Add TinyMLP backend

src/eyetrax/models/tiny_mlp.py

@@ -0,0 +1,46 @@
+from __future__ import annotations
+
+from sklearn.neural_network import MLPRegressor
+
+from . import register_model
+from .base import BaseModel
+
+
+class TinyMLPModel(BaseModel):
+    def __init__(
+        self,
+        *,
+        hidden_layer_sizes: tuple[int, ...] = (64, 32),
+        activation: str = "relu",
+        alpha: float = 1e-4,
+        learning_rate_init: float = 1e-3,
+        max_iter: int = 500,
+        early_stopping: bool = True,
+    ) -> None:
+        super().__init__()
+        self._init_native(
+            hidden_layer_sizes=hidden_layer_sizes,
+            activation=activation,
+            alpha=alpha,
+            learning_rate_init=learning_rate_init,
+            max_iter=max_iter,
+            early_stopping=early_stopping,
+        )
+
+    def _init_native(self, **kw):
+        self.model = MLPRegressor(
+            solver="adam",
+            batch_size="auto",
+            random_state=0,
+            verbose=False,
+            **kw,
+        )
+
+    def _native_train(self, X, y):
+        self.model.fit(X, y)
+
+    def _native_predict(self, X):
+        return self.model.predict(X)
+
+
+register_model("tiny_mlp", TinyMLPModel)