A framework preserves uncertainty in ensemble distillation.
problem Preserving uncertainty decomposition in ensemble distillation.
method General framework for distilling both regression and classification ensembles, preserving natural uncertainty decomposition.
result Framework maintains decomposition of predictive uncertainty.
Ensembles of models often yield improvements in system performance. These ensemble approaches have also been empirically shown to yield robust measures of uncertainty, and are capable of distinguishing between different \emph{forms} of uncertainty. However, ensembles come at a computational and memory cost which may be…
S2D efficiently trains models to estimate uncertainty without increasing resource costs.
problem Efficiently estimating uncertainty in deep learning models for safety-critical applications.
method Self-distribution distillation (S2D) approach to train a single model for uncertainty estimation.
result S2D models outperform standard models and Monte-Carlo dropout in uncertainty estimation.
New method improves ensemble inference for high-class tasks.
problem High inference costs for ensemble models.
method Proxy-Dirichlet target to minimize reverse KL-divergence.
result Resolves gradient issues for large-scale classification tasks.
Regression Prior Networks improve ensemble performance on regression tasks.
problem Improving ensemble performance on regression tasks.
method Extending Prior Networks and Ensemble Distribution Distillation (EnD2) to regression tasks using the Normal-Wishart distribution. result Regression Prior Networks yield performance competitive with ensemble approaches on regression tasks.
Ensembles of models have been empirically shown to improve predictive performance and to yield robust measures of uncertainty. However, they are expensive in computation and memory. Therefore, recent research has focused on distilling ensembles into a single compact model, reducing the computational and memory burden o…
FED distills ensemble predictions efficiently, improving accuracy and uncertainty.
problem Efficiently distilling ensemble predictions from limited data.
method Proposes Functional Ensemble Distillation (FED) using mixup augmentation.
result Significantly boosts performance in accuracy and uncertainty estimation.
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.
Paper proposes ensemble distillation for well-calibrated structured prediction.
problem Well-calibrated predictions are hard to achieve in structured prediction.
method Ensemble distillation framework for structured prediction.
result Ensemble distillation produces well-calibrated models with similar performance and calibration benefits to ensembles.
FAST-DAD distills complex ensemble models into faster, more accurate individual models.
problem Deploying complex AutoML ensemble predictors on tabular data is slow, large, and opaque.
method Data augmentation strategy based on Gibbs sampling from a self-attention pseudolikelihood estimator.
result FAST-DAD distillation produces significantly better individual models than standard training.
Top-performing machine learning systems, such as deep neural networks, large ensembles and complex probabilistic graphical models, can be expensive to store, slow to evaluate and hard to integrate into larger systems. Ideally, we would like to replace such cumbersome models with simpler models that perform equally well…
This paper connects RND, deep ensembles, and Bayesian inference, providing a unified theoretical perspective.
problem Uncertainty quantification in deep learning models.
method Analysis of Random Network Distillation (RND) within the neural tangent kernel framework.
result The uncertainty signal from RND is equivalent to the predictive variance of a deep ensemble and can be made to mirror the centered posterior predictive distribution of Bayesian inference.
Paper distills ensemble ENSO forecasts into simpler models for better diagnostics.
problem Interpreting complex ensemble ENSO forecasts.
method Aggregates eSPA ensemble members that correctly predict ENSO phase.
result Compact distilled models maintain forecast performance and enable diagnostics.
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.
Anti-Distillation improves reproducibility of deep networks by making ensemble predictions more diverse.
problem Deep networks are prone to high prediction differences, making them less reproducible.
method Anti-Distillation uses ensembles to force predictions to be more different and diverse.
result Anti-Distillation reduces prediction differences by making ensemble predictions more diverse.
Single model estimates ensemble uncertainty efficiently.
problem Efficient uncertainty quantification in deep learning models.
method Contextual similarity distillation, approximating ensemble variance with a single model.
result Single model estimates predictive variance with a single forward pass.
Distillation is an effective knowledge-transfer technique that uses predicted distributions of a powerful teacher model as soft targets to train a less-parameterized student model. A pre-trained high capacity teacher, however, is not always available. Recently proposed online variants use the aggregated intermediate pr…
Recent advances in deep learning have facilitated the demand of neural models for real applications. In practice, these applications often need to be deployed with limited resources while keeping high accuracy. This paper touches the core of neural models in NLP, word embeddings, and presents a new embedding distillati…
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…
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.
Techniques such as ensembling and distillation promise model quality improvements when paired with almost any base model. However, due to increased test-time cost (for ensembles) and increased complexity of the training pipeline (for distillation), these techniques are challenging to use in industrial settings. In this…
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…
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.
Paper introduces efficient uncertainty estimation in LLMs without multiple forward passes.
problem Accurate uncertainty quantification in LLMs remains challenging.
method Evidential Knowledge Distillation to create compact student models.
result Efficient uncertainty estimation achieved with single forward pass.
Compact Gaussian model approximates deep ensemble predictions.
problem Efficiently approximating deep ensemble models for image prediction.
method Sparse-structured multivariate Gaussian with Cholesky parameterization trained to match pre-trained ensemble outputs.
result Compact representation captures uncertainty and structured correlations explicitly.
New methods improve tree ensemble models by compressing them while maintaining accuracy.
problem Theoretical understanding and practical compression of tree ensembles like random forests and gradient boosting machines.
method Spectral perspective on tree ensembles, deriving minimax rates and developing compression schemes.
result Leading eigenfunctions/singular vectors capture dominant predictive directions, leading to smaller, competitive models.
In this study we examined the question of how error correction occurs in an ensemble of deep convolutional networks, trained for an important applied problem: segmentation of Electrocardiograms(ECG). We also explore the possibility of using the information about ensemble errors to evaluate a quality of data representat…
Compressed Federated Distillation reduces communication in federated learning.
problem Communication constraints in Federated Learning.
method Compressed Federated Distillation (CFD) leverages soft labels and quantization techniques.
result Reduces communication by more than 4 orders of magnitude compared to Federated Averaging.
A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to…
Federated learning improves by training central model with client model outputs.
problem Direct averaging of client models is limited in FL.
method Ensemble distillation for model fusion.
result Central model trained faster with fewer communication rounds.
FedAUX improves Federated Learning by better using unlabeled data.
problem Improving Federated Learning performance with unlabeled data.
method FedAUX modifies FD training by unsupervised pre-training and private certainty scoring.
result FedAUX outperforms state-of-the-art Federated Learning methods.
In this work we aim to obtain computationally-efficient uncertainty estimates with deep networks. For this, we propose a modified knowledge distillation procedure that achieves state-of-the-art uncertainty estimates both for in and out-of-distribution samples. Our contributions include a) demonstrating and adapting to …
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 distills Bayesian posterior expectations for deep neural networks.
problem Improving deep neural network performance and uncertainty quantification.
method Develops a framework for distilling expectations from Bayesian posterior distributions using Monte Carlo samples.
result The framework successfully distills posterior predictive distribution and expected entropy.
EM Distillation simplifies diffusion models to one-step generators.
problem Efficient sampling from complex diffusion models with minimal loss of quality.
method EM Distillation, a maximum likelihood approach based on Expectation-Maximization.
result EM Distillation outperforms existing one-step generative methods in FID scores.
KD2M unifies feature matching in neural networks.
problem Matching neural network activations for knowledge transfer.
method Formalizes distribution matching strategy.
result New theoretical results for KD.
A new distillation framework predicts stock trading volumes more accurately with less model size.
problem Predicting stock trading volumes using regression models without class correlations.
method Transformed regression model into a probabilistic forecasting model, matching distributions and correlational relationships.
result Framework achieves superior prediction accuracy with significantly smaller model size.
FedZKT enables resource-constrained devices to participate in federated learning with heterogeneous models.
problem Inequality in resource allocation hinders participation from resource-constrained devices in federated learning.
method Zero-shot knowledge transfer through a server-assigned distillation process.
result FedZKT effectively transfers knowledge across heterogeneous on-device models without requiring comparable local training efforts.
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.
UVU simplifies value uncertainty quantification in RL.
problem Estimating epistemic uncertainty in value functions for reinforcement learning.
method UVU uses squared prediction errors between an online learner and a fixed, randomly initialized target network, incorporating policy-conditional value uncertainty.
result UVU achieves equal performance to large ensembles on challenging offline RL settings, with computational savings.
Dataset distillation is a method for reducing dataset sizes by learning a small number of synthetic samples containing all the information of a large dataset. This has several benefits like speeding up model training, reducing energy consumption, and reducing required storage space. Currently, each synthetic sample is …
Extracurricular learning closes the accuracy gap in knowledge distillation.
problem Accuracy gap between teacher and student models after knowledge distillation.
method Modeling student and teacher output distributions, sampling from an extended data distribution, and matching over this set.
result Extracurricular learning reduces the accuracy gap by 46% to 68%.
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.
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.
We introduce a novel approach to feed-forward neural network interpretation based on partitioning the space of sequences of neuron activations. In line with this approach, we propose a model-specific interpretation method, called YASENN. Our method inherits many advantages of model-agnostic distillation, such as an abi…
Ensembles of deep neural networks significantly improve generalization accuracy. However, training neural network ensembles requires a large amount of computational resources and time. State-of-the-art approaches either train all networks from scratch leading to prohibitive training cost that allows only very small ens…
This paper investigates teacher hacking during language model distillation and proposes methods to mitigate it.
problem Teacher hacking during language model distillation, leading to suboptimal performance.
method A controlled experimental setup involving an oracle LM, teacher LM, and student LM, using fixed offline or online data generation techniques.
result Data diversity is the key factor in preventing teacher hacking during distillation.
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.