pyerualjetwork 4.3.7b3__py3-none-any.whl → 4.3.8__py3-none-any.whl

pyerualjetwork/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "4.3.7b3"
+__version__ = "4.3.8"
 __update__ = "* Changes: https://github.com/HCB06/PyerualJetwork/blob/main/CHANGES\n* PyerualJetwork Homepage: https://github.com/HCB06/PyerualJetwork/tree/main\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/activation_functions.py

@@ -217,8 +217,8 @@ def scaled_cubic(x, alpha=1.0):
 
 def sine_offset(x, beta=0.0):
     return np.sin(x + beta)
-    
-    
+
+
 def apply_activation(Input, activation_list):
     """
     Applies activation functions for inputs

pyerualjetwork/activation_functions_cuda.py

@@ -217,124 +217,73 @@ def scaled_cubic(x, alpha=1.0):
 def sine_offset(x, beta=0.0):
     return cp.sin(x + beta)
 
-
-def safe_add(current_sum, new_value):
-        try:
-            return current_sum + new_value
-        except OverflowError:
-            return cp.array(current_sum) + cp.array(new_value)
     
 def apply_activation(Input, activation_list):
     """
-        Applies a sequence of activation functions to the input.
-        
-        Args:
-            Input (cupy.ndarray): The input to apply activations to.
-            activation_list (list): A list of activation function names to apply.
-        
-        Returns:
-            cupy.ndarray: The input after all activations have been applied.
-    """
+    Applies activation functions for inputs
     
+    Args:
+        Input (cupy.ndarray):
+        activation_list (list):
+    """
     origin_input = cp.copy(Input)
+    
+    activation_functions = {
+        'sigmoid': Sigmoid,
+        'swish': swish,
+        'mod_circular': modular_circular_activation,
+        'tanh_circular': tanh_circular_activation,
+        'leaky_relu': leaky_relu,
+        'relu': Relu,
+        'softplus': softplus,
+        'elu': elu,
+        'gelu': gelu,
+        'selu': selu,
+        'tanh': tanh,
+        'sinakt': sinakt,
+        'p_squared': p_squared,
+        'sglu': lambda x: sglu(x, alpha=1.0),
+        'dlrelu': dlrelu,
+        'exsig': exsig,
+        'sin_plus': sin_plus,
+        'acos': lambda x: acos(x, alpha=1.0, beta=0.0),
+        'gla': lambda x: gla(x, alpha=1.0, mu=0.0),
+        'srelu': srelu,
+        'qelu': qelu,
+        'isra': isra,
+        'waveakt': waveakt,
+        'arctan': arctan,
+        'bent_identity': bent_identity,
+        'sech': sech,
+        'softsign': softsign,
+        'pwl': pwl,
+        'cubic': cubic,
+        'gaussian': gaussian,
+        'sine': sine,
+        'tanh_square': tanh_square,
+        'mod_sigmoid': mod_sigmoid,
+        'linear': lambda x: x,
+        'quartic': quartic,
+        'square_quartic': square_quartic,
+        'cubic_quadratic': cubic_quadratic,
+        'exp_cubic': exp_cubic,
+        'sine_square': sine_square,
+        'logarithmic': logarithmic,
+        'scaled_cubic': lambda x: scaled_cubic(x, 1.0),
+        'sine_offset': lambda x: sine_offset(x, 1.0),
+        'spiral': spiral_activation,
+        'circular': circular_activation
+    }
+    
+    try:
+        
+        valid_mask = cp.array([act in activation_functions for act in activation_list])
+        valid_activations = cp.array(activation_list)[valid_mask]
+
+        activation_outputs = cp.array([activation_functions[act](origin_input) for act in valid_activations])
 
-    for i in range(len(activation_list)):
-        try:
-            if activation_list[i] == 'sigmoid':
-                Input = safe_add(Input, Sigmoid(origin_input))
-            elif activation_list[i] == 'swish':
-                Input = safe_add(Input, swish(origin_input))
-            elif activation_list[i] == 'mod_circular':
-                Input = safe_add(Input, modular_circular_activation(origin_input))
-            elif activation_list[i] == 'tanh_circular':
-                Input = safe_add(Input, tanh_circular_activation(origin_input))
-            elif activation_list[i] == 'leaky_relu':
-                Input = safe_add(Input, leaky_relu(origin_input))
-            elif activation_list[i] == 'relu':
-                Input = safe_add(Input, Relu(origin_input))
-            elif activation_list[i] == 'softplus':
-                Input = safe_add(Input, softplus(origin_input))
-            elif activation_list[i] == 'elu':
-                Input = safe_add(Input, elu(origin_input))
-            elif activation_list[i] == 'gelu':
-                Input = safe_add(Input, gelu(origin_input))
-            elif activation_list[i] == 'selu':
-                Input = safe_add(Input, selu(origin_input))
-            elif activation_list[i] == 'tanh':
-                Input = safe_add(Input, tanh(origin_input))
-            elif activation_list[i] == 'sinakt':
-                Input = safe_add(Input, sinakt(origin_input))
-            elif activation_list[i] == 'p_squared':
-                Input = safe_add(Input, p_squared(origin_input))
-            elif activation_list[i] == 'sglu':
-                Input = safe_add(Input, sglu(origin_input, alpha=1.0))
-            elif activation_list[i] == 'dlrelu':
-                Input = safe_add(Input, dlrelu(origin_input))
-            elif activation_list[i] == 'exsig':
-                Input = safe_add(Input, exsig(origin_input))
-            elif activation_list[i] == 'sin_plus':
-                Input = safe_add(Input, sin_plus(origin_input))
-            elif activation_list[i] == 'acos':
-                Input = safe_add(Input, acos(origin_input, alpha=1.0, beta=0.0))
-            elif activation_list[i] == 'gla':
-                Input = safe_add(Input, gla(origin_input, alpha=1.0, mu=0.0))
-            elif activation_list[i] == 'srelu':
-                Input = safe_add(Input, srelu(origin_input))
-            elif activation_list[i] == 'qelu':
-                Input = safe_add(Input, qelu(origin_input))
-            elif activation_list[i] == 'isra':
-                Input = safe_add(Input, isra(origin_input))
-            elif activation_list[i] == 'waveakt':
-                Input = safe_add(Input, waveakt(origin_input))
-            elif activation_list[i] == 'arctan':
-                Input = safe_add(Input, arctan(origin_input))
-            elif activation_list[i] == 'bent_identity':
-                Input = safe_add(Input, bent_identity(origin_input))
-            elif activation_list[i] == 'sech':
-                Input = safe_add(Input, sech(origin_input))
-            elif activation_list[i] == 'softsign':
-                Input = safe_add(Input, softsign(origin_input))
-            elif activation_list[i] == 'pwl':
-                Input = safe_add(Input, pwl(origin_input))
-            elif activation_list[i] == 'cubic':
-                Input = safe_add(Input, cubic(origin_input))
-            elif activation_list[i] == 'gaussian':
-                Input = safe_add(Input, gaussian(origin_input))
-            elif activation_list[i] == 'sine':
-                Input = safe_add(Input, sine(origin_input))
-            elif activation_list[i] == 'tanh_square':
-                Input = safe_add(Input, tanh_square(origin_input))
-            elif activation_list[i] == 'mod_sigmoid':
-                Input = safe_add(Input, mod_sigmoid(origin_input))
-            elif activation_list[i] == 'linear':
-                Input = safe_add(Input, origin_input)
-            elif activation_list[i] == 'quartic':
-                Input = safe_add(Input, quartic(origin_input))
-            elif activation_list[i] == 'square_quartic':
-                Input = safe_add(Input, square_quartic(origin_input))
-            elif activation_list[i] == 'cubic_quadratic':
-                Input = safe_add(Input, cubic_quadratic(origin_input))
-            elif activation_list[i] == 'exp_cubic':
-                Input = safe_add(Input, exp_cubic(origin_input))
-            elif activation_list[i] == 'sine_square':
-                Input = safe_add(Input, sine_square(origin_input))
-            elif activation_list[i] == 'logarithmic':
-                Input = safe_add(Input, logarithmic(origin_input))
-            elif activation_list[i] == 'scaled_cubic':
-                Input = safe_add(Input, scaled_cubic(origin_input, 1.0))
-            elif activation_list[i] == 'sine_offset':
-                Input = safe_add(Input, sine_offset(origin_input, 1.0))
-            elif activation_list[i] == 'spiral':
-                Input = safe_add(Input, spiral_activation(origin_input))
-            elif activation_list[i] == 'circular':
-                Input = safe_add(Input, circular_activation(origin_input))
+        return Input + cp.sum(activation_outputs, axis=0)
         
-        except Exception as e:
-            warnings.warn(f"Error in activation {activation_list[i]}: {str(e)}", RuntimeWarning)
-            if not isinstance(Input, cp.ndarray):
-                Input = cp.array(Input)
-            if not isinstance(origin_input, cp.ndarray):
-                origin_input = cp.array(origin_input)
-            continue
- 
-    return Input
+    except Exception as e:
+        warnings.warn(f"Error in activation processing: {str(e)}", RuntimeWarning)
+        return Input

pyerualjetwork/data_operations_cuda.py

@@ -401,7 +401,7 @@ def standard_scaler(x_train=None, x_test=None, scaler_params=None, dtype=cp.floa
 
         return scaled_data  # sample data scaled
     
-    
+
 def normalization(
     Input,  # num: Input data to be normalized.
     dtype=cp.float32

pyerualjetwork/model_operations.py

@@ -258,7 +258,7 @@ def predict_model_ssd(Input, model_name, model_path='', dtype=np.float32):
     ndarray: Output from the model.
     """
     
-    from .plan import feed_forward
+    from .activation_functions import apply_activation
     from .data_operations import standard_scaler
     
     model = load_model(model_name, model_path)
@@ -269,12 +269,12 @@ def predict_model_ssd(Input, model_name, model_path='', dtype=np.float32):
 
     Input = standard_scaler(None, Input, scaler_params, dtype=dtype)
 
-    neural_layer = Input
-    neural_layer = np.array(neural_layer, copy=False)
-    neural_layer = neural_layer.ravel()
+    Input = np.array(Input, dtype=dtype, copy=False)
+    Input = Input.ravel()
 
     try:
-        neural_layer = feed_forward(neural_layer, np.copy(W.astype(dtype, copy=False)), is_training=False, Class='?', activation_potentiation=activation_potentiation)
+        Input = apply_activation(Input, activation_potentiation)
+        neural_layer = Input @ W.T
         return neural_layer
     except:
         print(Fore.RED + "ERROR: Unexpected Output or wrong model parameters from: predict_model_ssd." + Style.RESET_ALL)
@@ -304,7 +304,7 @@ def reverse_predict_model_ssd(output, model_name, model_path='', dtype=np.float3
     W = model[get_weights()]
 
     try:
-        Input = np.dot(output.astype(dtype, copy=False), np.copy(W.astype(dtype, copy=False)))
+        Input = W.T @ output
         return Input
     except:
         print(Fore.RED + "ERROR: Unexpected Output or wrong model parameters from: reverse_predict_model_ssd." + Style.RESET_ALL)
@@ -334,18 +334,18 @@ def predict_model_ram(Input, W, scaler_params=None, activation_potentiation=['li
     ndarray: Output from the model.
     """
 
-    from data_operations import standard_scaler
-    from plan import feed_forward
+    from .data_operations import standard_scaler
+    from .activation_functions import apply_activation
 
     Input = standard_scaler(None, Input, scaler_params, dtype=dtype)
     
+    Input = np.array(Input, dtype=dtype, copy=False)
+    Input = Input.ravel()
+
     try:
         
-        neural_layer = Input
-        neural_layer = np.array(neural_layer, copy=False)
-        neural_layer = neural_layer.ravel()
-
-        neural_layer = feed_forward(neural_layer, np.copy(W.astype(dtype, copy=False)), is_training=False, Class='?', activation_potentiation=activation_potentiation)
+        Input = apply_activation(Input, activation_potentiation)
+        neural_layer = Input @ W.T
 
         return neural_layer
     
@@ -371,7 +371,7 @@ def reverse_predict_model_ram(output, W, dtype=np.float32):
     """
 
     try:
-        Input = np.dot(output.astype(dtype, copy=False), np.copy(W.astype(dtype, copy=False)))
+        Input = W.T @ output
         return Input
     
     except:

pyerualjetwork/model_operations_cuda.py

@@ -263,10 +263,10 @@ def predict_model_ssd(Input, model_name, model_path='', dtype=cp.float32):
     Returns:
     ndarray: Output from the model.
     """
-            
+    
     Input = cp.array(Input, dtype=dtype, copy=False)
-        
-    from .plan_cuda import feed_forward
+    
+    from .activation_functions_cuda import apply_activation
     from .data_operations_cuda import standard_scaler
     
     model = load_model(model_name, model_path)
@@ -277,13 +277,14 @@ def predict_model_ssd(Input, model_name, model_path='', dtype=cp.float32):
 
     Input = standard_scaler(None, Input, scaler_params)
 
-    neural_layer = Input
-    neural_layer = cp.array(neural_layer)
-    neural_layer = neural_layer.ravel()
+    Input = cp.array(Input, dtype=dtype, copy=False)
+    Input = Input.ravel()
 
     try:
-        neural_layer = feed_forward(neural_layer, cp.copy(W), is_training=False, Class='?', activation_potentiation=activation_potentiation)
+        Input = apply_activation(Input, activation_potentiation)
+        neural_layer = Input @ W.T
         return neural_layer
+
     except:
         print(Fore.RED + "ERROR: Unexpected Output or wrong model parameters from: predict_model_ssd." + Style.RESET_ALL)
         sys.exit()
@@ -314,7 +315,7 @@ def reverse_predict_model_ssd(output, model_name, model_path='', dtype=cp.float3
     W = model[get_weights()]
 
     try:
-        Input = cp.dot(output, cp.copy(W))
+        Input = W.T @ output
         return Input
     except:
         print(Fore.RED + "ERROR: Unexpected Output or wrong model parameters from: reverse_predict_model_ssd." + Style.RESET_ALL)
@@ -344,19 +345,17 @@ def predict_model_ram(Input, W, scaler_params=None, activation_potentiation=['li
     """
 
     from .data_operations_cuda import standard_scaler
-    from .plan_cuda import feed_forward
+    from .activation_functions_cuda import apply_activation
 
-    Input = cp.array(Input, dtype=dtype, copy=False)
-    
     Input = standard_scaler(None, Input, scaler_params)
     
+    Input = cp.array(Input, dtype=dtype, copy=False)
+    Input = Input.ravel()
+
     try:
         
-        neural_layer = Input
-        neural_layer = cp.array(neural_layer)
-        neural_layer = neural_layer.ravel()
-
-        neural_layer = feed_forward(neural_layer, cp.copy(W), is_training=False, Class='?', activation_potentiation=activation_potentiation)
+        Input = apply_activation(Input, activation_potentiation)
+        neural_layer = Input @ W.T
 
         return neural_layer
     
@@ -384,7 +383,7 @@ def reverse_predict_model_ram(output, W, dtype=cp.float32):
     output = cp.array(output, dtype=dtype, copy=False)
 
     try:
-        Input = cp.dot(output, cp.copy(W))
+        Input = W.T @ output
         return Input
     
     except:

pyerualjetwork/parallel.py

@@ -0,0 +1,118 @@
+import multiprocessing as mp
+from functools import partial
+import random
+from .planeat import mutation, cross_over, second_parent_selection
+
+def run_process(policy, best_weight, best_activations, good_weights, 
+                good_activations, bad_weights, bad_activations, best_fitness, 
+                normalized_fitness, child_W, child_act, mutated_W, mutated_act,
+                cross_over_mode, activation_selection_add_prob, 
+                activation_selection_change_prob, activation_selection_threshold, 
+                bad_genomes_selection_prob, fitness_bias, epsilon, 
+                bad_genomes_mutation_prob, activation_mutate_prob, 
+                activation_mutate_add_prob, activation_mutate_delete_prob, 
+                activation_mutate_change_prob, weight_mutate_prob, 
+                weight_mutate_threshold, activation_mutate_threshold):
+    
+    process_func = partial(process_single, policy=policy, best_weight=best_weight, 
+                           best_activations=best_activations, good_weights=good_weights, 
+                           good_activations=good_activations, bad_weights=bad_weights, 
+                           bad_activations=bad_activations, best_fitness=best_fitness, 
+                           normalized_fitness=normalized_fitness, child_W=child_W, 
+                           child_act=child_act, mutated_W=mutated_W, mutated_act=mutated_act,
+                           cross_over_mode=cross_over_mode,
+                           activation_selection_add_prob=activation_selection_add_prob,
+                           activation_selection_change_prob=activation_selection_change_prob,
+                           activation_selection_threshold=activation_selection_threshold,
+                           bad_genomes_selection_prob=bad_genomes_selection_prob,
+                           fitness_bias=fitness_bias,
+                           epsilon=epsilon,
+                           bad_genomes_mutation_prob=bad_genomes_mutation_prob,
+                           activation_mutate_prob=activation_mutate_prob,
+                           activation_mutate_add_prob=activation_mutate_add_prob,
+                           activation_mutate_delete_prob=activation_mutate_delete_prob,
+                           activation_mutate_change_prob=activation_mutate_change_prob,
+                           weight_mutate_prob=weight_mutate_prob,
+                           weight_mutate_threshold=weight_mutate_threshold,
+                           activation_mutate_threshold=activation_mutate_threshold)
+
+    with mp.Pool() as pool:
+        results = pool.map(process_func, range(len(bad_weights)))
+
+    for i, new_child_W, new_child_act, new_mutated_W, new_mutated_act in results:
+        child_W[i] = new_child_W
+        child_act[i] = new_child_act
+        mutated_W[i] = new_mutated_W
+        mutated_act[i] = new_mutated_act
+
+
+    return child_W, child_act, mutated_W, mutated_act
+
+
+def process_single(i, policy, best_weight, best_activations, 
+                   good_weights, good_activations, 
+                   bad_weights, bad_activations, 
+                   best_fitness, normalized_fitness, 
+                   cross_over_mode, 
+                   activation_selection_add_prob, activation_selection_change_prob,
+                   activation_selection_threshold, 
+                   bad_genomes_selection_prob, fitness_bias,
+                   epsilon, bad_genomes_mutation_prob, 
+                   activation_mutate_prob, activation_mutate_add_prob,
+                   activation_mutate_delete_prob, activation_mutate_change_prob,
+                   weight_mutate_prob, weight_mutate_threshold,
+                   activation_mutate_threshold):
+
+    if policy == 'aggressive':
+        first_parent_W = best_weight
+        first_parent_act = best_activations
+    elif policy == 'explorer':
+        first_parent_W = good_weights[i]
+        first_parent_act = good_activations[i]
+    else:
+        raise ValueError("policy parameter must be: 'aggressive' or 'explorer'")
+        
+    second_parent_W, second_parent_act, s_i = second_parent_selection(
+        good_weights, bad_weights, good_activations, bad_activations, bad_genomes_selection_prob)
+
+    new_child_W, new_child_act = cross_over(
+        first_parent_W,
+        second_parent_W,
+        first_parent_act,
+        second_parent_act,
+        cross_over_mode=cross_over_mode,
+        activation_selection_add_prob=activation_selection_add_prob,
+        activation_selection_change_prob=activation_selection_change_prob,
+        activation_selection_threshold=activation_selection_threshold,
+        bad_genomes_selection_prob=bad_genomes_selection_prob,
+        first_parent_fitness=best_fitness,
+        fitness_bias=fitness_bias,
+        second_parent_fitness=normalized_fitness[s_i],
+        epsilon=epsilon
+    )
+
+    mutation_prob = random.uniform(0, 1)
+    if mutation_prob > bad_genomes_mutation_prob:
+        genome_W = good_weights[i]
+        genome_act = good_activations[i]
+        fitness_index = int(len(bad_weights) / 2 + i)
+    else:
+        genome_W = bad_weights[i]
+        genome_act = bad_activations[i]
+        fitness_index = i
+
+    new_mutated_W, new_mutated_act = mutation(
+        genome_W,
+        genome_act,
+        activation_mutate_prob=activation_mutate_prob,
+        activation_add_prob=activation_mutate_add_prob,
+        activation_delete_prob=activation_mutate_delete_prob,
+        activation_change_prob=activation_mutate_change_prob,
+        weight_mutate_prob=weight_mutate_prob,
+        weight_mutate_threshold=weight_mutate_threshold,
+        genome_fitness=normalized_fitness[fitness_index],
+        activation_mutate_threshold=activation_mutate_threshold,
+        epsilon=epsilon
+    )
+    
+    return (i, new_child_W, new_child_act, new_mutated_W, new_mutated_act)