PrototypeAdamOptimizer<T>
Models & Types · AiDotNet.Prototypes
Prototype Adam optimizer using vectorized operations via PrototypeVector. Demonstrates GPU acceleration through the Execution Engine pattern.
How It Works
This is a PROTOTYPE for Phase A validation. The production version will integrate with the existing AdamOptimizer class.
Key Difference from Current AdamOptimizer:
- BEFORE: Element-wise for-loops (CPU only, slow)
- AFTER: Vectorized operations (GPU accelerated when using float)
Example:
Constructors
| Constructor | Summary |
|---|---|
PrototypeAdamOptimizer(double learningRate, double beta1, double beta2, double epsilon) |
Initializes a new instance of the PrototypeAdamOptimizer. |
Properties
| Property | Summary |
|---|---|
int TimeStep { get; } |
Gets the current time step (number of updates performed). |
Methods
| Method | Summary |
|---|---|
PrototypeVector<T> UpdateParameters(PrototypeVector<T> parameters, PrototypeVector<T> gradient) |
Updates parameters using the Adam optimization algorithm with vectorized operations. |
string ToString() |
Returns a string representation of the optimizer configuration. |
void Reset() |
Resets the optimizer state (clears moment estimates). |