New Estimation interface Implementation

.gitignore

@@ -22,3 +22,4 @@ docs/build
 .coverage
 
 .jupyter
+.venv/

hmp/__init__.py

@@ -6,6 +6,7 @@
     basedata,
     crossvalidation,
     distributions,
+    estimators,
     io,
     models,
     patterndata,
@@ -22,4 +23,5 @@
 
 
 __all__ = ["models", "simulations", "utils", "visu", "io", "patterns", "projectors",
-           "patterndata", "basedata", "distributions" ,"crossvalidation", "__version__"]
+           "patterndata", "basedata", "distributions" ,"crossvalidation", "estimators",
+           "__version__"]

hmp/distributions.py

@@ -264,4 +264,3 @@ def mean_to_scale(self, mean: float) -> float:
             The calculated scale parameter.
         """
         return mean / self.gamma_func(1 + 1 / self.shape)
-

hmp/estimators/__init__.py

@@ -0,0 +1,10 @@
+"""Parameter estimation methods for HMP models."""
+
+from .base import BaseEstimator, EstimationResult
+from .em import EMEstimator
+
+__all__ = [
+    "BaseEstimator",
+    "EstimationResult",
+    "EMEstimator",
+]

hmp/estimators/base.py

@@ -0,0 +1,99 @@
+"""Base classes for parameter estimation in HMP models."""
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import Any, Optional
+
+import numpy as np
+
+from hmp.patterndata import PatternData
+
+
+@dataclass
+class EstimationResult:
+    """Results from parameter estimation.
+
+    Parameters
+    ----------
+    channel_pars : np.ndarray
+        Estimated channel parameters
+    time_pars : np.ndarray
+        Estimated time distribution parameters
+    likelihood : float
+        Final log-likelihood value
+    converged : bool
+        Whether estimation converged
+    n_iterations : int
+        Number of iterations performed
+    diagnostics : dict
+        Estimation-specific diagnostic information
+    uncertainty : dict, optional
+        Parameter uncertainty measures (for Bayesian methods)
+    """
+
+    channel_pars: np.ndarray
+    time_pars: np.ndarray
+    likelihood: float
+    converged: bool
+    n_iterations: int
+    diagnostics: dict[str, Any] = field(default_factory=dict)
+    uncertainty: Optional[dict[str, Any]] = None
+
+
+class BaseEstimator(ABC):
+    """Abstract base class for parameter estimation methods.
+
+    This class defines the interface that all parameter estimation methods
+    must implement to work with HMP models.
+    """
+
+    def __init__(self):
+        """Initialize the estimator."""
+        self.fitted = False
+
+    @abstractmethod
+    def fit(
+        self,
+        model,
+        pattern_data: PatternData,
+        initial_channel_pars: np.ndarray,
+        initial_time_pars: np.ndarray,
+        groups: np.ndarray = None,
+        cpus: int = 1,
+    ) -> EstimationResult:
+        """Estimate model parameters.
+
+        Parameters
+        ----------
+        model : BaseModel
+            Model providing the likelihood and the expectation step.
+        pattern_data : PatternData
+            Preprocessed data cross-correlated with the pattern of the model.
+        initial_channel_pars : np.ndarray
+            Initial channel parameter values, one per starting point.
+        initial_time_pars : np.ndarray
+            Initial time distribution parameter values, one per starting point.
+        groups : np.ndarray, optional
+            Array indicating the groups for grouping modeling. Default is None.
+        cpus : int, optional
+            Number of cores to use in multiprocessing functions. Default is 1.
+
+        Returns
+        -------
+        EstimationResult
+            Results of parameter estimation
+        """
+        pass
+
+    @property
+    def is_fitted(self) -> bool:
+        """Whether the estimator has been fitted."""
+        return self.fitted
+
+    def get_method_name(self) -> str:
+        """Get the name of the estimation method."""
+        return self.__class__.__name__
+
+    def supports_uncertainty(self) -> bool:
+        """Whether this estimator provides uncertainty estimates."""
+        return False