I want to fit a large amount of nonlinear data using multilayer perceptron neural network, what is the accurate, low computational cost and the fastest method. I used "Pytorch" (with "Cuda" which using GPUs) and "Scikit Learn" to train MLP But I am not very satisfied with the results in terms of accuracy and computational cost. What do you suggest?
Understanding what the data represent and that there exist associations a given multilayer perceptron can’t represent.
Well, the NNs are designed to deal with the nonlinear data and their correlation. Simply understand your dataset and target an NN for the required results. A bit more details of your application and data could help a lot in answering your question.
Dear Mohsen Rezaei ,
Any large set of data with nonlinear relation can be dealt with using NN however, it would be better if you can provide more details about the characteristics of your data set. That will help in answering and solving your problem.
Article A neural network analysis of Lifeways cross-generation imputed data
Article Multi-Bias Non-linear Activation in Deep Neural Networks
I think the above links may be helpful
Dear, Mohsen Rezaei!
The developed MLP models based on the Scikit Learn libraries are much inferior to the professional models developed on KERAS and TensorFlow. I personally experimented on the same architectures and hyperparameters, and got different results. The model developed on KARAS and TensorFlow surpassed the models on Scikit Learn. I recommend developing and optimizing the model on KERAS and TensorFlow.
As for a large amount of data, you can optimize the model training process by adjusting the number of buckets. You can also use generators to optimize the flow of the training data set.
If any help requires, you can feel free to contact me: [email protected]
Shahboz Sharifovich Qodirov , Thank you for your complete answer, Another question, How is Pytorch? Do you work with it?
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training