Learning Stochastic Recurrent Networks
About
Leveraging advances in variational inference, we propose to enhance recurrent neural networks with latent variables, resulting in Stochastic Recurrent Networks (STORNs). The model i) can be trained with stochastic gradient methods, ii) allows structured and multi-modal conditionals at each time step, iii) features a reliable estimator of the marginal likelihood and iv) is a generalisation of deterministic recurrent neural networks. We evaluate the method on four polyphonic musical data sets and motion capture data.
Justin Bayer, Christian Osendorfer• 2014
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Polyphonic music modeling | Nottingham (Nott) | NLL (nats)2.85 | 14 | |
| Polyphonic music modeling | JSB Chorales | Negative Log-Likelihood (nats)6.91 | 14 | |
| Polyphonic music modeling | Piano-midi.de | NLL (nats)7.13 | 12 | |
| Polyphonic music modeling | MuseData (Muse) | Negative Log-Likelihood (nats)6.16 | 12 | |
| Polyphonic Music Generation | Nottingham (test) | NLL2.85 | 11 | |
| Polyphonic Music Generation | Piano (test) | Negative Log-Likelihood7.13 | 5 | |
| Polyphonic Music Generation | Musedata (test) | NLL6.16 | 5 | |
| Polyphonic Music Generation | JSB (test) | Negative Log-Likelihood6.91 | 5 |
Showing 8 of 8 rows