arXiv subjects

Jaehoon Oh

Publications and source records attributed to Jaehoon Oh.

At least 19 recordsLinked to original sources

arXiv2025

OrthoRank: Token Selection via Sink Token Orthogonality for Efficient LLM inference

Attention mechanisms are central to the success of large language models (LLMs), enabling them to capture intricate token dependencies and implicitly assign importance to each token. Recent studies have revealed the sink token, which receives disproportionately high attention despite their limited semantic role. In this paper, we first expand the relationship between the sink token and other tokens, moving beyond attention to explore their similarity in hidden states, considering the layer depth. We observe that as the layers get deeper, the cosine similarity between the normalized hidden states of the sink token and those of other tokens increases, and that the normalized hidden states of the sink token exhibit negligible changes. These imply that other tokens consistently are directed toward the sink token throughout the layers. Next, we propose a dynamic token selection method, called OrthoRank, using these findings to select important tokens. Specifically, in a certain layer, we define token importance by the speed at which the token moves toward the sink token. This is converted into orthogonality with the sink token, meaning that tokens that are more orthogonal to the sink token are assigned greater importance. Finally, through extensive experiments, we demonstrated that our method results in lower perplexity and higher zero-shot accuracy compared to layer pruning methods at the same sparsity ratio with comparable throughput, while also achieving superior performance on LongBench.

cs.CL↗

arXiv2024

House of Cards: Massive Weights in LLMs

Massive activations, which manifest in specific feature dimensions of hidden states, introduce a significant bias in large language models (LLMs), leading to an overemphasis on the corresponding token. In this paper, we identify that massive activations originate not from the hidden state but from the intermediate state of a feed-forward network module in an early layer. Expanding on the previous observation that massive activations occur only in specific feature dimensions, we dive deep into the weights that cause massive activations. Specifically, we define top-$k$ massive weights as the weights that contribute to the dimensions with the top-$k$ magnitudes in the intermediate state. When these massive weights are set to zero, the functionality of LLMs is entirely disrupted. However, when all weights except for massive weights are set to zero, it results in a relatively minor performance drop, even though a much larger number of weights are set to zero. This implies that during the pre-training process, learning is dominantly focused on massive weights. Building on this observation, we propose a simple plug-and-play method called MacDrop (massive weights curriculum dropout), to rely less on massive weights during parameter-efficient fine-tuning. This method applies dropout to the pre-trained massive weights, starting with a high dropout probability and gradually decreasing it as fine-tuning progresses. Through various experiments, including zero-shot downstream tasks, long-context tasks, and ablation studies, we demonstrate that \texttt{MacDrop} generally improves performance and strengthens robustness.

cs.LG↗

arXiv2024

BAPO: Base-Anchored Preference Optimization for Overcoming Forgetting in Large Language Models Personalization

While learning to align Large Language Models (LLMs) with human preferences has shown remarkable success, aligning these models to meet the diverse user preferences presents further challenges in preserving previous knowledge. This paper examines the impact of personalized preference optimization on LLMs, revealing that the extent of knowledge loss varies significantly with preference heterogeneity. Although previous approaches have utilized the KL constraint between the reference model and the policy model, we observe that they fail to maintain general knowledge and alignment when facing personalized preferences. To this end, we introduce Base-Anchored Preference Optimization (BAPO), a simple yet effective approach that utilizes the initial responses of reference model to mitigate forgetting while accommodating personalized alignment. BAPO effectively adapts to diverse user preferences while minimally affecting global knowledge or general alignment. Our experiments demonstrate the efficacy of BAPO in various setups.

cs.AI↗

arXiv2023

FedFN: Feature Normalization for Alleviating Data Heterogeneity Problem in Federated Learning

Federated Learning (FL) is a collaborative method for training models while preserving data privacy in decentralized settings. However, FL encounters challenges related to data heterogeneity, which can result in performance degradation. In our study, we observe that as data heterogeneity increases, feature representation in the FedAVG model deteriorates more significantly compared to classifier weight. Additionally, we observe that as data heterogeneity increases, the gap between higher feature norms for observed classes, obtained from local models, and feature norms of unobserved classes widens, in contrast to the behavior of classifier weight norms. This widening gap extends to encompass the feature norm disparities between local and the global models. To address these issues, we introduce Federated Averaging with Feature Normalization Update (FedFN), a straightforward learning method. We demonstrate the superior performance of FedFN through extensive experiments, even when applied to pretrained ResNet18. Subsequently, we confirm the applicability of FedFN to foundation models.

cs.LG↗

arXiv2023

Cross-Modal Retrieval Meets Inference:Improving Zero-Shot Classification with Cross-Modal Retrieval

Contrastive language-image pre-training (CLIP) has demonstrated remarkable zero-shot classification ability, namely image classification using novel text labels. Existing works have attempted to enhance CLIP by fine-tuning on downstream tasks, but these have inadvertently led to performance degradation on unseen classes, thus harming zero-shot generalization. This paper aims to address this challenge by leveraging readily available image-text pairs from an external dataset for cross-modal guidance during inference. To this end, we propose X-MoRe, a novel inference method comprising two key steps: (1) cross-modal retrieval and (2) modal-confidence-based ensemble. Given a query image, we harness the power of CLIP's cross-modal representations to retrieve relevant textual information from an external image-text pair dataset. Then, we assign higher weights to the more reliable modality between the original query image and retrieved text, contributing to the final prediction. X-MoRe demonstrates robust performance across a diverse set of tasks without the need for additional training, showcasing the effectiveness of utilizing cross-modal features to maximize CLIP's zero-shot ability.

cs.CV↗

arXiv2023

FedSOL: Stabilized Orthogonal Learning with Proximal Restrictions in Federated Learning

Federated Learning (FL) aggregates locally trained models from individual clients to construct a global model. While FL enables learning a model with data privacy, it often suffers from significant performance degradation when clients have heterogeneous data distributions. This data heterogeneity causes the model to forget the global knowledge acquired from previously sampled clients after being trained on local datasets. Although the introduction of proximal objectives in local updates helps to preserve global knowledge, it can also hinder local learning by interfering with local objectives. To address this problem, we propose a novel method, Federated Stabilized Orthogonal Learning (FedSOL), which adopts an orthogonal learning strategy to balance the two conflicting objectives. FedSOL is designed to identify gradients of local objectives that are inherently orthogonal to directions affecting the proximal objective. Specifically, FedSOL targets parameter regions where learning on the local objective is minimally influenced by proximal weight perturbations. Our experiments demonstrate that FedSOL consistently achieves state-of-the-art performance across various scenarios.

cs.LG↗

arXiv2022

Synergy with Translation Artifacts for Training and Inference in Multilingual Tasks

Translation has played a crucial role in improving the performance on multilingual tasks: (1) to generate the target language data from the source language data for training and (2) to generate the source language data from the target language data for inference. However, prior works have not considered the use of both translations simultaneously. This paper shows that combining them can synergize the results on various multilingual sentence classification tasks. We empirically find that translation artifacts stylized by translators are the main factor of the performance gain. Based on this analysis, we adopt two training methods, SupCon and MixUp, considering translation artifacts. Furthermore, we propose a cross-lingual fine-tuning algorithm called MUSC, which uses SupCon and MixUp jointly and improves the performance. Our code is available at https://github.com/jongwooko/MUSC.

cs.CL↗

arXiv2022

Demystifying the Base and Novel Performances for Few-shot Class-incremental Learning

Few-shot class-incremental learning (FSCIL) has addressed challenging real-world scenarios where unseen novel classes continually arrive with few samples. In these scenarios, it is required to develop a model that recognizes the novel classes without forgetting prior knowledge. In other words, FSCIL aims to maintain the base performance and improve the novel performance simultaneously. However, there is little study to investigate the two performances separately. In this paper, we first decompose the entire model into four types of parameters and demonstrate that the tendency of the two performances varies greatly with the updated parameters when the novel classes appear. Based on the analysis, we propose a simple method for FSCIL, coined as NoNPC, which uses normalized prototype classifiers without further training for incremental novel classes. It is shown that our straightforward method has comparable performance with the sophisticated state-of-the-art algorithms.

cs.LG↗

arXiv2022

How to Fine-tune Models with Few Samples: Update, Data Augmentation, and Test-time Augmentation

Most of the recent few-shot learning (FSL) algorithms are based on transfer learning, where a model is pre-trained using a large amount of source data, and the pre-trained model is fine-tuned using a small amount of target data. In transfer learning-based FSL, sophisticated pre-training methods have been widely studied for universal representation. Therefore, it has become more important to utilize the universal representation for downstream tasks, but there are few studies on fine-tuning in FSL. In this paper, we focus on how to transfer pre-trained models to few-shot downstream tasks from the three perspectives: update, data augmentation, and test-time augmentation. First, we compare the two popular update methods, full fine-tuning (i.e., updating the entire network, FT) and linear probing (i.e., updating only a linear classifier, LP). We find that LP is better than FT with extremely few samples, whereas FT outperforms LP as training samples increase. Next, we show that data augmentation cannot guarantee few-shot performance improvement and investigate the effectiveness of data augmentation based on the intensity of augmentation. Finally, we adopt augmentation to both a support set for update (i.e., data augmentation) as well as a query set for prediction (i.e., test-time augmentation), considering support-query distribution shifts, and improve few-shot performance. The code is available at https://github.com/kimyuji/updating_FSL.

cs.LG↗

arXiv2022

ReFine: Re-randomization before Fine-tuning for Cross-domain Few-shot Learning

Cross-domain few-shot learning (CD-FSL), where there are few target samples under extreme differences between source and target domains, has recently attracted huge attention. Recent studies on CD-FSL generally focus on transfer learning based approaches, where a neural network is pre-trained on popular labeled source domain datasets and then transferred to target domain data. Although the labeled datasets may provide suitable initial parameters for the target data, the domain difference between the source and target might hinder fine-tuning on the target domain. This paper proposes a simple yet powerful method that re-randomizes the parameters fitted on the source domain before adapting to the target data. The re-randomization resets source-specific parameters of the source pre-trained model and thus facilitates fine-tuning on the target domain, improving few-shot performance.

cs.CV↗

arXiv2022

Understanding Cross-Domain Few-Shot Learning Based on Domain Similarity and Few-Shot Difficulty

Cross-domain few-shot learning (CD-FSL) has drawn increasing attention for handling large differences between the source and target domains--an important concern in real-world scenarios. To overcome these large differences, recent works have considered exploiting small-scale unlabeled data from the target domain during the pre-training stage. This data enables self-supervised pre-training on the target domain, in addition to supervised pre-training on the source domain. In this paper, we empirically investigate which pre-training is preferred based on domain similarity and few-shot difficulty of the target domain. We discover that the performance gain of self-supervised pre-training over supervised pre-training becomes large when the target domain is dissimilar to the source domain, or the target domain itself has low few-shot difficulty. We further design two pre-training schemes, mixed-supervised and two-stage learning, that improve performance. In this light, we present six findings for CD-FSL, which are supported by extensive experiments and analyses on three source and eight target benchmark datasets with varying levels of domain similarity and few-shot difficulty. Our code is available at https://github.com/sungnyun/understanding-cdfsl.

cs.LG↗

arXiv2021

TornadoAggregate: Accurate and Scalable Federated Learning via the Ring-Based Architecture

Federated learning has emerged as a new paradigm of collaborative machine learning; however, many prior studies have used global aggregation along a star topology without much consideration of the communication scalability or the diurnal property relied on clients' local time variety. In contrast, ring architecture can resolve the scalability issue and even satisfy the diurnal property by iterating nodes without an aggregation. Nevertheless, such ring-based algorithms can inherently suffer from the high-variance problem. To this end, we propose a novel algorithm called TornadoAggregate that improves both accuracy and scalability by facilitating the ring architecture. In particular, to improve the accuracy, we reformulate the loss minimization into a variance reduction problem and establish three principles to reduce variance: Ring-Aware Grouping, Small Ring, and Ring Chaining. Experimental results show that TornadoAggregate improved the test accuracy by up to 26.7% and achieved near-linear scalability.

cs.LG↗

arXiv2021

FedBABU: Towards Enhanced Representation for Federated Image Classification

Federated learning has evolved to improve a single global model under data heterogeneity (as a curse) or to develop multiple personalized models using data heterogeneity (as a blessing). However, little research has considered both directions simultaneously. In this paper, we first investigate the relationship between them by analyzing Federated Averaging at the client level and determine that a better federated global model performance does not constantly improve personalization. To elucidate the cause of this personalization performance degradation problem, we decompose the entire network into the body (extractor), which is related to universality, and the head (classifier), which is related to personalization. We then point out that this problem stems from training the head. Based on this observation, we propose a novel federated learning algorithm, coined FedBABU, which only updates the body of the model during federated training (i.e., the head is randomly initialized and never updated), and the head is fine-tuned for personalization during the evaluation process. Extensive experiments show consistent performance improvements and an efficient personalization of FedBABU. The code is available at https://github.com/jhoon-oh/FedBABU.

cs.LG↗

arXiv2021

Comparing Kullback-Leibler Divergence and Mean Squared Error Loss in Knowledge Distillation

Knowledge distillation (KD), transferring knowledge from a cumbersome teacher model to a lightweight student model, has been investigated to design efficient neural architectures. Generally, the objective function of KD is the Kullback-Leibler (KL) divergence loss between the softened probability distributions of the teacher model and the student model with the temperature scaling hyperparameter tau. Despite its widespread use, few studies have discussed the influence of such softening on generalization. Here, we theoretically show that the KL divergence loss focuses on the logit matching when tau increases and the label matching when tau goes to 0 and empirically show that the logit matching is positively correlated to performance improvement in general. From this observation, we consider an intuitive KD loss function, the mean squared error (MSE) between the logit vectors, so that the student model can directly learn the logit of the teacher model. The MSE loss outperforms the KL divergence loss, explained by the difference in the penultimate layer representations between the two losses. Furthermore, we show that sequential distillation can improve performance and that KD, particularly when using the KL divergence loss with small tau, mitigates the label noise. The code to reproduce the experiments is publicly available online at https://github.com/jhoon-oh/kd_data/.

cs.LG↗

arXiv2021

BOIL: Towards Representation Change for Few-shot Learning

Model Agnostic Meta-Learning (MAML) is one of the most representative of gradient-based meta-learning algorithms. MAML learns new tasks with a few data samples using inner updates from a meta-initialization point and learns the meta-initialization parameters with outer updates. It has recently been hypothesized that representation reuse, which makes little change in efficient representations, is the dominant factor in the performance of the meta-initialized model through MAML in contrast to representation change, which causes a significant change in representations. In this study, we investigate the necessity of representation change for the ultimate goal of few-shot learning, which is solving domain-agnostic tasks. To this aim, we propose a novel meta-learning algorithm, called BOIL (Body Only update in Inner Loop), which updates only the body (extractor) of the model and freezes the head (classifier) during inner loop updates. BOIL leverages representation change rather than representation reuse. This is because feature vectors (representations) have to move quickly to their corresponding frozen head vectors. We visualize this property using cosine similarity, CKA, and empirical results without the head. BOIL empirically shows significant performance improvement over MAML, particularly on cross-domain tasks. The results imply that representation change in gradient-based meta-learning approaches is a critical component.

cs.LG↗

arXiv2020

Accurate and Fast Federated Learning via IID and Communication-Aware Grouping

Federated learning has emerged as a new paradigm of collaborative machine learning; however, it has also faced several challenges such as non-independent and identically distributed(IID) data and high communication cost. To this end, we propose a novel framework of IID and communication-aware group federated learning that simultaneously maximizes both accuracy and communication speed by grouping nodes based on data distributions and physical locations of the nodes. Furthermore, we provide a formal convergence analysis and an efficient optimization algorithm called FedAvg-IC. Experimental results show that, compared with the state-of-the-art algorithms, FedAvg-IC improved the test accuracy by up to 22.2% and simultaneously reduced the communication time to as small as 12%.

cs.LG↗

arXiv2020

SIPA: A Simple Framework for Efficient Networks

With the success of deep learning in various fields and the advent of numerous Internet of Things (IoT) devices, it is essential to lighten models suitable for low-power devices. In keeping with this trend, MicroNet Challenge, which is the challenge to build efficient models from the view of both storage and computation, was hosted at NeurIPS 2019. To develop efficient models through this challenge, we propose a framework, coined as SIPA, consisting of four stages: Searching, Improving, Pruning, and Accelerating. With the proposed framework, our team, OSI AI, compressed 334x the parameter storage and 357x the math operation compared to WideResNet-28-10 and took 4th place in the CIFAR-100 track at MicroNet Challenge 2019 with the top 10% highly efficient computation. Our source code is available from https://github.com/Lee-Gihun/MicroNet_OSI-AI.

cs.LG↗

arXiv2018

Spectrogram-channels u-net: a source separation model viewing each channel as the spectrogram of each source

Sound source separation has attracted attention from Music Information Retrieval(MIR) researchers, since it is related to many MIR tasks such as automatic lyric transcription, singer identification, and voice conversion. In this paper, we propose an intuitive spectrogram-based model for source separation by adapting U-Net. We call it Spectrogram-Channels U-Net, which means each channel of the output corresponds to the spectrogram of separated source itself. The proposed model can be used for not only singing voice separation but also multi-instrument separation by changing only the number of output channels. In addition, we propose a loss function that balances volumes between different sources. Finally, we yield performance that is state-of-the-art on both separation tasks.

cs.SD↗