Searcharxiv⌕ Search

arXiv subjects

En-Hui Yang

Publications and source records attributed to En-Hui Yang.

16 recordsLinked to original sources
arXiv2026

Learned JPEG Compression for DNN Vision

JPEG, a lossy image compression technique designed for human viewers, has maintained its dominance for decades. However, in the era of artificial intelligence (AI), a substantial portion of image data, often compressed by JPEG, is and will continue to be consumed by deep neural networks (DNNs) instead of humans, thus creating a need to optimize JPEG for DNN inference performance. To this end, we propose learned JPEG compression for DNN vision (J4D), a novel training framework for determining JPEG encoding parameters to minimize compression rate while maximizing DNN inference performance. The major challenge of solving this optimization problem lies in representing the JPEG codec and compression rate in closed form. By incorporating a differentiable soft quantizer based on a probabilistic quantization scheme, we not only obtain a differentiable proxy for the JPEG codec, but are also able to compute the entropy of the coded source analytically, which is a close estimate of the actual compression rate. Equipped with both the differentiable JPEG codec and the information-theoretic rate estimator, we are then able to solve the aforementioned optimization problem with backpropagation. After training, the learned encoding parameters will be subsequently used in actual JPEG encoding based on probabilistic quantization. Extensive experimental results across multiple datasets and DNN architectures demonstrate that J4D consistently and significantly outperforms the default JPEG and other competitive JPEG codecs optimized for DNNs. Notably, compared to the default JPEG, J4D achieves an increase in accuracy by as much as 11.60% at the same rate, or a reduction of compression rate up to 80.05% at the same accuracy. Additionally, with the help of J4D, we show the potential to design universal JPEG encoding parameters for various DNN architectures for the first time.

cs.CV↗
arXiv2025

Coupled Data and Measurement Space Dynamics for Enhanced Diffusion Posterior Sampling

Inverse problems, where the goal is to recover an unknown signal from noisy or incomplete measurements, are central to applications in medical imaging, remote sensing, and computational biology. Diffusion models have recently emerged as powerful priors for solving such problems. However, existing methods either rely on projection-based techniques that enforce measurement consistency through heuristic updates, or they approximate the likelihood $p(\boldsymbol{y} \mid \boldsymbol{x})$, often resulting in artifacts and instability under complex or high-noise conditions. To address these limitations, we propose a novel framework called \emph{coupled data and measurement space diffusion posterior sampling} (C-DPS), which eliminates the need for constraint tuning or likelihood approximation. C-DPS introduces a forward stochastic process in the measurement space $\{\boldsymbol{y}_t\}$, evolving in parallel with the data-space diffusion $\{\boldsymbol{x}_t\}$, which enables the derivation of a closed-form posterior $p(\boldsymbol{x}_{t-1} \mid \boldsymbol{x}_t, \boldsymbol{y}_{t-1})$. This coupling allows for accurate and recursive sampling based on a well-defined posterior distribution. Empirical results demonstrate that C-DPS consistently outperforms existing baselines, both qualitatively and quantitatively, across multiple inverse problem benchmarks.

cs.LG↗
arXiv2025

Leveraging Conditional Mutual Information to Improve Large Language Model Fine-Tuning For Classification

Although large language models (LLMs) have demonstrated remarkable capabilities in recent years, the potential of information theory (IT) to enhance LLM development remains underexplored. This paper introduces the information theoretic principle of Conditional Mutual Information (CMI) to LLM fine-tuning for classification tasks, exploring its promise in two main ways: minimizing CMI to improve a model's standalone performance and maximizing CMI to enhance knowledge distillation (KD) for more capable student models. To apply CMI in LLM fine-tuning, we adapt the recently proposed CMI-constrained deep learning framework, which was initially developed for image classification, with some modification. By minimizing CMI during LLM fine-tuning, we achieve superior performance gains on 6 of 8 GLUE classification tasks compared to BERT. Additionally, maximizing CMI during the KD process results in significant performance improvements in 6 of 8 GLUE classification tasks compared to DistilBERT. These findings demonstrate CMI's adaptability for optimizing both standalone LLMs and student models, showcasing its potential as a robust framework for advancing LLM fine-tuning. Our work bridges the gap between information theory and LLM development, offering new insights for building high-performing language models.

cs.CL↗
arXiv2025

Conditional Mutual Information Based Diffusion Posterior Sampling for Solving Inverse Problems

Inverse problems are prevalent across various disciplines in science and engineering. In the field of computer vision, tasks such as inpainting, deblurring, and super-resolution are commonly formulated as inverse problems. Recently, diffusion models (DMs) have emerged as a promising approach for addressing noisy linear inverse problems, offering effective solutions without requiring additional task-specific training. Specifically, with the prior provided by DMs, one can sample from the posterior by finding the likelihood. Since the likelihood is intractable, it is often approximated in the literature. However, this approximation compromises the quality of the generated images. To overcome this limitation and improve the effectiveness of DMs in solving inverse problems, we propose an information-theoretic approach. Specifically, we maximize the conditional mutual information $\mathrm{I}(\boldsymbol{x}_0; \boldsymbol{y} | \boldsymbol{x}_t)$, where $\boldsymbol{x}_0$ represents the reconstructed signal, $\boldsymbol{y}$ is the measurement, and $\boldsymbol{x}_t$ is the intermediate signal at stage $t$. This ensures that the intermediate signals $\boldsymbol{x}_t$ are generated in a way that the final reconstructed signal $\boldsymbol{x}_0$ retains as much information as possible about the measurement $\boldsymbol{y}$. We demonstrate that this method can be seamlessly integrated with recent approaches and, once incorporated, enhances their performance both qualitatively and quantitatively.

cs.LG↗
arXiv2024

Enhancing Diffusion Models for Inverse Problems with Covariance-Aware Posterior Sampling

Inverse problems exist in many disciplines of science and engineering. In computer vision, for example, tasks such as inpainting, deblurring, and super resolution can be effectively modeled as inverse problems. Recently, denoising diffusion probabilistic models (DDPMs) are shown to provide a promising solution to noisy linear inverse problems without the need for additional task specific training. Specifically, with the prior provided by DDPMs, one can sample from the posterior by approximating the likelihood. In the literature, approximations of the likelihood are often based on the mean of conditional densities of the reverse process, which can be obtained using Tweedie formula. To obtain a better approximation to the likelihood, in this paper we first derive a closed form formula for the covariance of the reverse process. Then, we propose a method based on finite difference method to approximate this covariance such that it can be readily obtained from the existing pretrained DDPMs, thereby not increasing the complexity compared to existing approaches. Finally, based on the mean and approximated covariance of the reverse process, we present a new approximation to the likelihood. We refer to this method as covariance-aware diffusion posterior sampling (CA-DPS). Experimental results show that CA-DPS significantly improves reconstruction performance without requiring hyperparameter tuning. The code for the paper is put in the supplementary materials.

cs.CV↗
arXiv2024

Going Beyond Feature Similarity: Effective Dataset Distillation based on Class-Aware Conditional Mutual Information

Dataset distillation (DD) aims to minimize the time and memory consumption needed for training deep neural networks on large datasets, by creating a smaller synthetic dataset that has similar performance to that of the full real dataset. However, current dataset distillation methods often result in synthetic datasets that are excessively difficult for networks to learn from, due to the compression of a substantial amount of information from the original data through metrics measuring feature similarity, e,g., distribution matching (DM). In this work, we introduce conditional mutual information (CMI) to assess the class-aware complexity of a dataset and propose a novel method by minimizing CMI. Specifically, we minimize the distillation loss while constraining the class-aware complexity of the synthetic dataset by minimizing its empirical CMI from the feature space of pre-trained networks, simultaneously. Conducting on a thorough set of experiments, we show that our method can serve as a general regularization method to existing DD methods and improve the performance and training efficiency.

cs.CV↗
arXiv2024

JPEG Inspired Deep Learning

Although it is traditionally believed that lossy image compression, such as JPEG compression, has a negative impact on the performance of deep neural networks (DNNs), it is shown by recent works that well-crafted JPEG compression can actually improve the performance of deep learning (DL). Inspired by this, we propose JPEG-DL, a novel DL framework that prepends any underlying DNN architecture with a trainable JPEG compression layer. To make the quantization operation in JPEG compression trainable, a new differentiable soft quantizer is employed at the JPEG layer, and then the quantization operation and underlying DNN are jointly trained. Extensive experiments show that in comparison with the standard DL, JPEG-DL delivers significant accuracy improvements across various datasets and model architectures while enhancing robustness against adversarial attacks. Particularly, on some fine-grained image classification datasets, JPEG-DL can increase prediction accuracy by as much as 20.9%. Our code is available on https://github.com/AhmedHussKhalifa/JPEG-Inspired-DL.git.

cs.CV↗
arXiv2024

Knowledge Distillation Based on Transformed Teacher Matching

As a technique to bridge logit matching and probability distribution matching, temperature scaling plays a pivotal role in knowledge distillation (KD). Conventionally, temperature scaling is applied to both teacher's logits and student's logits in KD. Motivated by some recent works, in this paper, we drop instead temperature scaling on the student side, and systematically study the resulting variant of KD, dubbed transformed teacher matching (TTM). By reinterpreting temperature scaling as a power transform of probability distribution, we show that in comparison with the original KD, TTM has an inherent R\'enyi entropy term in its objective function, which serves as an extra regularization term. Extensive experiment results demonstrate that thanks to this inherent regularization, TTM leads to trained students with better generalization than the original KD. To further enhance student's capability to match teacher's power transformed probability distribution, we introduce a sample-adaptive weighting coefficient into TTM, yielding a novel distillation approach dubbed weighted TTM (WTTM). It is shown, by comprehensive experiments, that although WTTM is simple, it is effective, improves upon TTM, and achieves state-of-the-art accuracy performance. Our source code is available at https://github.com/zkxufo/TTM.

cs.LG↗
arXiv2024

Bayes Conditional Distribution Estimation for Knowledge Distillation Based on Conditional Mutual Information

It is believed that in knowledge distillation (KD), the role of the teacher is to provide an estimate for the unknown Bayes conditional probability distribution (BCPD) to be used in the student training process. Conventionally, this estimate is obtained by training the teacher using maximum log-likelihood (MLL) method. To improve this estimate for KD, in this paper we introduce the concept of conditional mutual information (CMI) into the estimation of BCPD and propose a novel estimator called the maximum CMI (MCMI) method. Specifically, in MCMI estimation, both the log-likelihood and CMI of the teacher are simultaneously maximized when the teacher is trained. Through Eigen-CAM, it is further shown that maximizing the teacher's CMI value allows the teacher to capture more contextual information in an image cluster. Via conducting a thorough set of experiments, we show that by employing a teacher trained via MCMI estimation rather than one trained via MLL estimation in various state-of-the-art KD frameworks, the student's classification accuracy consistently increases, with the gain of up to 3.32\%. This suggests that the teacher's BCPD estimate provided by MCMI method is more accurate than that provided by MLL method. In addition, we show that such improvements in the student's accuracy are more drastic in zero-shot and few-shot settings. Notably, the student's accuracy increases with the gain of up to 5.72\% when 5\% of the training samples are available to the student (few-shot), and increases from 0\% to as high as 84\% for an omitted class (zero-shot). The code is available at \url{https://github.com/iclr2024mcmi/ICLRMCMI}.

cs.LG↗
arXiv2024

AdaFed: Fair Federated Learning via Adaptive Common Descent Direction

Federated learning (FL) is a promising technology via which some edge devices/clients collaboratively train a machine learning model orchestrated by a server. Learning an unfair model is known as a critical problem in federated learning, where the trained model may unfairly advantage or disadvantage some of the devices. To tackle this problem, in this work, we propose AdaFed. The goal of AdaFed is to find an updating direction for the server along which (i) all the clients' loss functions are decreasing; and (ii) more importantly, the loss functions for the clients with larger values decrease with a higher rate. AdaFed adaptively tunes this common direction based on the values of local gradients and loss functions. We validate the effectiveness of AdaFed on a suite of federated datasets, and demonstrate that AdaFed outperforms state-of-the-art fair FL methods.

cs.LG↗
arXiv2023

Conditional Mutual Information Constrained Deep Learning for Classification

The concepts of conditional mutual information (CMI) and normalized conditional mutual information (NCMI) are introduced to measure the concentration and separation performance of a classification deep neural network (DNN) in the output probability distribution space of the DNN, where CMI and the ratio between CMI and NCMI represent the intra-class concentration and inter-class separation of the DNN, respectively. By using NCMI to evaluate popular DNNs pretrained over ImageNet in the literature, it is shown that their validation accuracies over ImageNet validation data set are more or less inversely proportional to their NCMI values. Based on this observation, the standard deep learning (DL) framework is further modified to minimize the standard cross entropy function subject to an NCMI constraint, yielding CMI constrained deep learning (CMIC-DL). A novel alternating learning algorithm is proposed to solve such a constrained optimization problem. Extensive experiment results show that DNNs trained within CMIC-DL outperform the state-of-the-art models trained within the standard DL and other loss functions in the literature in terms of both accuracy and robustness against adversarial attacks. In addition, visualizing the evolution of learning process through the lens of CMI and NCMI is also advocated.

cs.LG↗
arXiv2012

Jar Decoding: Non-Asymptotic Converse Coding Theorems, Taylor-Type Expansion, and Optimality

Recently, a new decoding rule called jar decoding was proposed; under jar decoding, a non-asymptotic achievable tradeoff between the coding rate and word error probability was also established for any discrete input memoryless channel with discrete or continuous output (DIMC). Along the path of non-asymptotic analysis, in this paper, it is further shown that jar decoding is actually optimal up to the second order coding performance by establishing new non-asymptotic converse coding theorems, and determining the Taylor expansion of the (best) coding rate $R_n (ε)$ of finite block length for any block length $n$ and word error probability $ε$ up to the second order. Finally, based on the Taylor-type expansion and the new converses, two approximation formulas for $R_n (ε)$ (dubbed "SO" and "NEP") are provided; they are further evaluated and compared against some of the best bounds known so far, as well as the normal approximation of $R_n (ε)$ revisited recently in the literature. It turns out that while the normal approximation is all over the map, i.e. sometime below achievable bounds and sometime above converse bounds, the SO approximation is much more reliable as it is always below converses; in the meantime, the NEP approximation is the best among the three and always provides an accurate estimation for $R_n (ε)$. An important implication arising from the Taylor-type expansion of $R_n (ε)$ is that in the practical non-asymptotic regime, the optimal marginal codeword symbol distribution is not necessarily a capacity achieving distribution.

cs.IT↗
arXiv2012

Non-asymptotic Equipartition Properties for Independent and Identically Distributed Sources

Given an independent and identically distributed source $X = \{X_i \}_{i=1}^{\infty}$ with finite Shannon entropy or differential entropy (as the case may be) $H(X)$, the non-asymptotic equipartition property (NEP) with respect to $H(X)$ is established, which characterizes, for any finite block length $n$, how close $-{1\over n} \ln p(X_1 X_2...X_n)$ is to $H(X)$ by determining the information spectrum of $X_1 X_2...X_n $, i.e., the distribution of $-{1\over n} \ln p(X_1 X_2...X_n)$. Non-asymptotic equipartition properties (with respect to conditional entropy, mutual information, and relative entropy) in a similar nature are also established. These non-asymptotic equipartition properties are instrumental to the development of non-asymptotic coding (including both source and channel coding) results in information theory in the same way as the asymptotic equipartition property to all asymptotic coding theorems established so far in information theory. As an example, the NEP with respect to $H(X)$ is used to establish a non-asymptotic fixed rate source coding theorem, which reveals, for any finite block length $n$, a complete picture about the tradeoff between the minimum rate of fixed rate coding of $X_1...X_n$ and error probability when the error probability is a constant, or goes to 0 with block length $n$ at a sub-polynomial, polynomial or sub-exponential speed. With the help of the NEP with respect to other information quantities, non-asymptotic channel coding theorems of similar nature will be established in a separate paper.

cs.IT↗
arXiv2012

Interactive Encoding and Decoding Based on Binary LDPC Codes with Syndrome Accumulation

Interactive encoding and decoding based on binary low-density parity-check codes with syndrome accumulation (SA-LDPC-IED) is proposed and investigated. Assume that the source alphabet is $\mathbf{GF}(2)$, and the side information alphabet is finite. It is first demonstrated how to convert any classical universal lossless code $\mathcal{C}_n$ (with block length $n$ and side information available to both the encoder and decoder) into a universal SA-LDPC-IED scheme. It is then shown that with the word error probability approaching 0 sub-exponentially with $n$, the compression rate (including both the forward and backward rates) of the resulting SA-LDPC-IED scheme is upper bounded by a functional of that of $\mathcal{C}_n$, which in turn approaches the compression rate of $\mathcal{C}_n$ for each and every individual sequence pair $(x^n,y^n)$ and the conditional entropy rate $\mathrm{H}(X |Y)$ for any stationary, ergodic source and side information $(X, Y)$ as the average variable node degree $\bar{l}$ of the underlying LDPC code increases without bound. When applied to the class of binary source and side information $(X, Y)$ correlated through a binary symmetrical channel with cross-over probability unknown to both the encoder and decoder, the resulting SA-LDPC-IED scheme can be further simplified, yielding even improved rate performance versus the bit error probability when $\bar{l}$ is not large. Simulation results (coupled with linear time belief propagation decoding) on binary source-side information pairs confirm the theoretic analysis, and further show that the SA-LDPC-IED scheme consistently outperforms the Slepian-Wolf coding scheme based on the same underlying LDPC code. As a by-product, probability bounds involving LDPC established in the course are also interesting on their own and expected to have implications on the performance of LDPC for channel coding as well.

cs.IT↗
arXiv2003

Universal Compression of Ergodic Quantum Sources

For a real number $r>0$, let $F(r)$ be the family of all stationary ergodic quantum sources with von Neumann entropy rates less than $r$. We prove that, for any $r>0$, there exists a blind, source-independent block compression scheme which compresses every source from $F(r)$ to $r n$ qubits per input block length $n$ with arbitrarily high fidelity for all large $n$. As our second result,we show that the stationarity and the ergodicity of a quantum source $\{ρ_m \}_{m=1}^{\infty}$ are preserved by any trace-preserving completely positive linear map of the tensor product form ${\cal E}^{\otimes m}$, where a copy of ${\cal E}$ acts locally on each spin lattice site. We also establish ergodicity criteria for so called classically-correlated quantum sources.

quant-ph↗