ISecondOrderGradientComputable<T, TInput, TOutput>
Interfaces · AiDotNet.Interfaces
Extended gradient computation interface for MAML meta-learning algorithms.
Implements: IGradientComputable<T, TInput, TOutput>
For Beginners
While basic gradient computation tells you how to improve on a single task, second-order gradients tell you how changing your starting point would affect learning on that task.
Think of it like this:
- First-order: “If I start here, which direction improves this task?”
- Second-order: “If I started slightly differently, how would my entire learning trajectory change?”
This is computationally expensive but more accurate for meta-learning.
How It Works
This interface extends IGradientComputable with second-order
gradient computation capability required for full MAML (Model-Agnostic Meta-Learning).
MAML Use Case: Full MAML uses second-order gradients to backpropagate through the inner loop adaptation, computing true meta-gradients that account for how the adaptation process itself changes. Reptile and first-order MAML approximate this with only first-order gradients.
Methods
| Method | Summary |
|---|---|
Vector<T> ComputeSecondOrderGradients(List<ValueTuple<TInput, TOutput>> adaptationSteps, TInput queryInput, TOutput queryTarget, ILossFunction<T> lossFunction, T innerLearningRate) |
Computes second-order gradients (Hessian-vector product) for full MAML meta-learning. |