Build final classification layer if needed

r"""Calculate the Area Under the Receiver Operating Characteristic Curve (ROC AUC)

r"""Calculate the F1 score, a weighted average of precision and recall

r"""Calculate the precision, the ability of the classifier not to label a sample as positive if it is negative

r"""Calculate the recall, the ability of the classifier to find all the positive samples

r"""Method to save the training process figure. Args: loss: Training loss. accuracy: Training accuracy. p

Recursively compare two outputs, supporting nested tuples/lists.

Recursively compare two outputs, supporting nested tuples/lists.

r"""Method to compute fitness based on the function, predictions, and labels. Args: func: Symbolic expression. pred:

r"""Method to perform mutation on a selected tree structure. Args: selected: The selected offspring. Returns:

r"""Method to perform crossover between two selected tree structures. Args: selected1: The selected tree structures 1.

Return the index of input function name.

基于 'provider/model' 创建具体客户端。 必填:config['model'](形如 'provider/model')。 选填:config['api_key']、config['base_url']。

Return the best discovered neuron formula.

r"""Method to generate offspring based on the given population and fitness. Args: popula: The given population. ftns:

Initialize factor matrices for CP/Tucker decomposition

解析模型字符串为 (provider, model)。 规则:第一个 '/' 之前为提供商(大小写不敏感),之后的全部为模型名(大小写敏感,允许包含 '/'). 示例: - "deepseek/deepseek-chat" -> ("deepseek", "dee

r"""Method to update the progress plot during training. Args: loss: Training loss. accuracy: Training accuracy.

r"""Method to update the progress plot during training. Args: loss: Training loss.

r"""Prepares the input data and splits it into training and validation sets. Args: X (numpy ndarray): Training data.

r"""Prepares the input data and splits it into training and validation sets. Args: X (numpy ndarray): Training data.

r"""Set the random seed for reproducibility of experiments. Args: seed: Random seed.

r"""Method to save the training process figure. Args: loss: Training loss. path: Path to save the figure.

r"""Method to render a program based on a given tree structure. Recursively converts the tree into a string expression. Args:

Kaiming 初始化权重，均匀初始化偏置。

r"""Initialize weights and bias using Kaiming uniform initialization.

Parameter initialisation (handled internally by TNLinear).

r"""Selects the training device based on the 'gpu' parameter. Args: gpu: GPU ID.

r"""Selects the training device based on the 'gpu' parameter. Args: gpu: GPU ID.

return a ResNet 101 object

Perform one training step on the model, save and reload it, then verify outputs and parameters.

Return Program object by parsing input text using Python AST.

Training loop with regularization

r"""Construct MLPClassifier with task-based neurons. Args: neurons (str): Neuronal expression layers_list (list): List of neuro

r"""Initialize the preprocessor with options for numerical and categorical feature processing. Args: num_features: Numerical feat

r"""Initialization method of the BaseModel class that sets up a visualization tool

r"""Initialization method of the BaseModel class that sets up a visualization tool

r""" Construct MLPRegressor with task-based neurons. Args: neurons (str): Neuronal expression layers_list (list): L

r"""Use vectorized symbolic regression algorithm to generate neuronal expression. Args: random_state (int): Seed for random numbe