arXiv subjects

Junhyuk Jo

Publications and source records attributed to Junhyuk Jo.

5 recordsLinked to original sources

arXiv2026

Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs

Layer pruning removes entire Transformer decoder blocks from large language models, but introduces a mismatch between the hidden state received by the next surviving layer and the distribution it was trained to process, leading to significant performance degradation. We propose Ghosted Layers, a training-free recovery module that addresses this issue by solving a boundary activation alignment problem. Our method derives a closed-form optimal linear operator from a small calibration set to reconstruct the activation discrepancy introduced by the pruned layers. We show that this solution corresponds to the unconstrained optimum of the alignment objective, whereas existing methods are restricted to constrained solutions over limited operator subspaces. Experiments across multiple LLM backbones and pruning strategies demonstrate that our method consistently improves accuracy and perplexity over prior training-free baselines, while preserving the efficiency gains of layer pruning. Official code repository: https://github.com/daniel-eai/ghosted_layers_official_repository/.

cs.LG↗

arXiv2026

Weight Concentration Regularization for Improving Pruning Robustness Under High Sparsity

Deep neural networks achieve outstanding performance across vision and language tasks, yet their large parameter counts limit deployment in resource-constrained settings. One-shot pruning reduces model size without retraining, but models trained with standard objectives often suffer substantial accuracy drops under aggressive sparsity. Prior work mitigates this drop along two directions: regularizers such as $\ell_1$ and DeepHoyer that shape the weight distribution during training, and pruning-robust optimizers such as SAM, CrAM, and S$^2$SAM that flatten the loss landscape. However, existing regularizers either shrink all weights uniformly ($\ell_1$) or induce scale-invariant sparsity (DeepHoyer), without concentrating weight energy onto a small set of informative parameters. We propose a Weight Concentration Regularizer (WCR), a training-time regularizer that amplifies the magnitude of a small subset of parameters while driving the remainder toward zero, so that magnitude pruning predominantly removes parameters with negligible functional contribution. We provide a convergence analysis and evaluate WCR on LLM fine-tuning, image classification, and medical segmentation, demonstrating consistent improvements in pruning robustness across architectures and compatibility with existing pruning-robust optimizers.

cs.LG↗

arXiv2026

Biased Local SGD for Efficient Deep Learning on Heterogeneous Systems

Most parallel neural network training methods assume homogeneous computing resources. For example, synchronous data-parallel SGD suffers from significant synchronization overhead under heterogeneous workloads, often forcing practitioners to rely only on the fastest devices (e.g., GPUs). In this work, we study local SGD for efficient parallel training on heterogeneous systems. We show that intentionally introducing bias in data sampling and model aggregation can effectively harmonize slower CPUs with faster GPUs. Our extensive empirical results demonstrate that a carefully controlled bias significantly accelerates local SGD while achieving comparable or even higher accuracy than synchronous SGD under the same epoch budget. For instance, our method trains ResNet20 on CIFAR-10 with 2 CPUs and 8 GPUs up to 32x faster than synchronous SGD, with nearly identical accuracy. These results provide practical insights into how to flexibly utilize diverse compute resources for deep learning.

cs.LG↗

arXiv2026

Enabling Weak Client Participation via On-device Knowledge Distillation in Heterogeneous Federated Learning

Online Knowledge Distillation (KD) is recently highlighted to train large models in Federated Learning (FL) environments. Many existing studies adopt the logit ensemble method to perform KD on the server side. However, they often assume that unlabeled data collected at the edge is centralized on the server. Moreover, the logit ensemble method personalizes local models, which can degrade the quality of soft targets, especially when data is highly non-IID. To address these critical limitations,we propose a novel on-device KD-based heterogeneous FL method. Our approach leverages a small auxiliary model to learn from labeled local data. Subsequently, a subset of clients with strong system resources transfers knowledge to a large model through on-device KD using their unlabeled data. Our extensive experiments demonstrate that our on-device KD-based heterogeneous FL method effectively utilizes the system resources of all edge devices as well as the unlabeled data, resulting in higher accuracy compared to SOTA KD-based FL methods.

cs.LG↗

arXiv2025

Asynchronous Sharpness-Aware Minimization For Fast and Accurate Deep Learning

Sharpness-Aware Minimization (SAM) is an optimization method that improves generalization performance of machine learning models. Despite its superior generalization, SAM has not been actively used in real-world applications due to its expensive computational cost. In this work, we propose a novel asynchronous-parallel SAM which achieves nearly the same gradient norm penalizing effect like the original SAM while breaking the data dependency between the model perturbation and the model update. The proposed asynchronous SAM can even entirely hide the model perturbation time by adjusting the batch size for the model perturbation in a system-aware manner. Thus, the proposed method enables to fully utilize heterogeneous system resources such as CPUs and GPUs. Our extensive experiments well demonstrate the practical benefits of the proposed asynchronous approach. E.g., the asynchronous SAM achieves comparable Vision Transformer fine-tuning accuracy (CIFAR-100) as the original SAM while having almost the same training time as SGD.

cs.LG↗