The good performance of the Vanilla RNN, which does not integrate the “long” aspect of the LSTM algorithm, implies that the time series follows a pattern that does not require much of a long-term memory.
Read moreIs there anything better than LSTM?
Temporal convolutional network (TCN) “outperform canonical recurrent networks such as LSTMs across a diverse range of tasks and datasets, while demonstrating longer effective memory”. Note 4: Related to this topic, is the fact that we know little of how our human brain learns and remembers sequences.
Read more