Recommendations or resources for neural network/deep learning for time series application?

Question

I know there are quite a few good deep learning books out there, but most explain neural networks and deep learning via application on images. If there are examples/code, they are often done on the MNIST data set.

I was wondering if there's a book out there that goes in depth into neural networks and is equally well written but explains it on non-image data. I am particularly interested in time series application, although some talk on cross-sectional application would be helpful as well. I am particularly interested in learning about:

• What types of layers/functions/structure are better suited for time-series data
• Various models for time series and pros/cons/applications of each (Convoluted NN, LTSM, etc...)
• Typical structures of your neural network (depth, sparse connections, etc...) that seem to work well on time series data
• Special considerations or settings you should have in your neural network when working with time series data
• Maybe some talk/examples on how time series prediction using traditional models like ARIMA, can be reproduced or done better using neural networks. Or side by side comparison of pros/cons of using one vs the other.

Thanks!

Answer 1

I'm currently working with Temporal Convolution Networks (TCNs) for making predictions with time series data (link to article here: https://medium.com/@raushan2807/temporal-convolutional-networks-bfea16e6d7d2). These types of networks, like other types of convolutional networks for time series, use a dilated convolution operation, which, unlike the standard convolution in these networks, is a causal operation that preserves the causality of the time series input.

I've also seen CNNs used in Temporal Differencing applications, but that may be more useful in the context of video processing rather than time series analysis.