optirs-learned 0.3.2

OptiRS learned optimizers and meta-learning
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125

//! Auto-generated module
//!
//! 🤖 Generated with [SplitRS](https://github.com/cool-japan/splitrs)

use crate::error::Result;
use scirs2_core::ndarray::Array1;
use scirs2_core::numeric::Float;
use std::collections::HashMap;
use std::fmt::Debug;

use super::types::{
    MetaLearningAlgorithm, MetaTask, MetaTrainingResult, QueryEvaluationResult,
    TaskAdaptationResult,
};

/// Meta-learner trait
pub trait MetaLearner<T: Float + Debug + Send + Sync + 'static>: Send + Sync {
    /// Perform meta-training step
    fn meta_train_step(
        &mut self,
        task_batch: &[MetaTask<T>],
        meta_parameters: &mut HashMap<String, Array1<T>>,
    ) -> Result<MetaTrainingResult<T>>;
    /// Adapt to new task
    fn adapt_to_task(
        &mut self,
        task: &MetaTask<T>,
        meta_parameters: &HashMap<String, Array1<T>>,
        adaptation_steps: usize,
    ) -> Result<TaskAdaptationResult<T>>;
    /// Evaluate on query set
    fn evaluate_query_set(
        &self,
        task: &MetaTask<T>,
        adapted_parameters: &HashMap<String, Array1<T>>,
    ) -> Result<QueryEvaluationResult<T>>;
    /// Get meta-learner type
    fn get_algorithm(&self) -> MetaLearningAlgorithm;
}
#[cfg(test)]
mod tests {
    use crate::meta_learning::types::{
        AdaptationStrategy, AntiForgettingStrategy, AugmentationStrategy,
        ContinualLearningSettings, DistanceMetric, FewShotAlgorithm, FewShotSettings,
        GradientBalancingMethod, InterferenceMitigationStrategy, MAMLConfig, MemoryReplaySettings,
        MemorySelectionCriteria, MetaLearningAlgorithm, MetaLearningConfig, MetricLearningSettings,
        MultiTaskSettings, ReplayStrategy, SharedRepresentationStrategy, SimilarityMeasure,
        TaskIdentificationMethod, TaskSamplingStrategy, TaskWeightingStrategy,
        TransferLearningSettings, TransferStrategy,
    };
    #[test]
    fn test_meta_learning_config() {
        let config = MetaLearningConfig {
            algorithm: MetaLearningAlgorithm::MAML,
            inner_steps: 5,
            outer_steps: 100,
            meta_learning_rate: 0.001,
            inner_learning_rate: 0.01,
            task_batch_size: 16,
            support_set_size: 10,
            query_set_size: 15,
            second_order: true,
            gradient_clip: 1.0,
            adaptation_strategies: vec![AdaptationStrategy::FullFineTuning],
            transfer_settings: TransferLearningSettings {
                domain_adaptation: true,
                source_domain_weights: vec![1.0],
                strategies: vec![TransferStrategy::FineTuning],
                similarity_measures: vec![SimilarityMeasure::CosineDistance],
                progressive_transfer: false,
            },
            continual_settings: ContinualLearningSettings {
                anti_forgetting_strategies: vec![
                    AntiForgettingStrategy::ElasticWeightConsolidation,
                ],
                memory_replay: MemoryReplaySettings {
                    buffer_size: 1000,
                    replay_strategy: ReplayStrategy::Random,
                    replay_frequency: 10,
                    selection_criteria: MemorySelectionCriteria::Random,
                },
                task_identification: TaskIdentificationMethod::Oracle,
                plasticity_stability_balance: 0.5,
            },
            multitask_settings: MultiTaskSettings {
                task_weighting: TaskWeightingStrategy::Uniform,
                gradient_balancing: GradientBalancingMethod::Uniform,
                interference_mitigation: InterferenceMitigationStrategy::OrthogonalGradients,
                shared_representation: SharedRepresentationStrategy::HardSharing,
            },
            few_shot_settings: FewShotSettings {
                num_shots: 5,
                num_ways: 5,
                algorithm: FewShotAlgorithm::MAML,
                metric_learning: MetricLearningSettings {
                    distance_metric: DistanceMetric::Euclidean,
                    embedding_dim: 64,
                    learned_metric: false,
                },
                augmentation_strategies: vec![AugmentationStrategy::Geometric],
            },
            enable_meta_regularization: true,
            meta_regularization_strength: 0.01,
            task_sampling_strategy: TaskSamplingStrategy::Uniform,
        };
        assert_eq!(config.inner_steps, 5);
        assert_eq!(config.task_batch_size, 16);
        assert!(config.second_order);
        assert!(matches!(config.algorithm, MetaLearningAlgorithm::MAML));
    }
    #[test]
    fn test_maml_config() {
        let config = MAMLConfig {
            second_order: true,
            inner_lr: 0.01f64,
            outer_lr: 0.001f64,
            inner_steps: 5,
            allow_unused: true,
            gradient_clip: Some(1.0),
        };
        assert!(config.second_order);
        assert_eq!(config.inner_steps, 5);
        assert_eq!(config.inner_lr, 0.01);
    }
}