluate_omniglot.py - Omniglot evaluation → Neural learner: uses learner
evaluate_omniglot.py - Omniglot evaluation → Model configuration: builds model config
evaluate_omniglot.py - Omniglot evaluation → Experiment tracking: uses tracking
evaluate_omniglot.py - ImageNet evaluation → Dataset selection: selects data
evaluate_imagenet.py - ImageNet evaluation → Mini-ImageNet dataset: uses dataset
evaluate_imagenet.py - ImageNet evaluation → Neural learner: uses learner
evaluate_imagenet.py - ImageNet evaluation → Model configuration: builds model config
evaluate_imagenet.py - ImageNet evaluation → Experiment tracking: uses tracking
evaluate_imagenet.py - Regression evaluation → Regression options: uses options
evaluate_regression.py - Regression evaluation → Task sampling: samples tasks
evaluate_regression.py - Regression evaluation → Meta-learning: uses meta-learner
evaluate_regression.py - Regression evaluation → Experiment tracking: uses tracking
evaluate_regression.py - Representation visualization → Dataset selection: selects data
visualize_representations.py - Representation visualization → Neural learner: uses learner
visualize_representations.py - Representation visualization → Model configuration: builds model config
visualize_representations.py - Representation visualization → Experiment tracking: uses tracking
visualize_representations.py - Dataset selection → Omniglot dataset: constructs dataset
datasets/datasetfactory.py - Neural learner → OML layer operations: uses layers
model/learner.py - Meta-learning → Neural learner: constructs learner
model/meta_learner.py