arXiv ScienceSearch

arXiv subjects

Peter Lee

Publications and source records attributed to Peter Lee.

16 recordsLinked to original sources

ATRACT: A Trustworthy Robotic Autonomous system to support Casualty Triage

At a time when drones are increasingly associated with hostile operations, we re-purpose them for humanitarian and life-saving applications. However, adapting search and rescue drones for battlefield triage remains extremely challenging; the technology must perform reliably to support frontline medics who are forced to operate under extreme uncertainty, restricted access, and significant personal risk. Due to growing vulnerabilities of casualty evacuation in conflicting zones, this paper presents ATRACT (A Trustworthy Robotic Autonomous system to support Casualty Triage), a novel human-in-the-loop decision support system to enable early battlefield triage during the critical post-trauma period. ATRACT integrates drone-captured video with wearable sensor input for multi-modal learning to support casualty-state assessment, thereby addressing the limitations of existing systems. Drone video captures fine-grained behavioural cues, such as pose, posture, while body-worn sensors provide complementary physiological signals, including heart rate, breathing rate, and movement. By combining two modalities, ATRACT provides evidence to support the early judgement of medics when direct access to the casualty is delayed, risky, or restricted. To mitigate the data realism gap pertaining to injured actions, a conditional variational autoencoder is devised for data augmentation. Experimental results on our drone captured dataset show that proposed pipeline achieves 85.7% accuracy for action classification; while our lightweight CNN visual encoder remains competitive with stronger pre-trained video backbones. Overall, the results support ATRACT as a practically meaningful step towards remote triage in contested environments, where multi-modal sensing, human oversight and trustworthy decision support can improve casualty prioritisation, and lessen the exposure of frontline medics.

cs.HC

Vision-Inspired Image Quality Assessment for Radar-Based Human Activity Representations

Radar-based human activity recognition has gained attention as a privacy-preserving alternative to vision and wearable sensors, especially in sensitive environments like long-term care facilities. Micro-Doppler spectrograms derived from FMCW radar signals are central to recognizing dynamic activities, but their effectiveness is limited by noise and clutter. In this work, we use a benchmark radar dataset to reimplement and assess three recent denoising and preprocessing techniques: adaptive preprocessing, adaptive thresholding, and entropy-based denoising. To illustrate the shortcomings of conventional metrics in low-SNR regimes, we evaluate performance using both perceptual image quality measures and standard error-based metrics. We additionally propose a novel framework for static activity recognition using range-angle feature maps to expand HAR beyond dynamic activities. We present two important contributions: a temporal tracking algorithm to enforce consistency and a no-reference quality scoring algorithm to assess RA-map fidelity. According to experimental findings, our suggested techniques enhance classification performance and interpretability for both dynamic and static activities, opening the door for more reliable radar-based HAR systems.

eess.IV

A global log for medical AI

Modern computer systems rely on syslog, a universal protocol that records critical events across heterogeneous infrastructure. Medicine's rapidly growing AI stack has no equivalent. As medicine deploys AI tools at scale, there is no standard way to record how, when, by whom, and for whom these models are used. Without such records, it is difficult to measure real-world performance and outcomes, detect adverse events, or identify bias and dataset drift. Here we introduce MedLog, a protocol for event-level logging of medical AI. Each time an AI model interacts with a human, another algorithm, or an automated workflow, MedLog creates a record. Each record contains nine core fields: header, model, user, target, inputs, artifacts, outputs, outcomes, and feedback. We apply MedLog across four deployments in the US, Switzerland, and Vietnam: ICU deterioration prediction, tetanus progression monitoring from wearable signals, automated sepsis quality reporting, and patient attendance prediction. MedLog records capture model behavior, workflow interactions, and downstream outcomes, including AI performance degradation during severe weather events in patient attendance prediction and increased laboratory testing after ICU deterioration alerts. MedLog limits the data footprint through risk-based sampling, lifecycle-aware retention policies, and write-behind caching, enabling deployment in low-resource settings. It also supports detailed traces for complex, agentic, or multi-stage workflows, creating a foundation for continuous monitoring, auditing, and improvement of medical AI.

cs.AI

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions.

cs.CL

Collaborative effects of wavefront shaping and optical clearing agent in optical coherence tomography

We demonstrate that simultaneous application of optical clearing agents (OCAs) and complex wavefront shaping in optical coherence tomography (OCT) can provide significant enhancement of the penetration depth and imaging quality. OCA reduces optical inhomogeneity of a highly scattering sample, and the wavefront shaping of illumination light controls multiple scattering, resulting in an enhancement of the penetration depth and signal-to-noise ratio. A tissue phantom study shows that concurrent applications of OCA and wavefront shaping successfully operate in OCT imaging. The penetration depth enhancement is further demonstrated for ex vivo mouse ears, revealing hidden structures inaccessible with conventional OCT imaging.

physics.optics

On the Action of the Symmetric Group on the Cohomology of Groups Related to (Virtual) Braids

In this paper we consider the cohomology of four groups related to the virtual braids of [Kauffman] and [Goussarov-Polyak-Viro], namely the pure and non-pure virtual braid groups (PvB_n and vB_n, respectively), and the pure and non-pure flat braid groups (PfB_n and fB_n, respectively). The cohomologies of PvB_n and PfB_n admit an action of the symmetric group S_n. We give a description of the cohomology modules H^i(PvB_n,Q) and H^i(PfB_n,Q) as sums of S_n-modules induced from certain one-dimensional representations of specific subgroups of S_n. This in particular allows us to conclude that H^i(PvB_n,Q) and H^i(PfB_n,Q) are uniformly representation stable, in the sense of [Church-Farb]. We also give plethystic formulas for the Frobenius characteristics of these S_n-modules. We then derive a number of constraints on which S_n irreducibles may appear in H^i(PvB_n,Q) and H^i(PfB_n,Q). In particular, we show that the multiplicity of the alternating representation in H^i(PvB_n,Q) and H^i(PfB_n,Q) is identical, and moreover is nil for sufficiently large $n$. We use this to recover the (previously known) fact that the multiplicity of the alternating representation in H^i(PB_n,Q) is nil (here PB_n is the ordinary pure braid group). We also give an explicit formula for H^i(vB_n,Q) and show that H^i(fB_n,Q)=0. Finally, we give Hilbert series for the character of the action of S_n on H^i(PvB_n,Q) and H^i(PfB_n,Q). An extension of the standard `Koszul formula' for the graded dimension of Koszul algebras to graded characters of Koszul algebras then gives Hilbert series for the graded characters of the respective quadratic dual algebras.

math.RT

Closed-Form Associators in Permutative Chord Diagrams

Construction of a universal finite-type invariant can be reduced, under suitable assumptions, to the solution of certain equations (the hexagon and pentagon equations) in a particular graded associative algebra of chord diagrams. An explicit, closed-form solution to these equations may, indirectly, give information about various interesting properties of knots, such as which knots are ribbon. However, while closed-form solutions (as opposed to solutions which can only be approximated to successively higher degrees) are needed for this purpose, such solutions have proven elusive, partly as a result of the non-commutative nature of the algebra. To make the problem more tractable, we restrict our attention to solutions of the equations in the subalgebra of horizontal chord diagrams, viewed as a graded unital permutative algebra -- where `permutative' means that u[x,y]=0 whenever u has degree at least 1. We show that this restriction leads in a straightforward and fairly short way to a reduction of the hexagon and pentagon equations to a simpler equation taken over the algebra of power series in two commuting variables. This equation had been found and solved explicitly by Kurlin under a superficially different set of reduction assumptions, which we show here are in fact equivalent to ours. This paper thus provides an alternative (simpler and shorter) derivation of Kurlin's equation.

math.QA

The Pure Virtual Braid Group Is Quadratic

If an augmented algebra K over Q is filtered by powers of its augmentation ideal I, the associated graded algebra grK need not in general be quadratic: although it is generated in degree 1, its relations may not be generated by homogeneous relations of degree 2. In this paper we give a sufficient criterion (called the PVH Criterion) for grK to be quadratic. When K is the group algebra of a group G, quadraticity is known to be equivalent to the existence of a (not necessarily homomorphic) universal finite type invariant for G. Thus the PVH Criterion also implies the existence of such a universal finite type invariant for the group G. We apply the PVH Criterion to the group algebra of the pure virtual braid group (also known as the quasi-triangular group), and show that the corresponding associated graded algebra is quadratic, and hence that these groups have a (not necessarily homomorphic) universal finite type invariant.

math.QA

Pixel multiplexing for high-speed multi-resolution fluorescence imaging

We introduce a imaging modality that works by transiently masking image-subregions during a single exposure of a CCD frame. By offsetting subregion exposure time, temporal information is embedded within each stored frame, allowing simultaneous acquisition of a full high spatial resolution image and a high-speed image sequence without increasing bandwidth. The technique is demonstrated by imaging calcium transients in heart cells at 250 Hz with a 10 Hz megapixel camera.

q-bio.QM

A Note on Cubic Interactions in PP-Wave Light Cone String Field Theory

We study the string modes in the pp-wave light-cone string field theory. First, we clarify the discrepancy between the Neumann coefficients for the supergravity vertex and the zero mode of the full string one. We also repeat our previous manipulation of the prefactor for the string modes and find that the prefactor reduces to the energy difference of the cos modes minus that of the sin modes. Finally, we discuss off-shell three-string processes.

hep-th

Cubic Interactions in PP-Wave Light Cone String Field Theory

We use the supergravity modes to clarify the role of the prefactor in the light-cone superstring field theory on PP-wave background. We verify some of the proposals of the recent paper hep-th/0205089 and give further evidence for the correspondence between N=4 SYM gauge theory and string theory on PP-wave. We also consider energy-preserving processes and find that they give vanishing cubic interaction Hamiltonian matrix.

hep-th

Open Strings in PP-Wave Background from Defect Conformal Field Theory

We consider open strings ending on a D5-brane in the pp-wave background, which is realized in the Penrose limit of $AdS_5 \times S^5$ with an $AdS_4\times S^2$ brane. A complete list of gauge invariant operators in the defect conformal field theory is constructed which is dual to the open string states.

hep-th

A Note on Noncommutative D-Brane Actions

We use the nonabelian action of N coincident D(-1) branes in constant background fields, in the N $\to \infty$ limit, to construct noncommutative D-brane actions in an arbitrary noncommutative description and comment on tachyon condensation from this perspective.

hep-th

Boundary States for AdS_2 Branes in AdS_3

We construct boundary states for the AdS_2 D-branes in AdS_3. We show that, in the semi-classical limit, the boundary states correctly reproduce geometric configurations of these branes. We use the boundary states to compute the one loop free energy of open string stretched between the branes. The result agrees precisely with the open string computation in hep-th/0106129.

hep-th

Open Strings on AdS_2 Branes

We study the spectrum of open strings on AdS_2 branes in AdS_3 in an NS-NS background, using the SL(2,R) WZW model. When the brane carries no fundamental string charge, the open string spectrum is the holomorphic square root of the spectrum of closed strings in AdS_3. It contains short and long strings, and is invariant under spectral flow. When the brane carries fundamental string charge, the open string spectrum again contains short and long strings in all winding sectors. However, branes with fundamental string charge break half the spectral flow symmetry. This has different implications for short and long strings. As the fundamental string charge increases, the brane approaches the boundary of AdS_3. In this limit, the induced electric field on the worldvolume reaches its critical value, producing noncommutative open string theory on AdS_2.

hep-th