Hello, I am relatively new in Machine Learning and I recently stumbled upon Deep Learning which (if I am right) Machine Learning Using Massive Neural Networks with several hidden layers. Can it be used for regression of real continuous functions?
Hello, yes you can use the deep neural networks for regression problem. Just replace the loss function at the end of the network.
For regression problem, usually the Euclidean loss function was used at the end of the network.
you may find this tutorial useful
http://machinelearningmastery.com/regression-tutorial-keras-deep-learning-library-python/
yes, it is a flexible non linear framework, so it can be used to approximate arbitrary functions
Lex Fridman of MIT has a series of lectures available on YouTube covering deep learning CNNs and self driving cars. Here is one of them: https://www.youtube.com/watch?v=1L0TKZQcUtA
https://www.youtube.com/watch?v=1L0TKZQcUtA
Yes, you can do regression with Deep Learning. I have a paper on counting records in handwritten documents using a Convolutional Neural Network with Euclidean Loss.
Conference Paper Record Counting in Historical Handwritten Documents with Con...
By replacing the loss function and adding the linear transfer function at output layer, the architecture can be used for regression tasks
I have implemented DBN-RBMs for time series prediction problem.
https://www.researchgate.net/publication/311705779_An_Application_of_Internet_Traffic_Prediction_with_Deep_Neural_Network
Conference Paper An Application of Internet Traffic Prediction with Deep Neural Network
https://www.researchgate.net/publication/311628294_A_Hybrid_Approach_for_Time_series_Forecasting_using_Deep_Learning_and_Nonlinear_Autoregressive_Neural_Network
Conference Paper A Hybrid Approach for Time series Forecasting using Deep Lea...
- Badges
- Science topic