Forecasting with deep state space models

US12353996B2 · kind B2 · utility

Assignee

• Robert Bosch GmbH · DE

Inventors

• Chen Qiu · Sindelfingen, DE
• Maja Rudolph · Rötteln, DE

Key dates

• Filing date: Aug 20, 2021
• Grant date: Jul 8, 2025
• Expiry date: Apr 10, 2044

Classification

• Technology area (CPC G): Physics
• CPC primary: G06V10/82
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

A computer-implemented method for training a deep state space model using machine learning. The deep state space model includes a generative model and a multi-modal inference model. The generative model includes a transition model, and an emission model. The method includes: a) receiving a training data set comprising a sequence of observation vectors. For a plurality of observation vectors, the method iterates between b), c), and d) in sequence: b) inferring, using the multi-modal inference model, a current latent state of the generative model; c) constructing, using the multi-modal inference model, a posterior approximation of the current latent state as a mixture density network. For a plurality of observation vectors comprised in the sequence of observation vectors, d) decoding, using the emission model, the plurality of approximated latent state vectors to provide a plurality of synthetic observations; and e) outputting the trained deep state space model.