摘要
针对液体中物质浓度预测模型,构建一种基于卷积神经网络的水质特征提取模型。首先,定义含有卷积层、采样层、全连接层的七层网络结构,选取适当的最优化方法和损失函数,对模型进行训练调整参数。接着分析了不同损失函数对模型训练和模型验证的影响。实验验证了在水质检测领域运用卷积神经网络回归的可行性。
In the light of the prediction model of mass concentration in liquid, a model of water quality feature extraction based on convolutional neural network was constructed. First of all, the model was trained and the parameters were adjusted by the definition of seven layer network structure including convolution layer, pooling sampling layer and fUlly connected layer and the selection of appropriate optimization method and loss function. Then, the influence of different loss functions on model tra-ining and model validation was analysed. Finally, the experiments verified the feasibility of applying the convolution neural net-work in the field of water quality detection.
出处
《信息通信》
2017年第12期61-63,共3页
Information & Communications
关键词
卷积神经网络回归
光谱分析
水质检测
损失函数
梯度下降
convolutional neural network regression
spectrum analysis
water qiiality monitoring
loss function
gradient descent
