arXiv subjects

Lan Dang

Publications and source records attributed to Lan Dang.

3 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↗

arXiv2024

How Homogenizing the Channel-wise Magnitude Can Enhance EEG Classification Model?

A significant challenge in the electroencephalogram EEG lies in the fact that current data representations involve multiple electrode signals, resulting in data redundancy and dominant lead information. However extensive research conducted on EEG classification focuses on designing model architectures without tackling the underlying issues. Otherwise, there has been a notable gap in addressing data preprocessing for EEG, leading to considerable computational overhead in Deep Learning (DL) processes. In light of these issues, we propose a simple yet effective approach for EEG data pre-processing. Our method first transforms the EEG data into an encoded image by an Inverted Channel-wise Magnitude Homogenization (ICWMH) to mitigate inter-channel biases. Next, we apply the edge detection technique on the EEG-encoded image combined with skip connection to emphasize the most significant transitions in the data while preserving structural and invariant information. By doing so, we can improve the EEG learning process efficiently without using a huge DL network. Our experimental evaluations reveal that we can significantly improve (i.e., from 2% to 5%) over current baselines.

eess.SP↗