This week we are making up some data and doing a couple of visualizations. We then modify the data to do a reality check on a simple algorithm. This allows us to gain confidence in our algorithm before we send it real data.
You can follow any comments to this entry through the RSS 2.0 feed.
You can skip to the end and leave a comment. Pinging is currently not allowed.
2 CommentsOldest to Newest
Thank you for the tutorials. Could you kindly offer me some tips on how to generate artificial data to test an algorithm written in matlab? I wish that the generated data will be monotone and non-linear also contain parameters. The aim is to obtain such parameters using the well-known nonlinear regression using a modified Neural Network training method. For example, I tried using
ydata = theta1*(g(theta2*x1 + theta3*x2)) + theta1*(g(theta4*x1 + theta5*x2))
but onbtained a linear relationship instead. Besides, the given equation represents a neural network, and using it in a reverse form might render the problem too simple to solve. I generated 100 data points for x1,x2 and obtained same number of points for ydata. The parameters theta1,theta2,theta3 and theta4 are all assigned to random numbers between 0 and 1. g/eta=(1/1+exp(-eta))
synthetic data generation