I'm currently reading Hoedt et al's paper on mass-conserving LSTM. In the corollary it is stated that

"[T]he memory cells, $c_k^\tau$, are bounded by the sum of mass inputs $\sum_{t=1}^\tau x^t+m_c^0$, that is $\vert c_k^\tau \vert \leq \sum_{t=1}^\tau x^t+m_c^0$. Furthermore, if the series of mass input converges, $\lim_{\tau \rightarrow \infty} \sum_{t=1}^\tau x^\tau=m_x^\infty$ , then also the sum of cell states converges."

I'm having trouble with the statement on convergence. Firstly, shouldn't the "exponent" in the sum be $t$ instead of $\tau$? I'm not even really sure what the index in the "sum of cell states" should be and which parameter is supposed to go to infinity.
Needless to say, I haven't been able to prove this statement, but I suspect it should be simple. The appendix says that the convergence follows with the comparison test.
I'd be grateful for any help on this matter!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.