R mxnet创建具有3x3隐藏层的3输入1输出回归

library(mxnet)

# 创建训练数据
train_data <- matrix(c(1, 2, 3, 4, 5, 6, 7, 8, 9), ncol = 3, byrow = TRUE)
train_label <- matrix(c(2, 4, 6), ncol = 1)

# 转换为mxnet数据格式
train_data <- mx.nd.array(train_data)
train_label <- mx.nd.array(train_label)

# 创建神经网络模型
model <- mx.symbol.Variable("data")
model <- mx.symbol.FullyConnected(data = model, num_hidden = 3, name = "fc1")
model <- mx.symbol.Activation(data = model, act_type = "relu")
model <- mx.symbol.FullyConnected(data = model, num_hidden = 3, name = "fc2")
model <- mx.symbol.Activation(data = model, act_type = "relu")
model <- mx.symbol.FullyConnected(data = model, num_hidden = 1, name = "fc3")

# 设置输出层
output <- mx.symbol.LinearRegressionOutput(data = model, label = "label")

# 创建模型迭代器
data_iter <- mx.io.arrayiter(data = train_data, label = train_label, batch.size = 1)

# 设置训练参数
model <- mx.model.FeedForward.create(output, X = data_iter, num.round = 10, array.batch.size = 1, learning.rate = 0.1, eval.metric = mx.metric.rmse)

# 进行预测
test_data <- mx.nd.array(matrix(c(10, 11, 12), ncol = 3, byrow = TRUE))
prediction <- predict(model, test_data)

print(prediction)