arXiv subjects

Thang Tran

Publications and source records attributed to Thang Tran.

4 recordsLinked to original sources

arXiv2026

Numbat: Building and Verifying a Self-Contained Machine-Learning Stack

Machine-learning systems are built almost exclusively on a few large Python-orchestrated frameworks, and they inherit those stacks' engineering costs: environments of hundreds of version-coupled packages, separate export toolchains for deployment, and the split between the language research is written in and the language products ship in. We report on the construction and verification of numbat, a machine-learning stack written in one general-purpose language (Zig) with no third-party runtime dependencies. The stack spans tensor computation, automatic differentiation, neural-network modules, mixed precision, multi-GPU training, data loading and monitoring; an SDK exposes it behind a stable, additively versioned C ABI of over 1,400 entry points, with bindings for six languages; and its clinical domain planes encode regulatory requirements as executable acceptance gates rather than documentation. Verifying such a stack is the harder half of building it: a defective training run rarely fails, it converges quietly to a slightly worse model. We treat a widely used reference implementation as an executable specification and verify against it at five levels, from operator gradient checks to an automated trajectory gate against a same-machine reference run - the arrangement our companion study formalizes as a trajectory-level differential oracle. The protocol surfaced ten silent recipe divergences, which we catalog with mechanisms and symptoms. As the acceptance test, we train a 25.9M-parameter detector of the YOLOv8m class from random initialization on COCO 2017 for the full 500-epoch schedule: the exported weights score 0.4956 mAP50-95 under the official protocol, scored by the reference stack's own validator (published endpoint 0.502), with single-GPU step time at parity on identical hardware. Weights, per-epoch metrics and the full run manifest are released.

cs.SE↗

arXiv2026

Cross-Stack Validation of Language-Model Training: A Clinical Fine-Tuning Case Study

Neural network training has an oracle problem: a run can converge normally and yield a usable model while the software beneath it computes something other than specified. Almost all such work runs on one stack, so there is rarely anything independent to check against. We study whether independently implemented training stacks can serve as differential oracles for a whole fine-tuning pipeline, rather than the operators and inference paths that prior differential testing targets. We define a trajectory-level protocol -- a shared specification, cross-check points spanning arithmetic, model loading, data rendering and the learning trajectory, and a separation of independence of the stack, the orchestration and the language runtime -- and apply it to a LoRA adaptation of Qwen3-0.6B over 168,574 clinical question-answer pairs under PyTorch and under numbat, an independent framework written in Zig, driven natively and through its C interface from six languages. Across 42 paired evaluations spanning a full epoch the two stacks' held-out cross-entropy differs by 0.134% on average, and four implementations end the epoch within 0.15% of one another. The comparison exposed 17 faults that single-implementation development had missed, two of them notable for software engineering. The fault with the largest effect on the trained model lay outside the numerical kernels: a mismatch in how clinical text was rendered moved held-out loss 0.15, some 500 times more than the arithmetic faults found beside it. And four faults were reachable only from a language whose memory model differs from the first two implementations: a scheduler migrating work across threads, a collector blind to device memory, an ownership discipline needing a primitive the interface lacked. Implementation diversity has several axes, and the runtime is one.

cs.SE↗

arXiv2021

Acoustic Cloak Design via Machine Learning

Acoustic metamaterials are engineered microstructures with special mechanical and acoustic properties enabling exotic effects such as wave steering, focusing and cloaking. The design of acoustic cloaks using scattering cancellation has traditionally involved the optimization of metamaterial structure based on direct computer simulations of the total scattering cross section (TSCS) for a large number of configurations. Here, we work with sets of cylindrical objects confined in a region of space and use machine learning methods to streamline the design of 2D configurations of scatterers with minimal TSCS demonstrating cloaking effect at discrete sets of wavenumbers. After establishing that artificial neural networks are capable of learning the TSCS based on the location of cylinders, we develop an inverse design algorithm, combining variational autoencoders and the Gaussian process, for predicting optimal arrangements of scatterers given the TSCS. We show results for up to eight cylinders and discuss the efficiency and other advantages of the machine learning approach.

cond-mat.dis-nn↗

arXiv2021

Real-Time ECG Interval Monitoring Using a Fully Disposable Wireless Patch Sensor

ECG interval monitoring provides key insights into the diagnosis of cardiac diseases. The standard 12-lead ECG is generally used, however, because of the current COVID-19 pandemic there is a strong need for a remote monitoring solution which will reduce exposure of health care providers to coronavirus. This article presents a disposable wireless patch biosensor (VitalPatch) and associated platform functionalities for real-time continuous measurement of clinically relevant ECG intervals including PR interval, QRS duration, QT interval, corrected QT interval by Bazett (QTb), and corrected QT interval by Fridericia (QTf). The performance of the VitalPatch is validated by comparing its automated algorithm interval measurements to the manually annotated global intervals of the 12-lead ECG device in 30 subjects. The accuracy of interval monitoring (in terms of mean timing error calculated by subtracting the VitalPatch measurements from the global intervals) is 2.7+/-15.94 ms, -1.97+/-12.29 ms, -14.6+/-12.97 ms, - 15.33+/-14.11 ms, and -15.08+/-13.69 ms for PR interval, QRS duration, QT interval, QTb, and QTf, respectively. These results demonstrate that the VitalPatch is a viable solution for measuring ECG intervals while taking advantage of its remote monitoring feature during the pandemic.

q-bio.QM↗