Skip to content

Latest commit

 

History

History
33 lines (28 loc) · 1.68 KB

README.md

File metadata and controls

33 lines (28 loc) · 1.68 KB

Modeling the behavior of Beet’s seeds in budding period with neural networks

Problem Statement

The objective of this research is to study the effect of different variables on the budding process of the beet plant. The variables are 11 different types of fertilizers. The land considered for the experiment was divided into 12 areas of the same size, the plants were planted and were treated the same way except for the fertilizer. One of the areas was considered as control area and the others did receive different treatments of the fertilizers. The progress of the plant’s budding was measured with three different criterion including FG%, GI and CUG and the measurements were repeated 10 times. The final step was analyzing the data to study the effect of the fertilizers on the budding of the plants.

Network Architecture

The network chosen for this task is the Multi-layer Perceptron (MLP) variant of the Shallow Feedforward Neural Network. Five different models were created with different configurations, Levenberg-Marquardt Algorithm (LMA) was used to train the network and adjust the weights. Number of hidden layer neurons of all the five networks: 12, 12, 12, 12, 28

The models schemas are as follows: Models Schemas

Notes

  1. In order for the Scripts to function, the input and output data must be present in the MATLAB workspace. To do so either drag and drop the data or use the command 'load' (e.g. load('Model 1 Data.mat')).
  2. The networks can be fine-tuned by changing the configuration, its an easy job using the scripts.
  3. More data can probably produce better results.

License

Licensed under the Apache License, Version 2.0