Timely Communications for Remote Inference

TL;DR

This work shows that inference performance in remote sensing can depend nontrivially on information freshness, even contradicting the common belief that newer data is always better. It introduces a general selection-from-buffer transmission model and develops scheduling policies that minimize general, possibly non-monotonic AoI functions for both single-source and multi-source settings. The authors provide an information-theoretic interpretation via $L$-entropy and cross-entropy, construct analysis using an $oldsymbol{ extepsilon}$-Markov framework, and prove the asymptotic optimality of a Whittle-index–based, dual-optimization scheduling policy. Data-driven experiments demonstrate substantial reductions in inference error compared to traditional generate-at-will and periodic updating approaches, highlighting the practical impact for next-generation networks and time-critical AI via efficient, timely communications.

Abstract

In this paper, we analyze the impact of data freshness on remote inference systems, where a pre-trained neural network blue infers a time-varying target (e.g., the locations of vehicles and pedestrians) based on features (e.g., video frames) observed at a sensing node (e.g., a camera). One might expect that the performance of a remote inference system degrades monotonically as the feature becomes stale. Using an information-theoretic analysis, we show that this is true if the feature and target data sequence can be closely approximated as a Markov chain, whereas it is not true if the data sequence is far from being Markovian. Hence, the inference error is a function of Age of Information (AoI), where the function could be non-monotonic. To minimize the inference error in real-time, we propose a new "selection-from-buffer" model for sending the features, which is more general than the "generate-at-will" model used in earlier studies. In addition, we design low-complexity scheduling policies to improve inference performance. For single-source, single-channel systems, we provide an optimal scheduling policy. In multi-source, multi-channel systems, the scheduling problem becomes a multi-action restless multi-armed bandit problem. For this setting, we design a new scheduling policy by integrating Whittle index-based source selection and duality-based feature selection-from-buffer algorithms. This new scheduling policy is proven to be asymptotically optimal. These scheduling results hold for minimizing general AoI functions (monotonic or non-monotonic). Data-driven evaluations demonstrate the significant advantages of our proposed scheduling policies.

Figures (12)

• Figure 1: Performance of video prediction experiment. The experimental results in (b) and (c) are regenerated from lee2018stochastic. The training and inference errors are non-decreasing functions of the AoI.
• Figure 2: Robot state prediction in a leader-follower robotic system. The leader robot uses a neural network to predict the follower robot's state $Y_t$ by using the leader robot's state $X_{t-\delta}$ generated $\delta$ time slots ago $(u=1)$. The training and inference errors decrease in the AoI $\leq 25$ and increase when AoI $\geq 25$.
• Figure 3: Performance of actuator state prediction under mechanical response delay. In the OpenAI CartPole-v1 task brockman2016openai, the pole angle $\psi_t$ is predicted by using $X_{t-\delta}=(v_{t-\delta}, v_{t-\delta-1}, \ldots, v_{t-\delta-u-1})$, where $v_{t}$ is the cart velocity at time $t$. The training error and inference error are non-monotonic in the AoI.
• Figure 4: Performance of temperature prediction. The training error and inference error are non-monotonic in AoI. As the feature sequence length $u$ increases, the errors tend closer to non-decreasing functions of the AoI.
• Figure 5: Performance of channel state information prediction. The training error and inference error are non-monotonic in AoI. As the feature sequence length $u$ increases, the errors tend closer to non-decreasing functions of the AoI.

Theorems & Definitions (47)

• Definition 1: $\epsilon$-Markov Chain
• Lemma 1
• proof
• Lemma 2: $\epsilon$-data processing inequality
• proof
• Theorem 1
• proof
• Definition 2: Univariate Stochastic Ordering
• Theorem 2
• proof