im a beginner.i use lstm to recognize this data, but loss is very big, recognition is bad.i dont know why, is lstm suit for the data(2133,32,2)?
- The 2 feature vectors are as follows
I don’t think this is really a question about Python, or about writing code, at all. I think it is a question about the underlying neural network theory.
I also can’t understand what you mean by “this data”. “(2133,32,2)” only tells us the dimensions of the data, not what’s actually in it. (This means more than a hundred thousand different values, so please do not try to include them in the post.) The feature vectors are also not the same thing as the data itself.
You are probably going to want to ask the question somewhere else, that is actually about neural networks. Somewhere that you can say something like “recognize this data” and it will actually mean something to the audience - it isn’t clear, in a general Python programming context, exactly what you expect to happen when the code runs, what concrete output you should get when the data is “recognized” well.
My last advice is that if you are a beginner to Python, you should gain more experience with Python before trying to use neural networks.
thanks for your reply. my question is about the network indeed. the data is from radar signal, i want recognize the signal pulse sequence,(2133,32,2) means each pulse sequence has 32 pulses,each pulse has 2 features, and we got 2133 sequence samples. and you can see the 2133x32x2 sequences in figure.
i just want to build a simple lstm netwrok to recognize it, but i got loss = nan or loss always be a big number and never get small.