Adaptive Skip Intervals: Temporal Abstraction for Recurrent Dynamical\n Models

2018/08/14 by Alexander Neitz, Giambattista Parascandolo, Neitz, Alexander +5
Computer Science · #Anomaly Detection Techniques and Applications #Data Stream Mining Techniques #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Reinforcement Learning in Robotics #Time Series Analysis and Forecasting

paper · pdf · doi:10.48550/arxiv.1808.04768

openalex publication_date 2018/08/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We introduce a method which enables a recurrent dynamics model to be\ntemporally abstract. Our approach, which we call Adaptive Skip Intervals (ASI),\nis based on the observation that in many sequential prediction tasks, the exact\ntime at which events occur is irrelevant to the underlying objective. Moreover,\nin many situations, there exist prediction intervals which result in\nparticularly easy-to-predict transitions. We show that there are prediction\ntasks for which we gain both computational efficiency and prediction accuracy\nby allowing the model to make predictions at a sampling rate which it can\nchoose itself.\n

Citations

• Adam: A Method for Stochastic Optimization
• Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning
• Action-Conditional Video Prediction using Deep Networks in Atari Games
• Recurrent Environment Simulators
• Value Prediction Network
• Imagination-Augmented Agents for Deep Reinforcement Learning
• Temporal Difference Models: Model-Free Deep RL for Model-Based Control

Related

• Adaptive Drift-Diffusion Process to Learn Time Intervals
• Learning Recurrent Models with Temporally Local Rules
• Time-Aware World Model for Adaptive Prediction and Control
• Deep Explicit Duration Switching Models for Time Series
• Bayesian Inference in Recurrent Explicit Duration Switching Linear Dynamical Systems