Menu
I am trying to use the convolutional residual network neural network architecture (ResNet). So far, I have implemented simple convolutions (conv1D) for time series data classification using Keras.
Now, I am trying to build ResNet using Keras but I'm having some difficulties trying to adapt it to time series data. Most of the implementations of ResNet or Nasnet in Keras (such as this one or that one) use conv2D for their implementation (which makes sense for images).
Could someone help me in implementing this for time series data?
938
asked Mar 17 '18 14:03
CNN is suitable for forecasting time-series because it offers dilated convolutions, in which filters can be used to compute dilations between cells. The size of the space between each cell allows the neural network to understand better the relationships between the different observations in the time-series [14].
Yes. ResNet is a way to handle the vanishing gradient problem in very deep CNNs. They work by skipping some layers assuming the fact that very deep networks should not produce a training error higher than its shallower counterparts. In an overall perspective they can be thought of as a model similar to LSTM in RNNs.
Do you know about the paper "Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline"? If not, you should check it out. The authors provide a very comprehensive overview of different models, including a ResNet implementation adjusted for time series classification.
Their Keras/Tensorflow implementation of ResNet can be found here.
Update: A more recent version of ResNet (and other classifiers) for time series data can be found here.
169
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
