I have 4 different experimental parameters using which I want to minimize my 5th parameter. The relation between the 4 parameters is unknown. For this problem, I want to implement ANN to get an optimised solution. Can someone please tell me how I can do it and which algorithms I need to use?
This is a general question and you need to consider a lot of factors. For instance to name a few
1. Can your data be represented by simpler statistical methods? (Statistical methods may even work better than ANN)
2. How many data points do you have? Is ANN worth try if you don't have many data points?
3. How complex are your data? Simpler ANN methods or advanced ones?
4. What optimization do you want to use? MSE, MAE, or a specific one? What does optimization mean in this context?
If you want to use ANN, I suggest you start with algorithms such as multilayer perceptrons. https://www.sciencedirect.com/topics/computer-science/multilayer-perceptron
Then you can make better models step by step.
Thank You, Mostafa Kiani Shahvandi, for your answer. For the questions you raised, I would like to add few points to clarify them:
1. I am thinking about ANN because my dataset has almost 10000 data points
considering which application of statistical method seems an impossible task.
2. Regarding the complexity of the dataset, I really didn't get your point, but as far as I understand, each parameter has a different range of values(but all positive)
3. What I mean by optimization is I have values of my 5th parameter for the range of the other 4 parameters. So I am looking for the minimum value of the 5th parameter of the best possible pair of the other 4 parameters.
4. I thought of implementing a genetic algorithm using a multilayered perception neural network as previously used in the literature for similar problems. But the problem here is, will it work? And if so, how to implement it?
Dear Suraj Lahase,
I have read your answers, but one thing is unclear:
3. What I mean by optimization is I have values of my 5th parameter for the range of the other 4 parameters. So I am looking for the minimum value of the 5th parameter of the best possible pair of the other 4 parameters.
You must formulate your optimization problem through the prism of a regression or classification problem, otherwise it will be difficult for you to solve it. ANN will perfectly cope with your task, since you have enough sample data. I am ready to help you if you can correctly formulate the optimization problem.
If you need to find the smallest values of the 5th parameter for all 4 parameters, then it is enough for you to write an algorithm from the logical conditions for finding such a combination (if I understood your idea correctly).
Shahboz Sharifovich Qodirov thank you for your answer. I will work on the points you mentioned here and will get back to you if anything is needed. Also, I cannot write a simple code as I am also looking for value prediction of my 5th parameter given that any one or more parameter values are changed.
- Badges
- Science topic