Summary:
Today, I first try the methods in papers and online to improve the ANN model, but the performance of the ANN model does not improve a lot.
Then, I watch some cases online and repeat cases in matlab to find that similar ANN model can perform well in these cases.
Afterwards, I use the codes in matlab to apply into my data, the results are not good. I have adjust every parameter of the model, but none of them works well.
The best performance of the ANN model is shown below. It is the comparison plot of the first parameter, namely the first amplitude. The blue scatter points are original data. The red lines are prediction data.
1. I should find more predictors (logging data) which are related to T2 distribution.
2. The 6 individual parameters I get from gaussian distributions fluctuate very randomly. So maybe it should be transformed into some other forms so that they do not fluctuate so much.
Tomorrow, I will try in the two ways to improve my data. There should be no problem of the model.
I want to see the prediction for test data and training data separately...
ReplyDeleteYou should record the the depths of training data and test data.. and then plot them separately
There is little difference between training and testing performance, and they all did badly. I will show them tonight if you want.
Deletethe predictors should be actual measurements and not some estimation. .... we need actual measured logs as inputs .... also we need to predict NMR with least number of input logs.
ReplyDeleteYes, I agree. But since our parameters fluctuate very randomly, maybe least number of input log are not enough to build a precise model.
Delete2. The 6 individual parameters I get from gaussian distributions fluctuate very randomly. So maybe it should be transformed into some other forms so that they do not fluctuate so much.
ReplyDelete------------------
I agree with you... they fluctuate a lot... in that case should the input logs and the six parameters be averaged as a moving average of let say 2 depth on top and 2 depths below?
So do you mean that we should average 5 continuous depths into one? I will consider about it.
Deleteanother suggestion.. can you predict the amplitude mean and variance associated with the biggest amplitude gaussian
ReplyDeletein other words, rather than predicting two peaks... can you predict only one peak ... the some of the log data might be more sensitive to either the smaller peaks or the bigger peaks
Due to the theory of ANN model, I think there is little difference between the performance of the bigger peak and the smaller peak. I think no matter how big or how small the peaks are, if they do not fluctuate a lot, the performance will be better.
ReplyDelete