如何创建一个使用几个客户端的权重的FL算法？

import numpy as np

# 初始化全局模型
global_model = np.random.rand(10, 1)

# 客户端数据
client_data = [np.random.rand(100, 10) for _ in range(5)]
client_labels = [np.random.randint(0, 2, (100, 1)) for _ in range(5)]

# 客户端权重（这里简单地按数据量分配权重）
client_weights = [len(data) for data in client_data]
total_samples = sum(client_weights)
client_weights = [weight / total_samples for weight in client_weights]

# 本地训练函数
def local_training(model, data, labels):
    # 这里使用简单的线性回归作为示例
    predictions = np.dot(data, model)
    error = labels - predictions
    gradient = -2 * np.dot(data.T, error)
    model -= 0.01 * gradient
    return model

# 联邦学习迭代
for round in range(10):
    local_models = []
    for i in range(len(client_data)):
        local_model = local_training(global_model.copy(), client_data[i], client_labels[i])
        local_models.append(local_model)
    
    # 加权聚合
    weighted_updates = np.zeros_like(global_model)
    for i in range(len(local_models)):
        weighted_updates += client_weights[i] * (local_models[i] - global_model)
    
    global_model += weighted_updates

print("Final Global Model:", global_model)