Searcharxiv⌕ Search

arXiv subjects

Chengjun Jin

Publications and source records attributed to Chengjun Jin.

7 recordsLinked to original sources

Retain-Neutral Surrogates for Min-Max Unlearning

Machine unlearning aims to make a model forget designated training data while retaining performance on the rest. In min-max unlearning, an inner adversary perturbs the model to maximize the forget loss, and the model is updated to minimize the retain loss at the resulting surrogate point. When the forget and retain gradients are positively coupled, the standard perturbation also moves toward higher retain loss, and the retain step begins by recovering that loss. We propose Retain-Orthogonal Surrogate Unlearning (ROSU), which redefines the adversary to admit only perturbations orthogonal to the retain gradient. The one that raises the forget loss fastest has a closed form, and ROSU uses it to set the surrogate point and amplifies it into the forgetting step. Under positive coupling the ROSU surrogate avoids the retain cost that coupling induces in the standard surrogate, and the difference shrinks as the gradients approach orthogonality. ROSU has the smallest accuracy gap to retraining in every vision setting (CIFAR-10, CIFAR-100, Tiny-ImageNet) and is competitive with the baselines on language benchmarks (TOFU, WMDP). The redefinition helps most under high positive coupling. Code is available at https://github.com/unlearn-ml/machine-unlearning.

cs.LG↗

Data-Efficient Unlearning-Based Backdoor Detection in Federated Learning

Detecting backdoor attacks in federated learning (FL) is challenging because malicious clients can manipulate the global model while preserving main-task performance. Static model-update statistics may therefore fail to clearly distinguish malicious from benign clients. Unlearning-based detectors instead probe client models through loss ascent on clean server data, but detection performance degrades with scarce proxy data. We propose Screening via Unlearning Response Analysis (SURA), a data-efficient detector. SURA reuses the same small clean proxy across multiple train-mode loss-ascent steps, strengthening client responses without additional clean data. For federated vision models with batch normalization (BN), repeated probing also changes BN running statistics and can confound endpoint measurements. SURA addresses this by restoring the shared pre-probe BN state only at readout while preserving the parameter changes induced by probing. This separates repeated probing from response measurement and removes BN drift and the interaction between parameter and BN-state changes from the readout. Across six backdoor attacks, SURA uses only 10 clean server images, outperforms methods under the same data budget, and approaches the robustness obtained with much larger proxy sets.

cs.CR↗

SCOPE: Entanglement Frontier Escape for Source-Free Class Unlearning

Source-free class unlearning erases whole classes using only the forget data, judged at the representation level, where features can leak a class the head no longer predicts. Existing feature-space erasers answer with one fixed projection, yet forget and retain classes share a representation, so deleting one disturbs the other where they overlap. We prove this tension is a frontier. Every fixed projection that deletes pays a retain cost of at least the retain-readout energy along the forget-discriminant subspace, and erasing that subspace alone attains the floor. The leading source-free erasers all instantiate the form it binds, so the frontier limits the whole class. Conditioning the erasure on the input escapes it. Spectral Conditional Projective Erasure (SCOPE) does so with a single gate, suppressing the forget subspace chiefly on inputs its frozen head's weight scores read as a forget class. It is closed form, needs no retain data or gradient training, and costs orders of magnitude less than retraining. Across five object, face, and speaker benchmarks spanning two modalities and both convolutional and transformer backbones, the frontier predicts the measured retain cost. SCOPE leads the source-free erasers on every benchmark and forget-set size, and at the hardest setting it tops every unlearner, trained methods included.

cs.LG↗

HCF: Hierarchical Cascade Framework for Distributed Multi-Stage Image Compression

Distributed multi-stage image compression -- where visual content traverses multiple processing nodes under varying quality requirements -- poses challenges. Progressive methods enable bitstream truncation but underutilize available compute resources; successive compression repeats costly pixel-domain operations and suffers cumulative quality loss and inefficiency; fixed-parameter models lack post-encoding flexibility. In this work, we developed the Hierarchical Cascade Framework (HCF) that achieves high rate-distortion performance and better computational efficiency through direct latent-space transformations across network nodes in distributed multi-stage image compression systems. Under HCF, we introduced policy-driven quantization control to optimize rate-distortion trade-offs, and established the edge quantization principle through differential entropy analysis. The configuration based on this principle demonstrates up to 0.6dB PSNR gains over other configurations. When comprehensively evaluated on the Kodak, CLIC, and CLIC2020-mobile datasets, HCF outperforms successive-compression methods by up to 5.56% BD-Rate in PSNR on CLIC, while saving up to 97.8% FLOPs, 96.5% GPU memory, and 90.0% execution time. It also outperforms state-of-the-art progressive compression methods by up to 12.64% BD-Rate on Kodak and enables retraining-free cross-quality adaptation with 7.13-10.87% BD-Rate reductions on CLIC2020-mobile.

cs.CV↗

Molecular Realization of a Quantum NAND Tree

The negative-AND (NAND) gate is universal for classical computation making it an important target for development. A seminal quantum computing algorithm by Farhi, Goldstone and Gutmann has demonstrated its realization by means of quantum scattering yielding a quantum algorithm that evaluates the output faster than any classical algorithm. Here, we derive the NAND outputs analytically from scattering theory using a tight-binding (TB) model and show the restrictions on the TB parameters in order to still maintain the NAND gate function. We map the quantum NAND tree onto a conjugated molecular system, and compare the NAND output with non-equilibrium Green's function (NEGF) transport calculations using density functional theory (DFT) and TB Hamiltonians for the electronic structure. Further, we extend our molecular platform to show other classical gates that can be realized for quantum computing by scattering on graphs.

quant-ph↗

Dynamical Image Charge Effect in Molecular Tunnel Junctions: Beyond Energy Level Alignment

When an electron tunnels between two metal contacts it temporarily induces an image charge (IC) in the electrodes which acts back on the tunneling electron. It is usually assumed that the IC forms instantaneously such that a static model for the image potential applies. Here we investigate how the finite IC formation time affects charge transport through a molecule suspended between two electrodes. For a single level model, an analytical treatment shows that the conductance is suppressed by a factor $Z^2$ (compared to the static IC approximation) where $Z$ is the quasiparticle renormalization factor. We show that $Z$ can be expressed either in terms of the plasma frequency of the electrode or as the overlap between the ground states of the electrode with and without an electron on the molecule. First-principles GW calculations for benzene-diamine connected to gold electrodes show that the dynamical corrections can reduce the conductance by more than a factor of two.

cond-mat.mes-hall↗

Quantitatively Accurate Calculations of Conductance and Thermopower of Molecular Junctions

Thermopower measurements of molecular junctions have recently gained interest as a characterization technique that supplements the more traditional conductance measurements. Here we investigate the electronic conductance and thermopower of benzenediamine (BDA) and benzenedicarbonitrile (BDCN) connected to gold electrodes using first-principles calculations. We find excellent agreement with experiments for both molecules when exchange-correlation effects are described by the many-body GW approximation. In contrast, results from standard density functional theory (DFT) deviate from experiments by up to two orders of magnitude. The failure of DFT is particularly pronounced for the n-type BDCN junction due to the severe underestimation of the lowest unoccupied molecular orbital (LUMO). The quality of the DFT results can be improved by correcting the molecular energy levels for self-interaction errors and image charge effects. Finally, we show that the conductance and thermopower of the considered junctions are relatively insensitive to the metal-molecule bonding geometry. Our results demonstrate that electronic and thermoelectric properties of molecular junctions can be predicted from first-principles calculations when exchange-correlation effects are taken properly into account.

cond-mat.mes-hall↗