Table 3 The architecture of BPNN and model parameter selection for weight learning function and node transfer function
From: Evaluating soil nutrients of Dacrydium pectinatum in China using machine learning techniques
Model parameter | Parameter quantity |
|---|---|
Training function | Tranilm |
Learning function | Learndm |
Performance function | MSE |
Hidden layer transfer function | Tansig |
Output layer transfer function | Purelin |
Number of neural elements nodes in input layer | 5 |
Number of nodes in output layer neural units | 1 |
Learning rate | 0.4 |
Momentum coefficient | 0.8 |
Iteration times | ≤ 50,000 |
Network convergence error | ≤ 0.05 |
Inertia factor | 0.5 |
Training target error | 0.001 |
Initial weight | [0 + 0.5] |
Learning coefficient | 0.05 |