I need to create a simple Recurrent Neural Network RNN or Long short-term memory (LSTM), which is specific type of RNN. After searching through examples and forums I haven't come across many applications that are similar enough to my project. I tried to implement a simple RNN to classify the ECG signals into two classes, but no luck so far. I'm confused about What is the correct input shape for the model?.
Let say I have 20 ECG record. My input data is such that I have m=20 signals each with n=7000 data points, and L=2 output labels that I am trying to learn. How I should structuring my input data. Is it true to make the input as (m,n,1) and output targets as (m,L)!.
I attempt to use the following Matlab toolboxes for building the RNN:
https://github.com/yechengxi/LightNet.
Please help!
I recommend to use SVM instead of neural networks, because most researchers proved that SVM is more powerful than neural networks in classification tasks. From my experience, SVM will be simpler to implement and test on your data and more understood.
Hey there, you can start with a simple sliding window approach. First prepare your sliding window, set how much is the length of your window and slide that sliding window one step at a time. These sliding window of your data would be the feature for your study. Now for each sliding window, place a desired class. This will be the labels.
You can then build your LSTM classifier using the sliding windows as input.
If you want an example you can check out this Jupyter notebook. The notebook contains a section on how to generate the sliding window from the data. You can either get that code and convert to Matlab or create a python script out of the snippet.
https://github.com/cadrev/lstm-flood-prediction/blob/master/lstm_flood.ipynb
What is the point of using a recurrent neural network for classifying ECG heartbeats ? Recurrent neural networks are useful for modeling dynamic processes, which, in your case, would just be a time series, without exogenous inputs (contrary to flood prediction mentioned above). The reason why you did not find any similar example in the literature is probably that it is a bad idea, in addition to the very sensible arguments provided by the first answer above.
There is a huge literature on hearbeat classification, using loads of different classifiers (including neural networks and SVMs); all state-of-the art classifiers provide basically the same results, which are very good because hearbeat classification is a simple problem if it is addressed correctly.
In my view, the key issue in heartbeat classification (as in most classification problems) is to use a maximum amount of expert knowledge in order to find an appropriate representation of the data. If you find an original data representation, more efficient than usual ones, you will make a nice scientific contribution; but I am afraid that trying to perform heartbeat classification with recurrent neural networks won't lead you anywhere. Sorry for being negative!
There's nothing less helpful than people chiming in on a message board like this to say "you don't want to do what you're trying to do" instead of actually attempting to answer the question as posed.
I stumbled on this page because I also want to understand the the input and output shapes of LSTM RNNs. I could not care less whether heartbeats are an appropriate application... in theory, it's a time series, so the tool is potentially useful.
Alex Teachey
There’s nothing more entertaining than people barging in on a message board like this and giving their priceless advice on subjects they know nothing about. Typically, astronomer (“exomoon hunter”) Alex Teachey giving – without anyone asking him to - his enlightened opinion on the machine-learning classification of heartbeats from ECG data. “It’s a time series, so the LSTM RNN is potentially useful”: yeah, sure. “Going from location A to location B is a transportation problem, so the airplane is potentially useful”: yeah, sure. Nevertheless, knowing that location A is three miles away from location B might help refrain from taking a sledge hammer to crush a tiny nut.
There are thousands of published papers on LSTMs and RNNs applied to time series; they should provide you with as a huge number of “input and output shapes of LSTM RNNs”, whatever that means. My bookmarked favorite is “Applying LSTM to time series predictable through time-window approaches” (ICANN 2001, doi: 10.1007/3-540-44668-0_93); in that paper, Jürgen Schmiedhuber (the inventor of LSTMs) and his collaborators report that LSTMs have no advantage over other methods for a specific class of time series – probably the only paper in the machine learning literature acknowledging that the authors’ own method is not universally superior to all others. The paper is about laser signals, about which, of course, you "could not care less”.
I must apologize deeply for having replied, three years ago, to Huda Da’s question on heartbeat classification by discussing heartbeat classification, despite the fact that you “could not care less whether heartbeats are an appropriate application”; obviously, I should have anticipated that. Sorry for wasting your invaluable time reading my worthless answer to Huda Da’s question, and thanks for a good laugh!
- Badges
- Science topic