New method regularizes deep networks by distilling self-knowledge.
problem Overfitting in deep neural networks.
method Self-knowledge distillation to regularize class-wise predictions.
result Significant improvement in generalization and calibration.
Self-distillation improves model performance but can lead to underfitting.
problem Understanding why self-distillation improves model performance and its limitations.
method Theoretical analysis of self-distillation in Hilbert space with ℓ2 regularization. result Self-distillation modifies regularization by limiting the number of basis functions, potentially leading to underfitting.
Study on ensemble, distillation, and self-distillation in deep learning models.
problem Improving test accuracy in deep learning models using ensemble and distillation methods.
method Formal study of ensemble and distillation, considering multi-view data structure.
result Proven that ensemble and distillation can improve test accuracy in deep learning models, and the superior performance can be distilled into a single model.
A new method for training GNNs without a teacher model.
problem Training over-parameterized GNN models is difficult and inefficient.
method GNN Self-Distillation (GNN-SD) with NDR and ADR.
result Improves GNN performance with less training cost and better generalization.
Two approaches extend knowledge distillation to Gaussian Processes, showing relationships to existing methods.
problem Applying knowledge distillation to Gaussian Processes for regression and classification.
method Data-centric and distribution-centric approaches to extend distillation to GPR and GPC.
result Distribution-centric approach for GPC approximately corresponds to data duplication and scaling.
This paper improves ASR performance by aligning frames more accurately.
problem Disagreement between teacher-student models in frame-level alignment.
method Introduces self-knowledge distillation (SKD) to guide frame-level alignment.
result Improves both resource efficiency and performance in ASR.
Improved generalization with iterative self-distillation using weighted ground-truth targets.
problem Improving generalization accuracy in neural networks.
method Iterative kernel regression with weighted ground-truth targets and ℓ2 regularization. result Closed-form solution for optimal weighting parameter and efficient estimation.
Repeated self-distillation improves model performance significantly.
problem How much gain is possible by applying multiple steps of self-distillation?
method Investigated linear regression tasks, applied multiple steps of self-distillation, analyzed excess risk reduction.
result Multi-step self-distillation reduces excess risk by a factor as large as d, where d is the input dimension. Convolutional neural networks have been widely deployed in various application scenarios. In order to extend the applications' boundaries to some accuracy-crucial domains, researchers have been investigating approaches to boost accuracy through either deeper or wider network structures, which brings with them the expon…
Since deep learning became a key player in natural language processing (NLP), many deep learning models have been showing remarkable performances in a variety of NLP tasks, and in some cases, they are even outperforming humans. Such high performance can be explained by efficient knowledge representation of deep learnin…
SAIL improves graph node representation learning by distilling knowledge between graphs.
problem Improving graph node representation learning with GNNs in unsupervised scenarios.
method SAIL framework with intra- and inter-graph knowledge distillation.
result SAIL consistently outperforms state-of-the-art baselines on various benchmark datasets.
Proposes LsrKD and MrKD to improve neural network training performance.
problem Improving neural network training performance, especially on deep networks.
method Extends Label Smoothing Regularization with Self-Knowledge Distillation, introducing LsrKD and MrKD.
result LsrKD and MrKD significantly improve training performance on deep neural networks.
PS-KD distills a model's own knowledge to soften hard targets during training.
problem Improving generalization of deep neural networks by softening hard targets.
method Progressive self-knowledge distillation (PS-KD) that progressively distills a model's own knowledge to soften hard targets.
result PS-KD improves accuracy and provides high quality of confidence estimates in terms of calibration and ordinal ranking.
To solve deep neural network (DNN)'s huge training dataset and its high computation issue, so-called teacher-student (T-S) DNN which transfers the knowledge of T-DNN to S-DNN has been proposed. However, the existing T-S-DNN has limited range of use, and the knowledge of T-DNN is insufficiently transferred to S-DNN. To …
Distillation is a method to transfer knowledge from one model to another and often achieves higher accuracy with the same capacity. In this paper, we aim to provide a theoretical understanding on what mainly helps with the distillation. Our answer is "early stopping". Assuming that the teacher network is overparameteri…
In real-world scenarios, data tends to exhibit a long-tailed distribution, which increases the difficulty of training deep networks. In this paper, we propose a novel self-paced knowledge distillation framework, termed Learning From Multiple Experts (LFME). Our method is inspired by the observation that networks traine…
Recent variants improve knowledge distillation performance.
problem Improving the performance of knowledge distillation.
method Introducing additional components or changing the learning process.
result These variants have shown promising results.
SPEQ improves quantized neural networks by stochastic precision sharing and cosine similarity loss.
problem Improving quantized deep neural networks for edge devices.
method SPEQ combines stochastic precision sharing and cosine similarity loss for knowledge distillation.
result SPEQ outperforms existing methods in various tasks.
SFAVEL distills features from language models for fact verification without annotations.
problem Fact verification with semantically meaningful and compact features.
method Self-supervised pretraining using contrastive loss with language models.
result Achieved state-of-the-art results on FB15k-237 and FEVER.
Survey of knowledge distillation for resource-limited devices.
problem Deploying large deep learning models on resource-limited devices.
method Knowledge distillation using a smaller model trained with information from a larger model.
result A new metric (distillation metric) for comparing different knowledge distillation algorithms.
This study improves knowledge distillation for RNN-T models with noisy labels.
problem Challenges in distilling knowledge from RNN-T models with variable quality teachers.
method Full-sum distillation and sequence-level knowledge distillation.
result Full-sum distillation outperforms other methods for RNN-T models, especially for bad teachers.
Improved self-distillation reduces label noise and enhances model accuracy.
problem Label noise in multi-class classification.
method Label averaging and refined partial labels.
result Single-round self-distillation achieves comparable performance to multi-round distillation.
Self-augmentation improves deep networks for few-shot learning with minimal training data.
problem Improving deep networks' generalization to unseen classes with limited training examples.
method Self-augmentation using self-mix and self-distillation techniques, combined with regional dropout and local representation learning.
result The method outperforms state-of-the-art few-shot learning methods on prevalent benchmarks.
Self-distillation optimally improves model performance in spiked covariance models.
problem Improving model performance in spiked covariance models.
method Developed spectral shrinkage estimators and analyzed self-distillation.
result Self-distillation achieves optimal performance among spectral shrinkage estimators for spiked covariance matrices.
RID framework quantifies and regularizes task-relevant knowledge in distillation.
problem Distilling irrelevant information can hinder student model performance.
method Partial Information Decomposition to quantify and regularize task-relevant knowledge.
result RID framework leads to more resilient distillation under nuisance teachers.
The paper explains knowledge distillation by analyzing visual concepts in DNNs.
problem Understanding how knowledge distillation affects the learning of visual concepts in deep neural networks.
method The paper proposes three hypotheses and designs mathematical metrics to evaluate feature representations of DNNs.
result The hypotheses were verified through experiments on various DNNs.
ProSMIN improves representation quality through probabilistic self-supervised learning.
problem Improving representation quality in self-supervised learning.
method ProSMIN uses two neural networks, online and target, to learn diverse representations through knowledge distillation and a modified scoring rule loss function.
result ProSMIN achieves superior accuracy and calibration on various downstream tasks.
Knowledge distillation (KD) is a technique to derive optimal performance from a small student network (SN) by distilling knowledge of a large teacher network (TN) and transferring the distilled knowledge to the small SN. Since a role of convolutional neural network (CNN) in KD is to embed a dataset so as to perform a g…
CAKD framework optimizes knowledge transfer by focusing on influential components of distillation.
problem Balancing and optimizing knowledge transfer in distillation models.
method Decouple KL divergence into BCD, SCD, and WCD; prioritize influential components.
result CAKD framework consistently outperforms baseline across diverse models and datasets.
This paper improves understanding of knowledge distillation and its effects.
problem Improving model quality with a fixed capacity budget.
method Categorizing teacher's knowledge into three levels and studying their effects on distillation.
result The three hierarchical levels of teacher's knowledge (universe, domain, instance) significantly impact knowledge distillation.
Self-distillation improves model performance in noisy label settings.
problem Improving model accuracy in supervised learning with noisy labels.
method Analyzes self-distillation in two supervised learning problems with noisy labels, using theoretical and empirical approaches.
result Optimal self-distillation parameter is greater than 1 in high label noise regimes, outperforming traditional methods.
ProSelfLC improves robustness of deep neural networks by automatically deciding trust in predictions.
problem Training robust deep neural networks requires addressing issues like label noise and low entropy predictions.
method ProSelfLC progressively increases trust in predicted labels over time, considering entropy and learning time.
result ProSelfLC demonstrates improved robustness in both clean and noisy settings through empirical validation.
Often we wish to transfer representational knowledge from one neural network to another. Examples include distilling a large network into a smaller one, transferring knowledge from one sensory modality to a second, or ensembling a collection of models into a single estimator. Knowledge distillation, the standard approa…
A new distillation method transfers channel information from teacher to student.
problem Transfer knowledge from teacher to student with fewer parameters and calculations.
method Channel Distillation (CD) and Guided Knowledge Distillation (GKD) with loss decay.
result Achieved 27.68% top-1 error on ImageNet with ResNet18, outperforming state-of-the-art methods.
Optimal self-distillation improves generative models' velocity risk and mode recovery.
problem Improving generative models' velocity risk and mode recovery.
method Proved optimal self-distillation for rectified flow via linear probing, derived mixing coefficient, and provided validation tuning.
result Optimal self-distillation improves velocity risk and mode recovery.
Knowledge distillation improves model accuracy by mimicking teacher model probabilities.
problem Improving model accuracy through model compression.
method Casting knowledge distillation as a semiparametric inference problem, deriving new guarantees, and developing enhancements.
result Enhancements improve student performance by mitigating teacher overfitting and underfitting.
Self-distillation improves model performance by increasing teacher diversity and smoothing predictions.
problem Improving model generalization and performance through self-distillation.
method Interpreting self-distillation as MAP estimation and proposing instance-specific label smoothing.
result Self-distillation enhances model performance by increasing teacher diversity and smoothing predictions.
Survey of knowledge distillation for model compression.
problem Deploying large deep models on resource-limited devices.
method Knowledge distillation from large teacher models to small student models.
result Effective model compression and acceleration.
RKD improves model compression by distilling residual knowledge from a deep teacher model.
problem Performance degradation due to the gap between student and teacher models.
method Introduces an assistant model to distill residual knowledge from the teacher model.
result RKD achieves better results on popular classification datasets than state-of-the-art methods.
The soaring demand for intelligent mobile applications calls for deploying powerful deep neural networks (DNNs) on mobile devices. However, the outstanding performance of DNNs notoriously relies on increasingly complex models, which in turn is associated with an increase in computational expense far surpassing mobile d…
PLD distills knowledge using choice-theoretic Plackett-Luce model.
problem Model compression and knowledge transfer between large and small networks.
method PLD uses a weighted list-wise ranking loss based on the Plackett-Luce model.
result PLD achieves consistent gains across diverse architectures and distillation methods.
We present a novel framework of knowledge distillation that is capable of learning powerful and efficient student models from ensemble teacher networks. Our approach addresses the inherent model capacity issue between teacher and student and aims to maximize benefit from teacher models during distillation by reducing t…
We study the use of knowledge distillation to compress the U-net architecture. We show that, while standard distillation is not sufficient to reliably train a compressed U-net, introducing other regularization methods, such as batch normalization and class re-weighting, in knowledge distillation significantly improves …
Knowledge distillation boosts simple models in banking without complexity.
problem Complexity and performance constraints in banking models.
method Soft Targets, Sample Selection, Data Augmentation.
result Improved model performance without altering model simplicity.
This paper improves speech recognition by distilling knowledge from acoustic models.
problem Improving speech recognition accuracy using ensemble models.
method Proposes multi-teacher distillation strategies for joint CTC-attention end-to-end ASR systems, integrating error rate metric for optimization.
result Reports state-of-the-art error rates on various datasets and languages.
The paper analyzes knowledge distillation in wide neural networks, providing theoretical insights and practical implications.
problem Lack of theoretical understanding of knowledge distillation in wide neural networks.
method Theoretical analysis of knowledge distillation in a linearized model of a wide neural network, introducing a metric of task training difficulty.
result For a perfect teacher, a high ratio of teacher's soft labels can be beneficial. For imperfect teacher, hard labels can correct wrong predictions.
Enhances model compression with multi-teacher knowledge distillation.
problem Uncertainty evaluation and diverse teacher expertise in model deployment.
method Bayesian inference and teacher-informed prior with entropy-based weighting.
result Improved predictive accuracy and robust uncertainty quantification.
Distilled models often fail to match teacher models, despite improving generalization.
problem The discrepancy between teacher and student predictive distributions remains large.
method Investigated the optimization difficulties and dataset details affecting student performance.
result Optimizing for matching the teacher does not always lead to better generalization.