I am implementing VAE based anomaly detection for multivariate timeseries using keras, I have ELBO (Evidence lower bound) which is combination of

$$-\ D_{KL}\left({\ q}_\varphi\left(z\middle| x^i\right)\parallel p_\theta\left(z\right)\ \right)\ +\ \ E_{q_\varphi(z|x^i)}\left[\log{p_\theta\left(x\middle| z\right)}\right]$$

The loss function is combination of minimizing the KL[Q(z|x)|p(z)] and maximizing log [P(X|z)] how do I implement the above loss function for the multivariate timeseries of shape (batch_size,time_steps,features) using Tensorflow ? appreciate any suggestions

  • $\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$
    – vl_knd
    Commented Feb 15 at 15:02
  • $\begingroup$ Hi yeah i am basically struggling with the implementation part of Loss function for VAE and my data is multivariate timeseries data. $\endgroup$
    – mmv_87
    Commented Feb 28 at 20:52


You must log in to answer this question.

Browse other questions tagged .