pyerualjetwork 4.2.0b1__py3-none-any.whl → 4.2.0b3__py3-none-any.whl

pyerualjetwork/__init__.py

@@ -48,7 +48,7 @@ for package_name in package_names:
 
 print(f"PyerualJetwork is ready to use with {err} errors")
 
-__version__ = "4.2.0b1"
+__version__ = "4.2.0b3"
 __update__ = "* Changes: https://github.com/HCB06/PyerualJetwork/blob/main/CHANGES\n* PyerualJetwork document: https://github.com/HCB06/PyerualJetwork/blob/main/Welcome_to_PyerualJetwork/PYERUALJETWORK_USER_MANUEL_AND_LEGAL_INFORMATION(EN).pdf\n* YouTube tutorials: https://www.youtube.com/@HasanCanBeydili"
 
 def print_version(__version__):

pyerualjetwork/planeat.py

@@ -248,6 +248,8 @@ def evolver(weights,
    else:
       raise ValueError("strategy parameter must be: 'normal_selective' or 'more_selective' or 'less_selective'")
    
+   if policy == 'explorer': fitness_bias = 0
+
    if ((activation_mutate_add_prob < 0 or activation_mutate_add_prob > 1) or 
       (activation_mutate_change_prob < 0 or activation_mutate_change_prob > 1) or 
       (activation_mutate_delete_prob < 0 or activation_mutate_delete_prob > 1) or 
@@ -303,15 +305,13 @@ def evolver(weights,
    normalized_fitness = abs(normalization(fitness, dtype=dtype))
 
    best_fitness = normalized_fitness[-1]
-
-   child_W = np.copy(bad_weights)
-   child_act = bad_activations.copy()
+   epsilon = np.finfo(float).eps
 
    for i in range(len(bad_weights)):
       second_parent_W, second_parent_act, s_i = second_parent_selection(good_weights, bad_weights, good_activations, bad_activations, bad_genomes_selection_prob)
 
       if policy == 'aggresive':
-         child_W[i], child_act[i] = cross_over(best_weights,
+         bad_weights[i], bad_activations[i] = cross_over(best_weights,
                                                          second_parent_W,
                                                          best_activations,
                                                          second_parent_act,
@@ -322,10 +322,11 @@ def evolver(weights,
                                                          bad_genomes_selection_prob=bad_genomes_selection_prob,
                                                          first_parent_fitness=best_fitness,
                                                          fitness_bias=fitness_bias,
-                                                         second_parent_fitness=normalized_fitness[s_i]
+                                                         second_parent_fitness=normalized_fitness[s_i],
+                                                         epsilon=epsilon
                                                          )
       elif policy == 'explorer':
-         child_W[i], child_act[i] = cross_over(good_weights[i],
+         bad_weights[i], bad_activations[i] = cross_over(good_weights[i],
                                                          second_parent_W, 
                                                          good_activations[i],
                                                          second_parent_act,
@@ -336,7 +337,8 @@ def evolver(weights,
                                                          bad_genomes_selection_prob=bad_genomes_selection_prob,
                                                          first_parent_fitness=normalized_fitness[i],
                                                          fitness_bias=fitness_bias,
-                                                         second_parent_fitness=normalized_fitness[s_i]
+                                                         second_parent_fitness=normalized_fitness[s_i],
+                                                         epsilon=epsilon
                                                          )
           
       else: raise ValueError("policy parameter must be: 'aggresive' or 'explorer'")
@@ -354,7 +356,8 @@ def evolver(weights,
                                                                activation_change_prob=activation_mutate_change_prob, 
                                                                weight_mutate_prob=weight_mutate_prob, 
                                                                threshold=weight_mutate_rate,
-                                                               genome_fitness=normalized_fitness[i]
+                                                               genome_fitness=normalized_fitness[i],
+                                                               epsilon=epsilon
                                                                )
             
          elif mutation_prob < bad_genomes_mutation_prob:
@@ -366,13 +369,14 @@ def evolver(weights,
                                                             activation_change_prob=activation_mutate_change_prob, 
                                                             weight_mutate_prob=weight_mutate_prob, 
                                                             threshold=weight_mutate_rate, 
-                                                            genome_fitness=normalized_fitness[i]
+                                                            genome_fitness=normalized_fitness[i],
+                                                            epsilon=epsilon
                                                             )
             
       if bar_status: progress.update(1)
 
-   weights = np.vstack((child_W, good_weights))
-   activation_potentiations = child_act + good_activations
+   weights = np.vstack((bad_weights, good_weights))
+   activation_potentiations = bad_activations + good_activations
 
    ### INFO PRINTING CONSOLE
    
@@ -492,7 +496,8 @@ def cross_over(first_parent_W,
                bad_genomes_selection_prob,
                first_parent_fitness,
                second_parent_fitness,
-               fitness_bias):
+               fitness_bias,
+               epsilon):
    """
     Performs a crossover operation on two sets of weights and activation functions.
     This function combines two individuals (represented by their weights and activation functions) 
@@ -500,22 +505,35 @@ def cross_over(first_parent_W,
 
     Args:
         first_parent_W (numpy.ndarray): The weight matrix of the first individual (parent).
+        
         second_parent_W (numpy.ndarray): The weight matrix of the second individual (parent).
+        
         first_parent_act (str or list): The activation function(s) of the first individual.
+        
         second_parent_act (str or list): The activation function(s) of the second individual.
+        
         cross_over_mode (str): Determines the crossover method to be used. Options:
             - 'tpm': Two-Point Matrix Crossover, where sub-matrices of weights are swapped between parents.
+        
         activation_selection_add_prob (float): Probability of adding new activation functions 
             from the second parent to the child genome.
+        
         activation_selection_change_prob (float): Probability of replacing an activation function in the child genome 
             with one from the second parent.
+        
         activation_selection_rate (float): Determines how quickly activation functions are added or replaced 
             during the crossover process.
+        
         bad_genomes_selection_prob (float): Probability of selecting a "bad" genome for replacement with the offspring.
+        
         first_parent_fitness (float): Fitness score of the first parent.
+        
         second_parent_fitness (float): Fitness score of the second parent.
+        
         fitness_bias (float): A bias factor used to favor fitter parents during crossover operations.
 
+        epsilon (float): Small epsilon constant
+
     Returns:
         tuple: A tuple containing:
             - child_W (numpy.ndarray): The weight matrix of the new individual created by crossover.
@@ -540,7 +558,8 @@ def cross_over(first_parent_W,
             bad_genomes_selection_prob=0.7,
             first_parent_fitness=0.9,
             second_parent_fitness=0.85,
-            fitness_bias=0.6
+            fitness_bias=0.6,
+            epsilon=np.finfo.eps
         )
         ```
     """
@@ -563,7 +582,7 @@ def cross_over(first_parent_W,
 
       undominant_parent_W = np.copy(second_parent_W)
       undominant_parent_act = second_parent_act
-      succes = second_parent_fitness
+      succes = second_parent_fitness + epsilon
 
    elif decision == 'second_parent':
       dominant_parent_W = np.copy(second_parent_W)
@@ -571,7 +590,7 @@ def cross_over(first_parent_W,
 
       undominant_parent_W = np.copy(first_parent_W)
       undominant_parent_act = first_parent_act
-      succes = first_parent_fitness
+      succes = first_parent_fitness + epsilon
 
    while True:
 
@@ -589,18 +608,17 @@ def cross_over(first_parent_W,
       selection_bias = random.uniform(0, 1)
 
       if fitness_bias > selection_bias:
-         row_cut_start = math.floor(row_cut_start * succes)
-         row_cut_end = math.ceil(row_cut_end * succes)
+         row_cut_start = math.floor(row_cut_start * (succes + epsilon))
+         row_cut_end = math.ceil(row_cut_end * (succes + epsilon))
 
-         col_cut_start = math.floor(col_cut_start * succes)
-         col_cut_end = math.ceil(col_cut_end * succes)
+         col_cut_start = math.floor(col_cut_start * (succes + epsilon))
+         col_cut_end = math.ceil(col_cut_end * (succes + epsilon))
 
    child_W = dominant_parent_W
 
    if cross_over_mode == 'tpm':
       child_W[row_cut_start:row_cut_end, col_cut_start:col_cut_end] = undominant_parent_W[row_cut_start:row_cut_end, col_cut_start:col_cut_end]
 
-
    if isinstance(dominant_parent_act, str): dominant_parent_act = [dominant_parent_act]
    if isinstance(undominant_parent_act, str): undominant_parent_act = [undominant_parent_act]
 
@@ -661,7 +679,8 @@ def mutation(weight,
              activation_change_prob, 
              weight_mutate_prob, 
              threshold, 
-             genome_fitness):
+             genome_fitness,
+             epsilon):
    """
     Performs mutation on the given weight matrix and activation functions.
     - The weight matrix is mutated by randomly changing its values based on the mutation probability.
@@ -669,14 +688,25 @@ def mutation(weight,
 
     Args:
         weight (numpy.ndarray): The weight matrix to mutate.
+        
         activations (list): The list of activation functions to mutate.
+        
         activation_mutate_prob (float): The overall probability of mutating activation functions.
+        
         activation_add_prob (float): Probability of adding a new activation function.
+        
         activation_delete_prob (float): Probability of removing an existing activation function.
+        
         activation_change_prob (float): Probability of replacing an existing activation function with a new one.
+        
         weight_mutate_prob (float): The probability of mutating weight matrix.
+        
         threshold (float): If the value you enter here is equal to the result of input layer * output layer, only a single weight will be mutated during each mutation process. If the value you enter here is half of the result of input layer * output layer, two weights in the weight matrix will be mutated.
+        
         genome_fitness (float): Fitness value of genome
+    
+        epsilon (float): Small epsilon constant
+        
     Returns:
         tuple: A tuple containing:
             - mutated_weight (numpy.ndarray): The weight matrix after mutation.
@@ -704,8 +734,8 @@ def mutation(weight,
       start = 0
       row_end = weight.shape[0]
       col_end = weight.shape[1]
-      
-      threshold = threshold * genome_fitness
+
+      threshold = threshold * (genome_fitness + epsilon)
       new_threshold = threshold
       
       while True:

pyerualjetwork/planeat_cuda.py

@@ -249,6 +249,8 @@ def evolver(weights,
    else:
       raise ValueError("strategy parameter must be: 'normal_selective' or 'more_selective' or 'less_selective'")
     
+   if policy =='explorer': fitness_bias = 0
+
    if ((activation_mutate_add_prob < 0 or activation_mutate_add_prob > 1) or 
       (activation_mutate_change_prob < 0 or activation_mutate_change_prob > 1) or 
       (activation_mutate_delete_prob < 0 or activation_mutate_delete_prob > 1) or 
@@ -303,15 +305,13 @@ def evolver(weights,
    normalized_fitness = abs(normalization(fitness, dtype=dtype))
 
    best_fitness = normalized_fitness[-1]
-
-   child_W = cp.copy(bad_weights)
-   child_act = bad_activations.copy()
+   epsilon = cp.finfo(float).eps
 
    for i in range(len(bad_weights)):
       second_parent_W, second_parent_act, s_i = second_parent_selection(good_weights, bad_weights, good_activations, bad_activations, bad_genomes_selection_prob)
 
       if policy == 'aggresive':
-         child_W[i], child_act[i] = cross_over(best_weights,
+         bad_weights[i], bad_activations[i] = cross_over(best_weights,
                                                          second_parent_W,
                                                          best_activations,
                                                          second_parent_act,
@@ -322,10 +322,11 @@ def evolver(weights,
                                                          bad_genomes_selection_prob=bad_genomes_selection_prob,
                                                          first_parent_fitness=best_fitness,
                                                          fitness_bias=fitness_bias,
-                                                         second_parent_fitness=normalized_fitness[s_i]
+                                                         second_parent_fitness=normalized_fitness[s_i],
+                                                         epsilon=epsilon
                                                          )
       elif policy == 'explorer':
-         child_W[i], child_act[i] = cross_over(good_weights[i],
+         bad_weights[i], bad_activations[i] = cross_over(good_weights[i],
                                                          second_parent_W, 
                                                          good_activations[i],
                                                          second_parent_act,
@@ -336,9 +337,10 @@ def evolver(weights,
                                                          bad_genomes_selection_prob=bad_genomes_selection_prob,
                                                          first_parent_fitness=normalized_fitness[i],
                                                          fitness_bias=fitness_bias,
-                                                         second_parent_fitness=normalized_fitness[s_i]
+                                                         second_parent_fitness=normalized_fitness[s_i],
+                                                         epsilon=epsilon
                                                          )
-          
+      
       else: raise ValueError("policy parameter must be: 'aggresive' or 'explorer'")
 
       if mutations is True:
@@ -354,7 +356,8 @@ def evolver(weights,
                                                                activation_change_prob=activation_mutate_change_prob, 
                                                                weight_mutate_prob=weight_mutate_prob, 
                                                                threshold=weight_mutate_rate,
-                                                               genome_fitness=normalized_fitness[i]
+                                                               genome_fitness=normalized_fitness[i],
+                                                               epsilon=epsilon
                                                                )
             
          elif mutation_prob < bad_genomes_mutation_prob:
@@ -366,13 +369,14 @@ def evolver(weights,
                                                             activation_change_prob=activation_mutate_change_prob, 
                                                             weight_mutate_prob=weight_mutate_prob, 
                                                             threshold=weight_mutate_rate, 
-                                                            genome_fitness=normalized_fitness[i]
+                                                            genome_fitness=normalized_fitness[i],
+                                                            epsilon=epsilon
                                                             )
             
       if bar_status: progress.update(1)
 
-   weights = cp.vstack((child_W, good_weights))
-   activation_potentiations = child_act + good_activations
+   weights = cp.vstack((bad_weights, good_weights))
+   activation_potentiations = bad_activations + good_activations
 
    ### INFO PRINTING CONSOLE
    
@@ -495,7 +499,8 @@ def cross_over(first_parent_W,
                bad_genomes_selection_prob,
                first_parent_fitness,
                second_parent_fitness,
-               fitness_bias):
+               fitness_bias,
+               epsilon):
    """
     Performs a crossover operation on two sets of weights and activation functions.
     This function combines two individuals (represented by their weights and activation functions) 
@@ -530,6 +535,8 @@ def cross_over(first_parent_W,
         
         fitness_bias (float): A bias factor used to favor fitter parents during crossover operations.
 
+        epsilon (float): Small epsilon constant
+
     Returns:
         tuple: A tuple containing:
             - child_W (numpy.ndarray): The weight matrix of the new individual created by crossover.
@@ -554,7 +561,8 @@ def cross_over(first_parent_W,
             bad_genomes_selection_prob=0.7,
             first_parent_fitness=0.9,
             second_parent_fitness=0.85,
-            fitness_bias=0.6
+            fitness_bias=0.6,
+            epsilon=cp.finfo.eps
         )
         ```
     """
@@ -577,7 +585,7 @@ def cross_over(first_parent_W,
 
       undominant_parent_W = cp.copy(second_parent_W)
       undominant_parent_act = second_parent_act
-      succes = second_parent_fitness
+      succes = second_parent_fitness + epsilon
 
    elif decision == 'second_parent':
       dominant_parent_W = cp.copy(second_parent_W)
@@ -585,7 +593,7 @@ def cross_over(first_parent_W,
 
       undominant_parent_W = cp.copy(first_parent_W)
       undominant_parent_act = first_parent_act
-      succes = first_parent_fitness
+      succes = first_parent_fitness + epsilon
 
    while True:
 
@@ -676,7 +684,8 @@ def mutation(weight,
              activation_change_prob, 
              weight_mutate_prob, 
              threshold, 
-             genome_fitness):
+             genome_fitness,
+             epsilon):
    """
     Performs mutation on the given weight matrix and activation functions.
     - The weight matrix is mutated by randomly changing its values based on the mutation probability.
@@ -701,6 +710,8 @@ def mutation(weight,
         
         genome_fitness (float): Fitness value of genome
     
+        epsilon (float): Small epsilon constant
+
     Returns:
         tuple: A tuple containing:
             - mutated_weight (numpy.ndarray): The weight matrix after mutation.
@@ -729,7 +740,7 @@ def mutation(weight,
       row_end = weight.shape[0]
       col_end = weight.shape[1]
       
-      threshold = threshold * genome_fitness
+      threshold = threshold * (genome_fitness + epsilon)
       new_threshold = threshold
       
       while True:

{pyerualjetwork-4.2.0b1.dist-info → pyerualjetwork-4.2.0b3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pyerualjetwork
-Version: 4.2.0b1
+Version: 4.2.0b3
 Summary: PyerualJetwork is a machine learning library written in Python for professionals, incorporating advanced, unique, new, and modern techniques.
 Author: Hasan Can Beydili
 Author-email: tchasancan@gmail.com

{pyerualjetwork-4.2.0b1.dist-info → pyerualjetwork-4.2.0b3.dist-info}/RECORD

@@ -1,4 +1,4 @@
-pyerualjetwork/__init__.py,sha256=UeD-cBkv904KlDncOWplEjPeNW9jPMkyjdaaTYJUDnY,2177
+pyerualjetwork/__init__.py,sha256=uTnwBPgD5ZsDGWn8CSyVs_DxNNdEbvJt0A1qrYmZtVM,2177
 pyerualjetwork/activation_functions.py,sha256=WWOdMd5pI6ZKe-ieKCIsKAYPQODHuXYxx7tzhA5xjes,11767
 pyerualjetwork/activation_functions_cuda.py,sha256=KmXJ5Cdig46XAMYakXFPEOlxSxtFJjD21-i3nGtxPjE,11807
 pyerualjetwork/data_operations.py,sha256=HjyW2QE18age6J8iG0jpbwqGOylL_nM-vE2CLbP9Wes,14690
@@ -13,12 +13,12 @@ pyerualjetwork/model_operations.py,sha256=hnhR8dtoICNJWIwGgJ65-LN3GYN_DYH4LMe6Yp
 pyerualjetwork/model_operations_cuda.py,sha256=XnKKq54ZLaqCm-NaJ6d8IToACKcKg2Ttq6moowVRRWo,13365
 pyerualjetwork/plan.py,sha256=EobwajGSIgbOujkzDKb-Kea0LGRHqpK3Xy1Le8VBAe8,34422
 pyerualjetwork/plan_cuda.py,sha256=iCcAHLzVw_VyjhkFHXzBWiedwbnpI1MCXNJgSDgZxWw,36065
-pyerualjetwork/planeat.py,sha256=--KO3cUAvUJ0L-qF2w2doHFX_UZvv0SsN-voPWJ7joY,41454
-pyerualjetwork/planeat_cuda.py,sha256=g-gdSAA-a1ML8GEUL3qOrSCd-XaFH7glkbqC0g_SC10,41655
+pyerualjetwork/planeat.py,sha256=nAVyCU78G00ac1cJ5XEsnmOnhMXuXCpbnS79pfaWqf8,42260
+pyerualjetwork/planeat_cuda.py,sha256=QorgaigpmtTSEjSvlfPy87jKH449Bve68AVOwRBYaiw,42212
 pyerualjetwork/ui.py,sha256=wu2BhU1k-w3Kcho5Jtq4SEKe68ftaUeRGneUOSCVDjU,575
 pyerualjetwork/visualizations.py,sha256=QaYSIyVkJZ8NqpBKArQKkI1y37nCQo_KIM98IMssnRc,28766
 pyerualjetwork/visualizations_cuda.py,sha256=F60vQ92AXlMgBka3InXnOtGoM25vQJAlBIU2AlYTwks,29200
-pyerualjetwork-4.2.0b1.dist-info/METADATA,sha256=ozbvSqNZWZ6eYKJVSRtkZHxHao8TuC0s2VHV53tsh2I,7795
-pyerualjetwork-4.2.0b1.dist-info/WHEEL,sha256=2wepM1nk4DS4eFpYrW1TTqPcoGNfHhhO_i5m4cOimbo,92
-pyerualjetwork-4.2.0b1.dist-info/top_level.txt,sha256=BRyt62U_r3ZmJpj-wXNOoA345Bzamrj6RbaWsyW4tRg,15
-pyerualjetwork-4.2.0b1.dist-info/RECORD,,
+pyerualjetwork-4.2.0b3.dist-info/METADATA,sha256=jPKsUsndSriMwL6FvZOQ2HThwHTATJ4eSHn-c63IqZ8,7795
+pyerualjetwork-4.2.0b3.dist-info/WHEEL,sha256=2wepM1nk4DS4eFpYrW1TTqPcoGNfHhhO_i5m4cOimbo,92
+pyerualjetwork-4.2.0b3.dist-info/top_level.txt,sha256=BRyt62U_r3ZmJpj-wXNOoA345Bzamrj6RbaWsyW4tRg,15
+pyerualjetwork-4.2.0b3.dist-info/RECORD,,