Decoupling Inference from State Updates in Low-Latency Feature Engines via Probabilistic Thinning

• URL: http://arxiv.org/abs/2606.16981v1
• Date: Mon, 15 Jun 2026 17:18:05 GMT
• Title: Decoupling Inference from State Updates in Low-Latency Feature Engines via Probabilistic Thinning
• Abstract summary: We decouple inference from state persistence in streaming Machine Learning pipelines via probabilistic thinning.<n>We show that persistence-path control is achievable without a high-frequency in-memory control plane or cross-worker.<n>Across experiments, up to 90% of events are excluded from the persistence path while preserving and in some cases improving downstream utility.
• Score: 6.081561157196023
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Streaming data systems increasingly underpin Machine Learning workflows that maintain large numbers of continuously updated aggregations. In production settings, each incoming event typically triggers read-modify-write operations to persistent storage, making high-frequency state updates a dominant source of latency, contention, and operational cost. In this work, we decouple inference from state persistence in streaming Machine Learning pipelines via probabilistic thinning: every event is scored, but durable state updates are selectively triggered by informative events. Unlike approaches that shed input or state, we show that persistence-path control is achievable without a high-frequency in-memory control plane or cross-worker coordination, relying exclusively on approximate statistics retrieved from disk-backed key-value stores. We model the resulting stochastic processes, derive bounds on filtering rates, and prove that common time-based aggregations remain unbiased under variance-aware formulations, preventing systemic error accumulation. We evaluate the approach in a controlled setting that isolates per-event costs, demonstrating substantial reductions in storage Input/Output and serialization overhead. Across experiments, up to 90% of events are excluded from the persistence path while preserving and in some cases improving downstream utility.

Related papers

• A Unified Particle Filter LSTM for Data-Driven Process Simulation [1.9473631530390483]
  Event logs provide only a partial view of the underlying process state.<n>We propose a Unified Particle Filter LSTM that maintains and sequentially updates a weighted set of recurrent-state hypotheses.<n>The framework is trained end-to-end from event-log data and evaluated on three real-world emergency department datasets.
  arXiv  Detail & Related papers  (2026-09-02T00:53:37Z)
• Enhancing Irregular Time Series Forecasting with Continuous-Time Modeling Framework [15.900704071560698]
  WrapFlow is a continuous-time modeling framework for irregular time series forecasting.<n>We develop a training paradigm for Residual Flow Matching, which learns conditional residual vector fields around base predictions.<n>Experiments on multiple real-world datasets demonstrate that WrapFlow achieves state-of-the-art performance.
  arXiv  Detail & Related papers  (2026-07-30T11:18:18Z)
• TimeLAVA: Learning-Agnostic Data Valuation for Time Series [48.76711740818533]
  TimeLAVA is a learning-agnostic framework that values temporal segments by their marginal contribution to minimize distributional discrepancy between evaluated and reference data.<n>We provide theoretical guarantees linking valuation to model-agnostic generalization and prove bounded sensitivity to contamination.
  arXiv  Detail & Related papers  (2026-06-17T06:16:16Z)
• CINDI: Conditional Imputation and Noisy Data Integrity with Flows in Power Grid Data [0.6991281327290525]
  Conditional Imputation and Noisy Data Integrity (CINDI) is an unsupervised probabilistic framework designed to restore data integrity in complex time series.<n>CINDI unifies anomaly detection and imputation into a single end-to-end system built on conditional normalizing flows.<n>We evaluate the framework using real-world grid loss data from a Norwegian power distribution operator.
  arXiv  Detail & Related papers  (2026-03-12T09:49:10Z)
• Causify DataFlow: A Framework For High-performance Machine Learning Stream Computing [0.0]
  We present DataFlow, a computational framework for building, testing, and deploying machine learning systems on unbounded time-series data.<n>Traditional data science assume finite datasets and require substantial reimplementation when moving from batch prototypes to streaming production systems.<n>DataFlow resolves these issues through a unified execution model based on acyclic graphs with point-in-time idempotency.
  arXiv  Detail & Related papers  (2025-12-30T04:24:04Z)
• Explainable Anomaly Detection for Industrial IoT Data Streams [1.2722697496405462]
  Industrial maintenance is being transformed by the Internet of Things and edge computing.<n>This paper presents a framework that integrates unsupervised anomaly detection with interactive, human-in-the-loop learning to support maintenance decisions.
  arXiv  Detail & Related papers  (2025-12-09T18:20:35Z)
• Steering Vision-Language-Action Models as Anti-Exploration: A Test-Time Scaling Approach [78.4812458793128]
  We propose textbfTACO, a test-time-scaling framework that applies a lightweight pseudo-count estimator as a high-fidelity verifier of action chunks.<n>Our method resembles the classical anti-exploration principle in offline reinforcement learning (RL), and being gradient-free, it incurs significant computational benefits.
  arXiv  Detail & Related papers  (2025-12-02T14:42:54Z)
• STDiff: A State Transition Diffusion Framework for Time Series Imputation in Industrial Systems [2.3895981099137535]
  We propose STDiff, which reframes imputation as learning how the system evolves from one state to the next.<n> STDiff consistently achieves the lowest errors, with its advantage increasing for longer gaps.<n>These results support dynamics-aware, explicitly conditioned imputation as a robust approach for industrial time series.
  arXiv  Detail & Related papers  (2025-08-26T13:14:53Z)
• Event Signal Filtering via Probability Flux Estimation [58.31652473933809]
  Events offer a novel paradigm for capturing scene dynamics via asynchronous sensing, but their inherent randomness often leads to degraded signal quality.<n>Event signal filtering is thus essential for enhancing fidelity by reducing this internal randomness and ensuring consistent outputs across diverse acquisition conditions.<n>This paper introduces a generative, online filtering framework called Event Density Flow Filter (EDFilter)<n>Experiments validate EDFilter's performance across tasks like event filtering, super-resolution, and direct event-based blob tracking.
  arXiv  Detail & Related papers  (2025-04-10T07:03:08Z)
• Continual Low-Rank Scaled Dot-product Attention [67.11704350478475]
  We introduce a new formulation of the Scaled-product Attention based on the Nystr"om approximation that is suitable for Continual Inference.<n>In experiments on Online Audio Classification and Online Action Detection tasks, the proposed Continual Scaled Dot-product Attention can lower the number of operations by up to three orders of magnitude.
  arXiv  Detail & Related papers  (2024-12-04T11:05:01Z)
• Value function estimation using conditional diffusion models for control [62.27184818047923]
  We propose a simple algorithm called Diffused Value Function (DVF) It learns a joint multi-step model of the environment-robot interaction dynamics using a diffusion model. We show how DVF can be used to efficiently capture the state visitation measure for multiple controllers.
  arXiv  Detail & Related papers  (2023-06-09T18:40:55Z)
• Deep Explicit Duration Switching Models for Time Series [84.33678003781908]
  We propose a flexible model that is capable of identifying both state- and time-dependent switching dynamics. State-dependent switching is enabled by a recurrent state-to-switch connection. An explicit duration count variable is used to improve the time-dependent switching behavior.
  arXiv  Detail & Related papers  (2021-10-26T17:35:21Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.