1 Algorithm Introduction
From the name, we can see that the BP neural network can be divided into two parts: bp and neural network. Here, bp is the abbreviation for Back Propagation, which means reverse propagation.
The BP network can learn and store a large number of input-output mapping relationships without needing to reveal the mathematical equations describing these relationships beforehand. Its learning rule uses the steepest descent method to continuously adjust the network’s weights and thresholds through backpropagation, minimizing the sum of squared errors. Its main characteristics are: signals are propagated forward, while errors are propagated backward.
For example, a manufacturer produces a product, and after launching it into the market, receives feedback from consumers. Based on this feedback, the manufacturer further upgrades and optimizes the product, repeating this process until achieving the ultimate goal—producing a product that satisfies consumers. The product launch is the “signal forward propagation,” while consumer feedback is the “error backward propagation.” This is the essence of the BP neural network.
2 Algorithm Principles
The workflow of the BP neural network algorithm is as follows: first, input samples are provided to the input layer neurons, then the signals are propagated forward layer by layer until the output layer produces results; then the output layer’s errors are calculated, and the errors are propagated backward to the hidden layer neurons. Finally, the connection weights and thresholds are adjusted based on the errors of the hidden layer neurons. This iterative process continues until the stopping condition is met.
The BP neural network is a feedforward neural network, structured as follows:
1. Network Layers: It consists of three main functional layers: input layer, hidden layer, and output layer, where the hidden layer can contain multiple sub-layers. 2. Number of Neurons: The number of neurons in the input layer equals the number of inputs; the number of neurons in the hidden layer is set by the user; the number of neurons in the output layer equals the number of outputs. 3. Network Weights: Each neuron in one layer is connected to the neurons in the next layer with weights. 4. Neural Network Threshold: Each neuron in the hidden layer and output layer has its own threshold. 5. Activation Function: Each neuron in the hidden layer and output layer has its own activation function.
The above is just a general structural form. When using it, we need to set specific parameters to give it a concrete structure, such as the number of hidden layers and the number of hidden neurons in each layer. Although the general structure allows for multiple hidden layers and different activation functions, in practice, the most common setup is one hidden layer, with the activation function of the hidden layer neurons set to the tansig function and the activation function of the output layer neurons set to the purelin function. Thus, the structure is as shown in the following figure:
3 Algorithm Applications
Regression prediction (can perform fitting, data processing analysis, event prediction, control, etc.), classification recognition (for type classification, pattern recognition, etc.). However, regardless of the network or method used, the accuracy of problem-solving cannot reach 100%, but this does not affect its use because many complex problems in reality have no meaningful precise explanation; meaningful analysis will inevitably lose precision.
In the field of traditional Chinese medicine, the BP neural network can be used for correlation analysis between traditional Chinese medicine syndromes, symptoms, and Western medical indicators, quantitative description of acupuncture efficacy, analysis of electrical signal characteristics of acupoints, and improving the accuracy of acupuncture robot’s acupoint searching behavior. For example, in the research on constructing a traditional Chinese medicine intelligent syndrome differentiation model for perimenopausal syndrome, the researchers built a neural network classifier by stacking layers of the torch.nn module. The neural network includes 1 input layer, 2 hidden layers, and 1 output layer. The activation function chosen between the input layer and hidden layers, as well as between the hidden layers, is the Relu activation function, while the output layer uses the Softmax function for multi-class output, with the training iteration period set to 50, and other hyperparameters chosen for optimal combinations. In each training cycle (epoch), the training batch is obtained through the Data Loader method, using the cross-entropy loss function as the model’s learning strategy, and the Adam optimizer (Adaptive Moment Estimation) is chosen for optimization. The final model results are shown in the following figures.
4 Conclusion
The BP neural network is widely used in the field of traditional Chinese medicine research, involving information analysis and syndrome identification of the four diagnostic methods of TCM, intelligent prescription selection and efficacy prediction of prescriptions, optimization, identification, and quality control of traditional Chinese medicine processes, and more. Because the BP neural network has a certain fault tolerance for the types and distributions of research data, it can gradually achieve complex mapping relationships between input and output variables through self-learning and self-adjustment. Therefore, for some data that traditional statistical methods cannot fully adapt to, applying BP neural networks can also facilitate objective analysis.
This ability to handle various data types with minimal restrictions, along with its proficiency in dealing with fuzzy, nonlinear, and noisy data characteristics, combined with its capability to utilize all available information for “learning” and “adaptation,” gives the BP neural network a certain advantage in data mining within the field of traditional Chinese medicine research. Artificial intelligence algorithms like the BP neural network will undoubtedly empower the great era of traditional Chinese medicine, promoting the informatization and intelligent development of TCM, thereby supporting the inheritance and development of the TCM health industry.
References:
[1] Yao Shuaijun, Yan Jinglai, Du Caifeng, et al. Research on the Construction of an Intelligent Syndrome Differentiation Model for Perimenopausal Syndrome Based on Machine Learning [J]. China Journal of Traditional Chinese Medicine Information, 2023, 30(06):68-75. DOI:10.19879/j.cnki.1005-5304.202207558.
[2] Fu Shufei, Liu Hongyan, Ren Haiyan, et al. Application of BP Neural Network in the Field of Traditional Chinese Medicine Research [J]. Medical Information, 2021, 34(12):12-14+18.
[3] Zhihu Column. “Understanding BP Neural Network in One Article—From Principles to Applications.” Accessed on November 22, 2023. https://zhuanlan.zhihu.com/p/485348369.
[4] Zhihu Column. “Characteristics, Applications, and Defects of BP Neural Network.” Accessed on November 22, 2023. https://zhuanlan.zhihu.com/p/536622739.
[5] Zhihu Column. “An Introduction to BP Neural Network Model.” Accessed on November 22, 2023. https://zhuanlan.zhihu.com/p/620327513.
Reinforcement Learning—Enabling Machines to Learn Autonomously
Latent Dirichlet Allocation (LDA)—Allocation, Distribution, and Classification
Recurrent Neural Network (RNN)—Neural Networks with Memory Functions
Ancient and Modern Medical Case Cloud Platform
Providing search services for over 500,000 ancient and modern medical cases
Supports manual, voice, OCR, and batch structured entry of medical cases
Designs nine major analysis modules, closely aligned with clinical practical needs
Supports collaborative analysis of massive medical cases and personal cases on the platform
EDC Traditional Chinese Medicine Research Case Collection System
Supports multi-center, online random grouping, data entry
SDV, inspection trails, SMS reminders, data statistics
Analysis and other functions
Supports customized form design
Users can log in at: https://www.yiankb.com/edc
Free trial!
Institute of Traditional Chinese Medicine Information, China Academy of Chinese Medical Sciences
Intelligent R&D Center for Traditional Chinese Medicine Health
Big Data R&D Department
Phone: 010-64089619
13522583261
QQ: 2778196938
https://www.yiankb.com
