oikan 0.0.3.2__py3-none-any.whl → 0.0.3.4__py3-none-any.whl

oikan/exceptions.py

@@ -4,4 +4,28 @@ class OIKANError(Exception):
 
 class ModelNotFittedError(OIKANError):
     """Raised when a method requires a fitted model."""
-    pass
+    pass
+
+class InvalidParameterError(OIKANError):
+    """Raised when an invalid parameter value is provided."""
+    pass
+
+class DataDimensionError(OIKANError):
+    """Raised when input data has incorrect dimensions."""
+    pass
+
+class NumericalInstabilityError(OIKANError):
+    """Raised when numerical computations become unstable."""
+    pass
+
+class FeatureExtractionError(OIKANError):
+    """Raised when feature extraction or transformation fails."""
+    pass
+
+class ModelSerializationError(OIKANError):
+    """Raised when model saving/loading operations fail."""
+    pass
+
+class ConvergenceError(OIKANError):
+    """Raised when the model fails to converge during training."""
+    pass    

oikan/model.py

@@ -3,13 +3,14 @@ import torch
 import torch.nn as nn
 import torch.optim as optim
 from sklearn.preprocessing import PolynomialFeatures
-from sklearn.linear_model import Lasso
+from sklearn.linear_model import ElasticNet
 from abc import ABC, abstractmethod
 import json
 from .neural import TabularNet
-from .utils import evaluate_basis_functions, get_features_involved
+from .utils import evaluate_basis_functions, get_features_involved, sympify_formula, get_latex_formula
 from sklearn.model_selection import train_test_split
 from sklearn.metrics import r2_score, accuracy_score
+from .exceptions import *
 import sys
 
 class OIKAN(ABC):
@@ -24,12 +25,12 @@ class OIKAN(ABC):
         Activation function for the neural network ('relu', 'tanh', 'leaky_relu', 'elu', 'swish', 'gelu').
     augmentation_factor : int, optional (default=10)
         Number of augmented samples per original sample.
-    polynomial_degree : int, optional (default=2)
-        Maximum degree of polynomial features for symbolic regression.
     alpha : float, optional (default=0.1)
         L1 regularization strength for Lasso in symbolic regression.
     sigma : float, optional (default=0.1)
         Standard deviation of Gaussian noise for data augmentation.
+    top_k : int, optional (default=5)
+        Number of top features to select in hierarchical symbolic regression.
     epochs : int, optional (default=100)
         Number of epochs for neural network training.
     lr : float, optional (default=0.001)
@@ -42,12 +43,30 @@ class OIKAN(ABC):
         Whether to evaluate neural network performance before full training.
     """
     def __init__(self, hidden_sizes=[64, 64], activation='relu', augmentation_factor=10, 
-                 polynomial_degree=2, alpha=0.1, sigma=0.1, epochs=100, lr=0.001, batch_size=32, 
-                 verbose=False, evaluate_nn=False):
+                 alpha=0.1, sigma=0.1, epochs=100, lr=0.001, batch_size=32, 
+                 verbose=False, evaluate_nn=False, top_k=5):
+        if not isinstance(hidden_sizes, list) or not all(isinstance(x, int) and x > 0 for x in hidden_sizes):
+            raise InvalidParameterError("hidden_sizes must be a list of positive integers")
+        if activation not in ['relu', 'tanh', 'leaky_relu', 'elu', 'swish', 'gelu']:
+            raise InvalidParameterError(f"Unsupported activation function: {activation}")
+        if not isinstance(augmentation_factor, int) or augmentation_factor < 1:
+            raise InvalidParameterError("augmentation_factor must be a positive integer")
+        if not isinstance(top_k, int) or top_k < 1:
+            raise InvalidParameterError("top_k must be a positive integer")
+        if not 0 < lr < 1:
+            raise InvalidParameterError("Learning rate must be between 0 and 1")
+        if not isinstance(batch_size, int) or batch_size < 1:
+            raise InvalidParameterError("batch_size must be a positive integer")
+        if not isinstance(epochs, int) or epochs < 1:
+            raise InvalidParameterError("epochs must be a positive integer")
+        if not 0 <= alpha <= 1:
+            raise InvalidParameterError("alpha must be between 0 and 1")
+        if sigma <= 0:
+            raise InvalidParameterError("sigma must be positive")
+        
         self.hidden_sizes = hidden_sizes
         self.activation = activation
         self.augmentation_factor = augmentation_factor
-        self.polynomial_degree = polynomial_degree
         self.alpha = alpha
         self.sigma = sigma
         self.epochs = epochs
@@ -55,6 +74,7 @@ class OIKAN(ABC):
         self.batch_size = batch_size
         self.verbose = verbose
         self.evaluate_nn = evaluate_nn
+        self.top_k = top_k
         self.neural_net = None
         self.symbolic_model = None
         self.evaluation_done = False
@@ -67,23 +87,53 @@ class OIKAN(ABC):
     def predict(self, X):
         pass
 
-    def get_formula(self):
-        """Returns the symbolic formula(s) as a string (regression) or list of strings (classification)."""
+    def get_formula(self, type='original'): 
+        """
+        Returns the symbolic formula(s) as a string (regression) or list of strings (classification). 
+        
+        Parameter:
+        --------
+        type : str, optional (default='original') other options: 'sympied', 'latex'
+            'original' returns the original formula with coefficients, 'sympied' returns sympy simplified formula.
+        """
+        if type.lower() not in ['original', 'sympied', 'latex']:
+            raise InvalidParameterError("Invalid type. Choose 'original', 'sympied', 'latex'.")
         if self.symbolic_model is None:
             raise ValueError("Model not fitted yet.")
         basis_functions = self.symbolic_model['basis_functions']
-        if 'coefficients' in self.symbolic_model:
-            coefficients = self.symbolic_model['coefficients']
-            formula = " + ".join([f"{coefficients[i]:.3f}*{basis_functions[i]}" 
-                                for i in range(len(coefficients)) if coefficients[i] != 0])
-            return formula if formula else "0"
+        if type.lower() == 'original':
+            if 'coefficients' in self.symbolic_model:
+                coefficients = self.symbolic_model['coefficients']
+                formula = " + ".join([f"{coefficients[i]:.6f}*{basis_functions[i]}" 
+                                    for i in range(len(coefficients)) if coefficients[i] != 0])
+                return formula if formula else "0"
+            else:
+                formulas = []
+                for c, coef in enumerate(self.symbolic_model['coefficients_list']):
+                    formula = " + ".join([f"{coef[i]:.6f}*{basis_functions[i]}" 
+                                        for i in range(len(coef)) if coef[i] != 0])
+                    formulas.append(f"Class {self.classes_[c]}: {formula if formula else '0'}")
+                return formulas
+        elif type.lower() == 'sympied':
+            if 'coefficients' in self.symbolic_model:
+                formula = sympify_formula(self.symbolic_model['basis_functions'], self.symbolic_model['coefficients'], self.symbolic_model['n_features'])
+                return formula
+            else: 
+                formulas = []
+                for c, coef in enumerate(self.symbolic_model['coefficients_list']):
+                    formula = sympify_formula(self.symbolic_model['basis_functions'], coef, self.symbolic_model['n_features'])
+                    formulas.append(f"Class {self.classes_[c]}: {formula}")
+                return formulas
         else:
-            formulas = []
-            for c, coef in enumerate(self.symbolic_model['coefficients_list']):
-                formula = " + ".join([f"{coef[i]:.3f}*{basis_functions[i]}" 
-                                    for i in range(len(coef)) if coef[i] != 0])
-                formulas.append(f"Class {self.classes_[c]}: {formula if formula else '0'}")
-            return formulas
+            if 'coefficients' in self.symbolic_model:
+                formula = get_latex_formula(self.symbolic_model['basis_functions'], self.symbolic_model['coefficients'], self.symbolic_model['n_features'])
+                return formula
+            else: 
+                formulas = []
+                for c, coef in enumerate(self.symbolic_model['coefficients_list']):
+                    formula = get_latex_formula(self.symbolic_model['basis_functions'], coef, self.symbolic_model['n_features'])
+                    formulas.append(f"Class {self.classes_[c]}: {formula}")
+                return formulas
 
     def feature_importances(self):
         """
@@ -129,27 +179,32 @@ class OIKAN(ABC):
             File path to save the model. Should end with .json
         """
         if self.symbolic_model is None:
-            raise ValueError("Model not fitted yet.")
-            
+            raise ModelNotFittedError("Model must be fitted before saving")
+        
         if not path.endswith('.json'):
             path = path + '.json'
-            
-        # Convert numpy arrays and other non-serializable types to lists
-        model_data = {
-            'n_features': self.symbolic_model['n_features'],
-            'degree': self.symbolic_model['degree'],
-            'basis_functions': self.symbolic_model['basis_functions']
-        }
         
-        if 'coefficients' in self.symbolic_model:
-            model_data['coefficients'] = self.symbolic_model['coefficients']
-        else:
-            model_data['coefficients_list'] = [coef for coef in self.symbolic_model['coefficients_list']]
-            if hasattr(self, 'classes_'):
-                model_data['classes'] = self.classes_.tolist()
+        try:    
+            # Convert numpy arrays and other non-serializable types to lists
+            model_data = {
+                'n_features': self.symbolic_model['n_features'],
+                'basis_functions': self.symbolic_model['basis_functions']
+            }
+            
+            if 'coefficients' in self.symbolic_model:
+                model_data['coefficients'] = self.symbolic_model['coefficients']
+            else:
+                model_data['coefficients_list'] = [coef for coef in self.symbolic_model['coefficients_list']]
+                if hasattr(self, 'classes_'):
+                    model_data['classes'] = self.classes_.tolist()
+            
+            with open(path, 'w') as f:
+                json.dump(model_data, f, indent=2)
+        except Exception as e:
+            raise ModelSerializationError(f"Failed to save model: {str(e)}")
         
-        with open(path, 'w') as f:
-            json.dump(model_data, f, indent=2)
+        if self.verbose:
+            print(f"Model saved to {path}")
 
     def load(self, path):
         """
@@ -162,22 +217,27 @@ class OIKAN(ABC):
         """
         if not path.endswith('.json'):
             path = path + '.json'
+        
+        try:
+            with open(path, 'r') as f:
+                model_data = json.load(f)
+                
+            self.symbolic_model = {
+                'n_features': model_data['n_features'],
+                'basis_functions': model_data['basis_functions']
+            }
             
-        with open(path, 'r') as f:
-            model_data = json.load(f)
-            
-        self.symbolic_model = {
-            'n_features': model_data['n_features'],
-            'degree': model_data['degree'],
-            'basis_functions': model_data['basis_functions']
-        }
+            if 'coefficients' in model_data:
+                self.symbolic_model['coefficients'] = model_data['coefficients']
+            else:
+                self.symbolic_model['coefficients_list'] = model_data['coefficients_list']
+                if 'classes' in model_data:
+                    self.classes_ = np.array(model_data['classes'])
+        except Exception as e:
+            raise ModelSerializationError(f"Failed to load model: {str(e)}")
         
-        if 'coefficients' in model_data:
-            self.symbolic_model['coefficients'] = model_data['coefficients']
-        else:
-            self.symbolic_model['coefficients_list'] = model_data['coefficients_list']
-            if 'classes' in model_data:
-                self.classes_ = np.array(model_data['classes'])
+        if self.verbose:
+            print(f"Model loaded from {path}")
 
     def _evaluate_neural_net(self, X, y, output_size, loss_fn):
         """Evaluates neural network performance on train-test split."""
@@ -185,7 +245,6 @@ class OIKAN(ABC):
         
         input_size = X.shape[1]
         self.neural_net = TabularNet(input_size, self.hidden_sizes, output_size, self.activation)
-        optimizer = optim.Adam(self.neural_net.parameters(), lr=self.lr)
         
         # Train on the training set
         self._train_neural_net(X_train, y_train, output_size, loss_fn)
@@ -253,7 +312,6 @@ class OIKAN(ABC):
 
     def _generate_augmented_data(self, X):
         """Generates augmented data by adding Gaussian noise."""
-        n_samples = X.shape[0]
         X_aug = []
         for _ in range(self.augmentation_factor):
             noise = np.random.normal(0, self.sigma, X.shape)
@@ -262,37 +320,105 @@ class OIKAN(ABC):
         return np.vstack(X_aug)
 
     def _perform_symbolic_regression(self, X, y):
-        """Performs symbolic regression using polynomial features and Lasso."""
-        poly = PolynomialFeatures(degree=self.polynomial_degree, include_bias=True)
-        X_poly = poly.fit_transform(X)
-        model = Lasso(alpha=self.alpha, fit_intercept=False)
-        model.fit(X_poly, y)
+        """
+        Performs hierarchical symbolic regression using a two-stage approach.
+        
+        Parameters:
+        -----------
+        X : array-like of shape (n_samples, n_features)
+            Input data.
+        y : array-like of shape (n_samples,) or (n_samples, n_classes)
+            Target values or logits.
+        """
+        n_features = X.shape[1]
+        self.top_k = min(self.top_k, n_features)
+        
+        if self.top_k < 1:
+            raise InvalidParameterError("top_k must be at least 1")
+
+        if np.any(np.isnan(X)) or np.any(np.isnan(y)):
+            raise NumericalInstabilityError("Input data contains NaN values")
+
+        if np.any(np.isinf(X)) or np.any(np.isinf(y)):
+            raise NumericalInstabilityError("Input data contains infinite values")
+
+        # Stage 1: Coarse Model
+        coarse_degree = 2  # Fixed low degree for coarse model
+        poly_coarse = PolynomialFeatures(degree=coarse_degree, include_bias=True)
+        X_poly_coarse = poly_coarse.fit_transform(X)
+        model_coarse = ElasticNet(alpha=self.alpha, fit_intercept=False)
+        model_coarse.fit(X_poly_coarse, y)
+
+        # Compute feature importances for original features
+        basis_functions_coarse = poly_coarse.get_feature_names_out()
+        if len(y.shape) == 1 or y.shape[1] == 1:
+            coef_coarse = model_coarse.coef_.flatten()
+        else:
+            coef_coarse = np.sum(np.abs(model_coarse.coef_), axis=0)
+
+        importances = np.zeros(X.shape[1])
+        for i, func in enumerate(basis_functions_coarse):
+            features_involved = get_features_involved(func)
+            for idx in features_involved:
+                importances[idx] += np.abs(coef_coarse[i])
+        
+        if np.all(importances == 0):
+            raise FeatureExtractionError("Failed to compute feature importances - all values are zero")
+
+        # Select top K features
+        top_k_indices = np.argsort(importances)[::-1][:self.top_k]
+
+        # Stage 2: Refined Model
+        # ~ generate additional non-linear features for top K features
+        additional_features = []
+        additional_names = []
+        for i in top_k_indices:
+            # Higher-degree polynomial
+            additional_features.append(X[:, i]**3)
+            additional_names.append(f'x{i}^3')
+            # Non-linear transformations
+            additional_features.append(np.log1p(np.abs(X[:, i])))
+            additional_names.append(f'log1p_x{i}')
+            additional_features.append(np.exp(np.clip(X[:, i], -10, 10)))
+            additional_names.append(f'exp_x{i}')
+            additional_features.append(np.sin(X[:, i]))
+            additional_names.append(f'sin_x{i}')
+        
+        # Combine features
+        X_additional = np.column_stack(additional_features)
+        X_refined = np.hstack([X_poly_coarse, X_additional])
+        basis_functions_refined = list(basis_functions_coarse) + additional_names
+
+        # Fit refined model
+        model_refined = ElasticNet(alpha=self.alpha, fit_intercept=False)
+        model_refined.fit(X_refined, y)
+
+        # Store symbolic model
         if len(y.shape) == 1 or y.shape[1] == 1:
-            coef = model.coef_.flatten()
-            selected_indices = np.where(np.abs(coef) > 1e-6)[0]
+            # Regression
+            coef_refined = model_refined.coef_.flatten()
+            selected_indices = np.where(np.abs(coef_refined) > 1e-6)[0]
             self.symbolic_model = {
                 'n_features': X.shape[1],
-                'degree': self.polynomial_degree,
-                'basis_functions': poly.get_feature_names_out()[selected_indices].tolist(),
-                'coefficients': coef[selected_indices].tolist()
+                'basis_functions': [basis_functions_refined[i] for i in selected_indices],
+                'coefficients': coef_refined[selected_indices].tolist()
             }
         else:
+            # Classification
             coefficients_list = []
-            # Note: Using the same basis functions across classes for simplicity
             selected_indices = set()
             for c in range(y.shape[1]):
-                coef = model.coef_[c]
+                coef = model_refined.coef_[c]
                 indices = np.where(np.abs(coef) > 1e-6)[0]
                 selected_indices.update(indices)
             selected_indices = list(selected_indices)
-            basis_functions = poly.get_feature_names_out()[selected_indices].tolist()
+            basis_functions = [basis_functions_refined[i] for i in selected_indices]
             for c in range(y.shape[1]):
-                coef = model.coef_[c]
+                coef = model_refined.coef_[c]
                 coef_selected = coef[selected_indices].tolist()
                 coefficients_list.append(coef_selected)
             self.symbolic_model = {
                 'n_features': X.shape[1],
-                'degree': self.polynomial_degree,
                 'basis_functions': basis_functions,
                 'coefficients_list': coefficients_list
             }

oikan/utils.py

@@ -1,4 +1,7 @@
 import numpy as np
+import sympy as sp
+import json
+from functools import lru_cache
 
 def evaluate_basis_functions(X, basis_functions, n_features):
     """
@@ -9,7 +12,7 @@ def evaluate_basis_functions(X, basis_functions, n_features):
     X : array-like of shape (n_samples, n_features)
         Input data.
     basis_functions : list
-        List of basis function strings (e.g., '1', 'x0', 'x0^2', 'x0 x1').
+        List of basis function strings (e.g., '1', 'x0', 'x0^2', 'x0 x1', 'log1p_x0').
     n_features : int
         Number of input features.
     
@@ -22,15 +25,26 @@ def evaluate_basis_functions(X, basis_functions, n_features):
     for i, func in enumerate(basis_functions):
         if func == '1':
             X_transformed[:, i] = 1
+        elif func.startswith('log1p_x'):
+            idx = int(func.split('_')[1][1:])
+            X_transformed[:, i] = np.log1p(np.abs(X[:, idx]))
+        elif func.startswith('exp_x'):
+            idx = int(func.split('_')[1][1:])
+            X_transformed[:, i] = np.exp(np.clip(X[:, idx], -10, 10))
+        elif func.startswith('sin_x'):
+            idx = int(func.split('_')[1][1:])
+            X_transformed[:, i] = np.sin(X[:, idx])
         elif '^' in func:
             var, power = func.split('^')
             idx = int(var[1:])
             X_transformed[:, i] = X[:, idx] ** int(power)
         elif ' ' in func:
-            var1, var2 = func.split(' ')
-            idx1 = int(var1[1:])
-            idx2 = int(var2[1:])
-            X_transformed[:, i] = X[:, idx1] * X[:, idx2]
+            vars = func.split(' ')
+            result = np.ones(X.shape[0])
+            for var in vars:
+                idx = int(var[1:])
+                result *= X[:, idx]
+            X_transformed[:, i] = result
         else:
             idx = int(func[1:])
             X_transformed[:, i] = X[:, idx]
@@ -43,21 +57,200 @@ def get_features_involved(basis_function):
     Parameters:
     -----------
     basis_function : str
-        String representation of the basis function, e.g., 'x0', 'x0^2', 'x0 x1'.
+        String representation of the basis function, e.g., 'x0', 'x0^2', 'x0 x1', 'log1p_x0'.
     
     Returns:
     --------
     set : Set of feature indices involved.
     """
-    if basis_function == '1':  # Constant term involves no features
+    if basis_function == '1':
         return set()
     features = set()
-    for part in basis_function.split():  # Split by space for interaction terms
-        if part.startswith('x'):
-            if '^' in part:  # Handle powers, e.g., 'x0^2'
-                var = part.split('^')[0]  # Take 'x0'
-            else:
-                var = part  # Take 'x0' as is
-            idx = int(var[1:])  # Extract index, e.g., 0
+    if '_' in basis_function:  # Handle non-linear functions like 'log1p_x0'
+        parts = basis_function.split('_')
+        if len(parts) == 2 and parts[1].startswith('x'):
+            idx = int(parts[1][1:])
             features.add(idx)
-    return features
+    elif '^' in basis_function:  # Handle powers, e.g., 'x0^2'
+        var = basis_function.split('^')[0]
+        idx = int(var[1:])
+        features.add(idx)
+    elif ' ' in basis_function:  # Handle interactions, e.g., 'x0 x1'
+        for part in basis_function.split():
+            idx = int(part[1:])
+            features.add(idx)
+    elif basis_function.startswith('x'):
+        idx = int(basis_function[1:])
+        features.add(idx)
+    return features
+
+@lru_cache(maxsize=1000)
+def _cached_sympify_formula(basis_functions_tuple, coefficients_tuple, n_features, threshold):
+    """
+    Internal function to perform SymPy formula simplification with caching.
+    
+    Parameters:
+    -----------
+    basis_functions_tuple : tuple
+        Tuple of basis function strings.
+    coefficients_tuple : tuple
+        Tuple of coefficients.
+    n_features : int
+        Number of input features.
+    threshold : float
+        Coefficients with absolute value below this are excluded.
+    
+    Returns:
+    --------
+    str
+        Simplified formula as a string, or '0' if empty.
+    """
+    # Convert tuples back to lists
+    basis_functions = list(basis_functions_tuple)
+    coefficients = list(coefficients_tuple)
+    
+    # Define symbolic variables
+    x = sp.symbols(f'x0:{n_features}')
+    expr = 0
+    
+    # Build the expression
+    for coef, func in zip(coefficients, basis_functions):
+        if abs(coef) < threshold:
+            continue  # Skip negligible coefficients
+        if func == '1':
+            term = coef
+        elif func.startswith('log1p_x'):
+            idx = int(func.split('_')[1][1:])
+            term = coef * sp.log(1 + sp.Abs(x[idx]))
+        elif func.startswith('exp_x'):
+            idx = int(func.split('_')[1][1:])
+            term = coef * sp.exp(x[idx])
+        elif func.startswith('sin_x'):
+            idx = int(func.split('_')[1][1:])
+            term = coef * sp.sin(x[idx])
+        elif '^' in func:
+            var, power = func.split('^')
+            idx = int(var[1:])
+            term = coef * x[idx]**int(power)
+        elif ' ' in func:
+            vars = func.split(' ')
+            term = coef
+            for var in vars:
+                idx = int(var[1:])
+                term *= x[idx]
+        else:
+            idx = int(func[1:])
+            term = coef * x[idx]
+        expr += term
+    
+    # Simplify the expression
+    simplified_expr = sp.simplify(expr)
+    
+    # Convert to string with rounded coefficients
+    def format_term(term):
+        if term.is_Mul:
+            coeff = 1
+            factors = []
+            for factor in term.args:
+                if factor.is_Number:
+                    coeff *= float(factor)
+                else:
+                    factors.append(str(factor))
+            if abs(coeff) < threshold:
+                return None
+            return f"{coeff:.5f}*{'*'.join(factors)}" if factors else f"{coeff:.5f}"
+        elif term.is_Add:
+            return None  # Handle in recursion
+        elif term.is_Number:
+            return f"{float(term):.5f}" if abs(float(term)) >= threshold else None
+        else:
+            return f"{1.0:.5f}*{term}" if abs(1.0) >= threshold else None
+
+    terms = []
+    if simplified_expr.is_Add:
+        for term in simplified_expr.args:
+            formatted = format_term(term)
+            if formatted:
+                terms.append(formatted)
+    else:
+        formatted = format_term(simplified_expr)
+        if formatted:
+            terms.append(formatted)
+    
+    formula = " + ".join(terms).replace("+ -", "- ")
+    return formula if formula else "0"
+
+def sympify_formula(basis_functions, coefficients, n_features, threshold=0.00005):
+    """
+    Simplifies a symbolic formula using SymPy with caching.
+    
+    Parameters:
+    -----------
+    basis_functions : list
+        List of basis function strings (e.g., 'x0', 'x0^2', 'x0 x1', 'exp_x0').
+    coefficients : list
+        List of coefficients corresponding to each basis function.
+    n_features : int
+        Number of input features.
+    threshold : float, optional (default=0.00005)
+        Coefficients with absolute value below this are excluded.
+    
+    Returns:
+    --------
+    str
+        Simplified formula as a string, or '0' if empty.
+    """
+    # Convert inputs to hashable types
+    basis_functions_tuple = tuple(basis_functions)
+    coefficients_tuple = tuple(coefficients)
+    
+    # Call cached function
+    return _cached_sympify_formula(basis_functions_tuple, coefficients_tuple, n_features, threshold)
+
+@lru_cache(maxsize=1000)
+def _cached_get_latex_formula(formula):
+    """
+    Internal function to convert a simplified formula to LaTeX with caching.
+    
+    Parameters:
+    -----------
+    formula : str
+        Simplified formula string.
+    
+    Returns:
+    --------
+    str
+        LaTeX formula as a string.
+    """
+    return sp.latex(sp.sympify(formula))
+
+def get_latex_formula(basis_functions, coefficients, n_features, threshold=0.00005):
+    """
+    Generates a LaTeX formula from the basis functions and coefficients with caching.
+    
+    Parameters:
+    -----------
+    basis_functions : list
+        List of basis function strings (e.g., 'x0', 'x0^2', 'x0 x1', 'exp_x0').
+    coefficients : list
+        List of coefficients corresponding to each basis function.
+    n_features : int
+        Number of input features.
+    threshold : float, optional (default=0.00005)
+        Coefficients with absolute value below this are excluded.
+    
+    Returns:
+    --------
+    str
+        LaTeX formula as a string, or '0' if empty.
+    """
+    # Get simplified formula (cached)
+    formula = sympify_formula(basis_functions, coefficients, n_features, threshold)
+    # Convert to LaTeX (cached)
+    return _cached_get_latex_formula(formula)
+
+if __name__ == "__main__":
+    with open('outputs/california_housing_model.json', 'r') as f:
+        model = json.load(f)
+    print('Sympified formula:', sympify_formula(model['basis_functions'], model['coefficients'], model['n_features']))
+    print('LaTeX formula:', get_latex_formula(model['basis_functions'], model['coefficients'], model['n_features']))

{oikan-0.0.3.2.dist-info → oikan-0.0.3.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: oikan
-Version: 0.0.3.2
+Version: 0.0.3.4
 Summary: OIKAN: Neuro-Symbolic ML for Scientific Discovery
 Author: Arman Zhalgasbayev
 License: MIT
@@ -14,6 +14,7 @@ Requires-Dist: torch
 Requires-Dist: numpy
 Requires-Dist: scikit-learn
 Requires-Dist: tqdm
+Requires-Dist: sympy
 Dynamic: license-file
 
 <!-- logo in the center -->
@@ -57,7 +58,7 @@ OIKAN implements a modern interpretation of the Kolmogorov-Arnold Representation
 
 2. **Neural Implementation**: OIKAN uses a specialized architecture combining:
    - Feature transformation layers with interpretable basis functions
-   - Symbolic regression for formula extraction
+   - Symbolic regression for formula extraction (ElasticNet-based)
    - Automatic pruning of insignificant terms
    
    ```python
@@ -76,15 +77,19 @@ OIKAN implements a modern interpretation of the Kolmogorov-Arnold Representation
    SYMBOLIC_FUNCTIONS = {
        'linear': 'x',           # Direct relationships
        'quadratic': 'x^2',      # Non-linear patterns
+       'cubic': 'x^3',         # Higher-order relationships
        'interaction': 'x_i x_j', # Feature interactions
-       'higher_order': 'x^n'    # Polynomial terms
+       'higher_order': 'x^n',    # Polynomial terms
+       'trigonometric': 'sin(x)', # Trigonometric functions
+       'exponential': 'exp(x)',  # Exponential growth
+       'logarithmic': 'log(x)'  # Logarithmic relationships
    }
    ```
 
 4. **Formula Extraction Process**:
    - Train neural network on raw data
    - Generate augmented samples for better coverage
-   - Perform L1-regularized symbolic regression
+   - Perform L1-regularized symbolic regression (alpha)
    - Prune terms with coefficients below threshold
    - Export human-readable mathematical expressions
 
@@ -114,13 +119,14 @@ model = OIKANRegressor(
     hidden_sizes=[32, 32], # Hidden layer sizes
     activation='relu', # Activation function (other options: 'tanh', 'leaky_relu', 'elu', 'swish', 'gelu')
     augmentation_factor=5, # Augmentation factor for data generation
-    polynomial_degree=2, # Degree of polynomial basis functions
-    alpha=0.1, # L1 regularization strength
+    alpha=0.1, # L1 regularization strength (Symbolic regression)
     sigma=0.1, # Standard deviation of Gaussian noise for data augmentation
+    top_k=5, # Number of top features to select (Symbolic regression)
     epochs=100, # Number of training epochs
     lr=0.001, # Learning rate
     batch_size=32, # Batch size for training
-    verbose=True # Verbose output during training
+    verbose=True, # Verbose output during training
+    evaluate_nn=True # Validate neural network performance before full process
 )
 
 # Fit the model
@@ -134,7 +140,7 @@ mse = mean_squared_error(y_test, y_pred)
 print("Mean Squared Error:", mse)
 
 # Get symbolic formula
-formula = model.get_formula()
+formula = model.get_formula() # default: type='original' -> returns all formula without pruning | other options: 'sympied' -> simplified formula using sympy; 'latex' -> LaTeX format
 print("Symbolic Formula:", formula)
 
 # Get feature importances
@@ -162,13 +168,14 @@ model = OIKANClassifier(
     hidden_sizes=[32, 32], # Hidden layer sizes
     activation='relu', # Activation function (other options: 'tanh', 'leaky_relu', 'elu', 'swish', 'gelu')
     augmentation_factor=10, # Augmentation factor for data generation
-    polynomial_degree=2, # Degree of polynomial basis functions
-    alpha=0.1, # L1 regularization strength
+    alpha=0.1, # L1 regularization strength (Symbolic regression)
     sigma=0.1, # Standard deviation of Gaussian noise for data augmentation
+    top_k=5, # Number of top features to select (Symbolic regression)
     epochs=100, # # Number of training epochs
     lr=0.001, # Learning rate
     batch_size=32, # Batch size for training
-    verbose=True # Verbose output during training
+    verbose=True, # Verbose output during training
+    evaluate_nn=True # Validate neural network performance before full process
 )
 
 # Fit the model
@@ -182,7 +189,7 @@ accuracy = model.score(X_test, y_test)
 print("Accuracy:", accuracy)
 
 # Get symbolic formulas for each class
-formulas = model.get_formula()
+formulas = model.get_formula() # default: type='original' -> returns all formula without pruning | other options: 'sympied' -> simplified formula using sympy; 'latex' -> LaTeX format
 for i, formula in enumerate(formulas):
     print(f"Class {i} Formula:", formula)
    
@@ -204,6 +211,60 @@ loaded_model.load("outputs/model.json")
 
 ![OIKAN v0.0.3(1) Architecture](https://raw.githubusercontent.com/silvermete0r/oikan/main/docs/media/oikan-v0.0.3(1)-architecture-oop.png)
 
+## OIKAN Symbolic Model Compilers
+
+OIKAN provides a set of symbolic model compilers to convert the symbolic formulas generated by the OIKAN model into different programming languages. 
+
+*Currently, we support: `Python`, `C++`, `C`, `JavaScript`, `Rust`, and `Go`. This allows users to easily integrate the generated formulas into their applications or systems.*
+
+All compilers: [model_compilers/](model_compilers)
+
+### Example of Python Compiler
+
+1. Regression Model:
+```python
+import numpy as np
+import json
+
+def predict(X, symbolic_model):
+    X = np.asarray(X)
+    X_transformed = evaluate_basis_functions(X, symbolic_model['basis_functions'], 
+                                            symbolic_model['n_features'])
+    return np.dot(X_transformed, symbolic_model['coefficients'])
+
+if __name__ == "__main__":
+    with open('outputs/california_housing_model.json', 'r') as f:
+        symbolic_model = json.load(f)
+    X = np.random.rand(10, symbolic_model['n_features'])
+    y_pred = predict(X, symbolic_model)
+    print(y_pred)
+```
+
+2. Classification Model:
+```python
+import numpy as np
+import json
+
+def predict(X, symbolic_model):
+    X = np.asarray(X)
+    X_transformed = evaluate_basis_functions(X, symbolic_model['basis_functions'], 
+                                            symbolic_model['n_features'])
+    logits = np.dot(X_transformed, np.array(symbolic_model['coefficients_list']).T)
+    probabilities = np.exp(logits) / np.sum(np.exp(logits), axis=1, keepdims=True)
+    return np.argmax(probabilities, axis=1)
+
+if __name__ == "__main__":
+    with open('outputs/iris_model.json', 'r') as f:
+        symbolic_model = json.load(f)
+    X = np.array([[5.1, 3.5, 1.4, 0.2],
+                  [7.0, 3.2, 4.7, 1.4],
+                  [6.3, 3.3, 6.0, 2.5]])
+    y_pred = predict(X, symbolic_model)
+    print(y_pred)
+```
+
+
+
 ## Contributing
 
 We welcome contributions! Key areas of interest:

oikan-0.0.3.4.dist-info/RECORD

@@ -0,0 +1,10 @@
+oikan/__init__.py,sha256=zEzhm1GYLT4vNaIQ4CgZcNpUk3uo8SWnoaHYtHW_XSQ,628
+oikan/exceptions.py,sha256=GhHWqy2Q5LVBcteTy4ngnqxr7FOoLNyD8dNt1kfRXyw,901
+oikan/model.py,sha256=-EqCxTMeejdOCh2T08ibc87YIDYWUvybFe7jfb1XYbA,23167
+oikan/neural.py,sha256=wxmGgzmtpwJ3lvH6u6D4i4BiAzg018czrIdw49phSCY,1558
+oikan/utils.py,sha256=7UCm9obO-8Q2zhetdAkukMDOZvGSBWUL_dSF04XqM7k,8808
+oikan-0.0.3.4.dist-info/licenses/LICENSE,sha256=75ASVmU-XIpN-M4LbVmJ_ibgbzbvRLVti8FhnR0BTf8,1096
+oikan-0.0.3.4.dist-info/METADATA,sha256=P-2T0xtWDyTNhfZYknMK0TPm9EcvCUBD-O2bdfCwpFc,11138
+oikan-0.0.3.4.dist-info/WHEEL,sha256=DnLRTWE75wApRYVsjgc6wsVswC54sMSJhAEd4xhDpBk,91
+oikan-0.0.3.4.dist-info/top_level.txt,sha256=XwnwKwTJddZwIvtrUsAz-l-58BJRj6HjAGWrfYi_3QY,6
+oikan-0.0.3.4.dist-info/RECORD,,

{oikan-0.0.3.2.dist-info → oikan-0.0.3.4.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.3.1)
+Generator: setuptools (80.4.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

oikan-0.0.3.2.dist-info/RECORD

@@ -1,10 +0,0 @@
-oikan/__init__.py,sha256=zEzhm1GYLT4vNaIQ4CgZcNpUk3uo8SWnoaHYtHW_XSQ,628
-oikan/exceptions.py,sha256=Is0jG4apxO8QJQREIiJQYMjANYWibWeS-103q9KWbfg,192
-oikan/model.py,sha256=oZtx5uFxMj4q89ODKDBeTJ0whbqiMIR2tKwgmYVXHUY,16887
-oikan/neural.py,sha256=wxmGgzmtpwJ3lvH6u6D4i4BiAzg018czrIdw49phSCY,1558
-oikan/utils.py,sha256=xMGRa1qhn8BWn9UxpVeJIuGb-UvQmbjiFSsvAdF0bMU,2095
-oikan-0.0.3.2.dist-info/licenses/LICENSE,sha256=75ASVmU-XIpN-M4LbVmJ_ibgbzbvRLVti8FhnR0BTf8,1096
-oikan-0.0.3.2.dist-info/METADATA,sha256=PPsMSll3Ds6E9J3ZnXxo8Yno0ZsHrSb55kZ-035jJZE,8441
-oikan-0.0.3.2.dist-info/WHEEL,sha256=0CuiUZ_p9E4cD6NyLD6UG80LBXYyiSYZOKDm5lp32xk,91
-oikan-0.0.3.2.dist-info/top_level.txt,sha256=XwnwKwTJddZwIvtrUsAz-l-58BJRj6HjAGWrfYi_3QY,6
-oikan-0.0.3.2.dist-info/RECORD,,