Search Results for « Latent computing »

Abstract Novel applications will require extending traditional cloud computing infrastructure with compute resources deployed close to the end user. Edge and fog computing tightly integrated with carrier networks can fulfill this demand. The emphasis is on integration: the rigorous delay constraints, ensuring reliability on the distributed, remote compute nodes, and the sheer scale of the system altogether call for a powerful resource provisioning platform that offers the applications the best of the underlying infrastructure. We therefore propose Kubernetes-edge-scheduler that provides high reliability for applications in the edge, while provisioning less than 10% of resources for this purpose, and at the same time, it guarantees compliance with the latency requirements that end users expect. We present a novel topology clustering method that considers application latency requirements, and enables scheduling applications even on a worldwide scale of edge clusters. We demonstrate that in a potential use case, a distributed stream analytics application, our orchestration system can reduce the job completion time to 40% of the baseline provided by the default Kubernetes scheduler.

Abstract Time stretch instruments have been exceptionally successful in discovering single-shot ultrafast phenomena such as optical rogue waves and have led to record-speed microscopy, spectroscopy, lidar, etc. These instruments encode the ultrafast events into the spectrum of a femtosecond pulse and then dilate the time scale of the data using group velocity dispersion. Generating as much as Tbit per second of data, they are ideal partners for deep learning networks which by their inherent complexity, require large datasets for training. However, the inference time scale of neural networks in the millisecond regime is orders of magnitude longer than the data acquisition rate of time stretch instruments. This underscores the need to explore means where some of the lower-level computational tasks can be done while the data is still in the optical domain. The Nonlinear Schrödinger Kernel computing addresses this predicament. It utilizes optical nonlinearities to map the data onto a new domain in which classification accuracy is enhanced, without increasing the data dimensions. One limitation of this technique is the fixed optical transfer function, which prevents training and generalizability. Here we show that the optical kernel can be effectively tuned and trained by utilizing digital phase encoding of the femtosecond laser pulse leading to a reduction of the error rate in data classification.

To improve low-latency computing and communication services, a new type of mobile edge computing architecture named multi-layer cloud radio access network (Multi-layer CRAN) is designed in this paper. In Multi-layer CRAN, a high-level edge cloud is deployed next to base band unit pool to handle the computing tasks of user equipment (UE) in centralized way. Meanwhile, a low-level edge cloud is deployed in each remote radio head (RRH) to locally handle UEs' computing tasks in a distributed way. Based upon Multi-layer CRAN, a cooperative communication and computation resource allocation (3C-RA) algorithm is further designed for lower service latency and energy cost, and higher network throughput in this paper. 3C-RA utilizes a distributed RRH cell coloring algorithm to enable each RRH to work out the resource allocation in an efficient and distributed way. 3C-RA employs a proportional fairness-based approach to allocate communication and computation resource in each RRH cell. A series of simulations on Multi-layer CRAN with 3C-RA were carried out. The simulation results validate that Multi-layer CRAN is more capable of providing low-latency computing and communication services, and 3C-RA enables Multi-layer CRAN to have lower service latency and energy cost and higher network throughput.

The rapid development of real-time hardware generation has led to low latency computing becoming one of the factors to improve system performance.This paper discusses the delay challenges in real-time hardware generation, including low efficiency of hardware resource scheduling, bandwidth limitation of data transmission, memory access bottleneck and insufficient collaborative optimization of hardware and software, and proposes to use dynamic scheduling to improve the efficiency of hardware resource use, upgrade the data transmission interface to improve bandwidth, and realize intelligent cache management to improve memory access speed.And deepen the hardware and software co-design to improve the running degree, effectively reduce the delay time, improve the comprehensive performance of real-time hardware generation.

Although individual neurons and neural populations exhibit the phenomenon of representational drift, perceptual and behavioral outputs of many neural circuits can remain stable across time scales over which representational drift is substantial. These observations motivate a dynamical systems framework for neural network activity that focuses on the concept of \emph{latent processing units,} core elements for robust coding and computation embedded in collective neural dynamics. Our theoretical treatment of these latent processing units yields five key attributes of computing through neural network dynamics. First, neural computations that are low-dimensional can nevertheless generate high-dimensional neural dynamics. Second, the manifolds defined by neural dynamical trajectories exhibit an inherent coding redundancy as a direct consequence of the universal computing capabilities of the underlying dynamical system. Third, linear readouts or decoders of neural population activity can suffice to optimally subserve downstream circuits controlling behavioral outputs. Fourth, whereas recordings from thousands of neurons may suffice for near optimal decoding from instantaneous neural activity patterns, experimental access to millions of neurons may be necessary to predict neural ensemble dynamical trajectories across timescales of seconds. Fifth, despite the variable activity of single cells, neural networks can maintain stable representations of the variables computed by the latent processing units, thereby making computations robust to representational drift. Overall, our framework for latent computation provides an analytic description and empirically testable predictions regarding how large systems of neurons perform robust computations via their collective dynamics.

Modern AI architectures operate natively in continuous high-dimensional geometric spaces (R^N, N >= 10,000), yet current programming paradigms force them to communicate via discrete sequential text channels — a mismatch we call the impedance gap. Recent empirical work has demonstrated that eliminating this gap substantially improves multi-agent coordination: Interlat [Du et al. 2025] shows that transmitting hidden states instead of tokens carries approximately 40,000 bits of information per state versus ~15 bits per language token — a 2,667× semantic bandwidth gain — while LatentMAS [ICML 2026 Spotlight] achieves superior multi-agent performance and major wall-clock speedups via training-free latent alignment. We investigate whether a formal programming language can be specified directly over the geometric space of transformer and SSM activations, without serialization to text. We propose POLYDIM, a language whose fundamental program unit is a geometric transformation T: R^N -> R^N. We prove four algebraic theorems establishing POLYDIM's formal foundations: associativity of composition (T1), linearity of superposition (T2), functoriality of projection to executors (T3, proved unconditionally for three contracts — COMPILE and EXPORT via the Subspace Commutativity Lemma, RENDER via the Flutter Algebraic Isomorphism), and uniqueness of fixed points (T4). We formalize a two-layer architecture separating the invariant algebraic core (COMPOSE, MIX, FIXPOINT, PROJECT) from architecture-specific implementation primitives (ATTEND for transformers, RECUR for SSMs such as Mamba-3 [ICLR 2026]), guaranteeing that the language specification is robust to architectural transitions. A Python bootstrap with 29/29 tests and +64.5% semantic improvement validates the conceptual framework. POLYDIM fills the gap between empirical latent communication research and formal language design: the specification that systems like Interlat and LatentMAS implicitly require but do not provide.

A disposable vape contains a 24 MHz CPU, 24 KB of storage, and 3 KB of RAM. An embedded engineer recently demonstrated that this is enough to run network protocols, serve web pages, maintain counters, and expose APIs. The security implications are not limited to vapes. They generalize to every microcontroller-bearing device that organizations treat as a non-computer—sensors, badges, peripherals, consumer gadgets, and any hardware with hidden or under-documented compute. This paper presents five artifacts for governing latent compute in AI-era environments: a governance principle (Scarcity-Forged Compute Awareness), a threat-modeling pattern (Hidden Compute Node), a case study demonstrating the failure chain from trivial device to covert infrastructure, a teaching example for AI-coding governance, and a formal control family (Latent Compute and Embedded Surfaces, LCES) with five invariants, five required controls, and five failure signatures. The governing insight is architectural: if it has a CPU, it is part of the execution substrate and must be governed as such.

Computation off-loading in mobile edge computing (MEC) systems constitutes an efficient paradigm of supporting resource-intensive applications on mobile devices. However, the benefit of MEC cannot be fully exploited, when the communications link used for off-loading computational tasks is hostile. Fortunately, the propagation-induced impairments may be mitigated by intelligent reflecting surfaces (IRS), which are capable of enhancing both the spectral- and energy-efficiency. Specifically, an IRS comprises an IRS controller and a large number of passive reflecting elements, each of which may impose a phase shift on the incident signal, thus collaboratively improving the propagation environment. In this paper, the beneficial role of IRSs is investigated in MEC systems, where single-antenna devices may opt for off-loading a fraction of their computational tasks to the edge computing node via a multi-antenna access point with the aid of an IRS. Pertinent latency-minimization problems are formulated for both single-device and multi-device scenarios, subject to practical constraints imposed on both the edge computing capability and the IRS phase shift design. To solve this problem, the block coordinate descent (BCD) technique is invoked to decouple the original problem into two subproblems, and then the computing and communications settings are alternatively optimized using low-complexity iterative algorithms. It is demonstrated that our IRS-aided MEC system is capable of significantly outperforming the conventional MEC system operating without IRSs. Quantitatively, about 20 % computational latency reduction is achieved over the conventional MEC system in a single cell of a 300 m radius and 5 active devices, relying on a 5-antenna access point.

Fog computing is an emerging architecture intended for alleviating the network burdens at the cloud and the core network by moving resource-intensive functionalities such as computation, communication, storage, and analytics closer to the End Users (EUs). In order to address the issues of energy efficiency and latency requirements for the time-critical Internet-of-Things (IoT) applications, fog computing systems could apply intelligence features in their operations to take advantage of the readily available data and computing resources. In this paper, we propose an approach that involves device-driven and human-driven intelligence as key enablers to reduce energy consumption and latency in fog computing via two case studies. The first one makes use of the machine learning to detect user behaviors and perform adaptive low-latency Medium Access Control (MAC)-layer scheduling among sensor devices. In the second case study on task offloading, we design an algorithm for an intelligent EU device to select its offloading decision in the presence of multiple fog nodes nearby, at the same time, minimize its own energy and latency objectives. Our results show a huge but untapped potential of intelligence in tackling the challenges of fog computing.

Analytics based on big data computing can benefit today's banking and financial organizations on many aspects, and provide much valuable information for organizations to achieve more intelligent trading, which can help them to gain a great competitive advantage. However, the large scale of data and the critical latency analytics requirement in finance poses a great challenge for current system architecture. In this paper, we first analyze the challenges brought by the financial latency critical big data computing, then propose a discussion on how to handle these challenges from a perspective of multi-level system. We also talk about current researches on low latency in different system levels. The discussions and conclusions in the paper can be useful to the banking and financial organizations with the critical latency requirement of big data analytics.