Interpretable Representation Learning for Motion Forecasting

35.51 €English / Englisch

Royden Wagner · KIT Scientific Publishing

Description

We address interpretable representation learning for motion forecasting in self-driving cars. Rather than treating transformers as black boxes, we develop methods to interpret and modify learned representations. We introduce self-supervised pre-training with interpretable objectives. Moreover, we probe latent spaces of forecasting models and reveal interpretable features, allowing us to make targeted interventions. Finally, we uncover retrocausal mechanisms, which enable goal-based instructions.

Ausgabe

ISBN: 9783731514749 · KIT Scientific Publishing

Additional information

Weight 0.33 kg
Dimensions 21 × 14.8 cm
Autor / Mitwirkende

Royden Wagner

Verlag

KIT Scientific Publishing

ISBN

9783731514749

Erscheinungsdatum

2026-05-13

Zustand

Neu

Buchsprache / Book language

English / Englisch