What would be a ideal Machine Configuration to run a Recurrent Neural Network

neural_network

#1

I am attempting to set up a Recurrent Neural Network for Time Series analysis for Financial markets. I would be implementing this in Theano. The setup up would be on a single high end machine. What should be the ideal configuration assuming that i would be writing code targetting GPU.
The MongoDB database size would be approximately 100MB or for a csv file it would be 500MB. I would appreciate a rough estimate of the probable time to train a deep belief net with six predictors basically Open, High, Low, Close, Volume, Time.


#2

I have just posted this 2 hours back and if one google’s for "ideal machine configuration for recurrent neural networks"
The first link displayed by google is this query. The indexing done by Google appears to be very fast.


#3

Hi @Ramesh_Ramachandran,

There is no “ideal” machine configs to run an NN algorithm, but it is generally recommended that the higher the specs you have, the better it is. There are some guidelines mentioned in this article, do go through it.