Temporally Folded Convolutional Neural Networks for Sequence Forecasting
This work addresses sequence forecasting problems, potentially benefiting domains like video analysis or music prediction, but it appears incremental as it builds on existing CNN and LSTM methods without a major paradigm shift.
The paper tackles time series forecasting by proposing a novel approach using convolutional neural networks, where the time dimension is treated as an input to a spatiotemporal network, and reports that it may outperform conventional recurrent strategies like convolutional LSTMs and LSTMs on datasets such as sequential MNIST and JSB chorals.
In this work we propose a novel approach to utilize convolutional neural networks for time series forecasting. The time direction of the sequential data with spatial dimensions $D=1,2$ is considered democratically as the input of a spatiotemporal $(D+1)$-dimensional convolutional neural network. Latter then reduces the data stream from $D +1 \to D$ dimensions followed by an incriminator cell which uses this information to forecast the subsequent time step. We empirically compare this strategy to convolutional LSTM's and LSTM's on their performance on the sequential MNIST and the JSB chorals dataset, respectively. We conclude that temporally folded convolutional neural networks (TFC's) may outperform the conventional recurrent strategies.