Federated learning(FL) is used for distributed scenarios to train artificial intelligence(AI) models while ensuring users' privacy. In federated learning scenario, the server generally never knows about users' data. This type of concept makes the AI training process efficient in terms of data privacy. However, regarding model performance, federated AI models may not sufficiently satisfy AI users' expectations. Furthermore, AI users have a wide range of different needs. It is not easy to satisfy the whole users needs. These types of issues can be addressed through AI model optimization, fine-tuning, or personalization to achieve optimal model performance. To address model optimization challenges, we propose reference model-based federated learning for optimal fine-tuning, which overcomes catastrophic forgetting in each round. This method is derived from Bayesian parameter-efficient transfer learning, which includes an optimal proximal term and enables overcoming the catastrophic forgetting issue in each round by utilizing a reference model that incorporates previous model parameters. As a result, this method achieves both high model performance and low computing cost.
Recently, vision-language pretraining has emerged as a transformative technique that integrates the strengths of both visual and textual modalities, resulting in powerful vision-language models (VLMs). Leveraging web-scale pretraining data, these models exhibit strong zero-shot capabilities. However, their performance often deteriorates when confronted with domain-specific or specialized generalization tasks. To address this, a growing body of research focuses on transferring or generalizing the rich knowledge embedded in VLMs to various downstream applications. This survey aims to comprehensively summarize the generalization settings, methodologies, benchmarking and results in VLM literatures. Delving into the typical VLM structures, current literatures are categorized into prompt-based, parameter-based and feature-based methods according to the transferred modules. The differences and characteristics in each category are furthered summarized and discussed by revisiting the typical transfer learning (TL) settings, providing novel interpretations for TL in the era of VLMs. Popular benchmarks for VLM generalization are further introduced with thorough performance comparisons among the reviewed methods. Following the advances in large-scale generalizable pretraining, this survey also discusses the relations and differences between VLMs and up-to-date multimodal large language models (MLLM), e.g., DeepSeek-VL. By systematically reviewing the surging literatures in vision-language research from a novel and practical generalization prospective, this survey contributes to a clear landscape of current and future multimodal researches.
This work introduces a growable and modular neural network architecture that naturally avoids catastrophic forgetting and interference in continual reinforcement learning. The structure of each module allows the selective combination of previous policies along with its internal policy, accelerating the learning process on the current task. Unlike previous growing neural network approaches, we show that the number of parameters of the proposed approach grows linearly with respect to the number of tasks, and does not sacrifice plasticity to scale. Experiments conducted in benchmark continuous control and visual problems reveal that the proposed approach achieves greater knowledge transfer and performance than alternative methods.
Explicit density learners are becoming an increasingly popular technique for generative models because of their ability to better model probability distributions. They have advantages over Generative Adversarial Networks due to their ability to perform density estimation and having exact latent-variable inference. This has many advantages, including: being able to simply interpolate, calculate sample likelihood, and analyze the probability distribution. The downside of these models is that they are often more difficult to train and have lower sampling quality. Normalizing flows are explicit density models, that use composable bijective functions to turn an intractable probability function into a tractable one. In this work, we present novel knowledge distillation techniques to increase sampling quality and density estimation of smaller student normalizing flows. We seek to study the capacity of knowledge distillation in Compositional Normalizing Flows to understand the benefits and weaknesses provided by these architectures. Normalizing flows have unique properties that allow for a non-traditional forms of knowledge transfer, where we can transfer that knowledge within intermediate layers. We find that through this distillation, we can make students significantly smaller while making substantial performance gains over a non-distilled student. With smaller models there is a proportionally increased throughput as this is dependent upon the number of bijectors, and thus parameters, in the network.
Atypical mitoses mark a deviation in the cell division process that can be an independent prognostically relevant marker for tumor malignancy. However, their identification remains challenging due to low prevalence, at times subtle morphological differences from normal mitoses, low inter-rater agreement among pathologists, and class imbalance in datasets. Building on the Atypical Mitosis dataset for Breast Cancer (AMi-Br), this study presents a comprehensive benchmark comparing deep learning approaches for automated atypical mitotic figure (AMF) classification, including baseline models, foundation models with linear probing, and foundation models fine-tuned with low-rank adaptation (LoRA). For rigorous evaluation, we further introduce two new hold-out AMF datasets - AtNorM-Br, a dataset of mitoses from the The TCGA breast cancer cohort, and AtNorM-MD, a multi-domain dataset of mitoses from the MIDOG++ training set. We found average balanced accuracy values of up to 0.8135, 0.7696, and 0.7705 on the in-domain AMi-Br and the out-of-domain AtNorm-Br and AtNorM-MD datasets, respectively, with the results being particularly good for LoRA-based adaptation of the Virchow-line of foundation models. Our work shows that atypical mitosis classification, while being a challenging problem, can be effectively addressed through the use of recent advances in transfer learning and model fine-tuning techniques. We make available all code and data used in this paper in this github repository: https://github.com/DeepMicroscopy/AMi-Br_Benchmark.
Contrastive learning has proven to be highly efficient and adaptable in shaping representation spaces across diverse modalities by pulling similar samples together and pushing dissimilar ones apart. However, two key limitations persist: (1) Without explicit regulation of the embedding distribution, semantically related instances can inadvertently be pushed apart unless complementary signals guide pair selection, and (2) excessive reliance on large in-batch negatives and tailored augmentations hinders generalization. To address these limitations, we propose Variational Supervised Contrastive Learning (VarCon), which reformulates supervised contrastive learning as variational inference over latent class variables and maximizes a posterior-weighted evidence lower bound (ELBO) that replaces exhaustive pair-wise comparisons for efficient class-aware matching and grants fine-grained control over intra-class dispersion in the embedding space. Trained exclusively on image data, our experiments on CIFAR-10, CIFAR-100, ImageNet-100, and ImageNet-1K show that VarCon (1) achieves state-of-the-art performance for contrastive learning frameworks, reaching 79.36% Top-1 accuracy on ImageNet-1K and 78.29% on CIFAR-100 with a ResNet-50 encoder while converging in just 200 epochs; (2) yields substantially clearer decision boundaries and semantic organization in the embedding space, as evidenced by KNN classification, hierarchical clustering results, and transfer-learning assessments; and (3) demonstrates superior performance in few-shot learning than supervised baseline and superior robustness across various augmentation strategies.
Link prediction in multilayer networks is a key challenge in applications such as recommendation systems and protein-protein interaction prediction. While many techniques have been developed, most rely on assumptions about shared structures and require access to raw auxiliary data, limiting their practicality. To address these issues, we propose a novel transfer learning framework for multilayer networks using a bi-level model averaging method. A $K$-fold cross-validation criterion based on edges is used to automatically weight inter-layer and intra-layer candidate models. This enables the transfer of information from auxiliary layers while mitigating model uncertainty, even without prior knowledge of shared structures. Theoretically, we prove the optimality and weight convergence of our method under mild conditions. Computationally, our framework is efficient and privacy-preserving, as it avoids raw data sharing and supports parallel processing across multiple servers. Simulations show our method outperforms others in predictive accuracy and robustness. We further demonstrate its practical value through two real-world recommendation system applications.
In this paper, we develop a novel transfer learning methodology to tackle the challenge of data scarcity in functional linear models. The methodology incorporates samples from the target model (target domain) alongside those from auxiliary models (source domains), transferring knowledge of coefficient shape from the source domains to the target domain. This shape-based knowledge transfer offers two key advantages. First, it is robust to covariate scaling, ensuring effectiveness despite variations in data distributions across different source domains. Second, the notion of coefficient shape homogeneity represents a meaningful advance beyond traditional coefficient homogeneity, allowing the method to exploit a wider range of source domains and achieve significantly improved model estimation. We rigorously analyze the convergence rates of the proposed estimator and examine the minimax optimality. Our findings show that the degree of improvement depends not only on the similarity of coefficient shapes between the target and source domains, but also on coefficient magnitudes and the spectral decay rates of the functional covariates covariance operators. To address situations where only a subset of auxiliary models is informative for the target model, we further develop a data-driven procedure for identifying such informative sources. The effectiveness of the proposed methodology is demonstrated through comprehensive simulation studies and an application to occupation time analysis using physical activity data from the U.S. National Health and Nutrition Examination Survey.
Lengthy subject- or session-specific data acquisition and calibration remain a key barrier to deploying electroencephalography (EEG)-based brain-computer interfaces (BCIs) outside the laboratory. Previous work has shown that cross subject, cross-session invariant features exist in EEG. We propose a transfer learning pipeline based on a two-layer convolutional neural network (CNN) that leverages these invariants to reduce the burden of data acquisition and calibration. A baseline model is trained on EEG data from five able-bodied individuals and then rapidly updated with a small amount of data from a sixth, holdout subject. The remaining holdout data were used to test the performance of both the baseline and updated models. We repeated this procedure via a leave-one-subject out (LOSO) validation framework. Averaged over six LOSO folds, the updated model improved classification accuracy upon the baseline by 10.0, 18.8, and 22.1 percentage points on two binary and one ternary classification tasks, respectively. These results demonstrate that decoding accuracy can be substantially improved with minimal subject-specific data. They also indicate that a CNN-based decoder can be personalized rapidly, enabling near plug-and-play BCI functionality for neurorehabilitation and other time-critical EEG applications.
Electrocardiogram (ECG), as a crucial find-grained cardiac feature, has been successfully recovered from radar signals in the literature, but the performance heavily relies on the high-quality radar signal and numerous radar-ECG pairs for training, restricting the applications in new scenarios due to data scarcity. Therefore, this work will focus on radar-based ECG recovery in new scenarios with limited data and propose a cardio-focusing and -tracking (CFT) algorithm to precisely track the cardiac location to ensure an efficient acquisition of high-quality radar signals. Furthermore, a transfer learning model (RFcardi) is proposed to extract cardio-related information from the radar signal without ECG ground truth based on the intrinsic sparsity of cardiac features, and only a few synchronous radar-ECG pairs are required to fine-tune the pre-trained model for the ECG recovery. The experimental results reveal that the proposed CFT can dynamically identify the cardiac location, and the RFcardi model can effectively generate faithful ECG recoveries after using a small number of radar-ECG pairs for training. The code and dataset are available after the publication.
Multi-condition fault diagnosis is prevalent in industrial systems and presents substantial challenges for conventional diagnostic approaches. The discrepancy in data distributions across different operating conditions degrades model performance when a model trained under one condition is applied to others. With the recent advancements in deep learning, transfer learning has been introduced to the fault diagnosis field as a paradigm for addressing multi-condition fault diagnosis. Among these methods, domain generalization approaches can handle complex scenarios by extracting condition-invariant fault features. Although many studies have considered fault diagnosis in specific multi-condition scenarios, the extent to which operating conditions affect fault information has been scarcely studied, which is crucial. However, the extent to which operating conditions affect fault information has been scarcely studied, which is crucial. When operating conditions have a significant impact on fault features, directly applying domain generalization methods may lead the model to learn condition-specific information, thereby reducing its overall generalization ability. This paper investigates the performance of existing end-to-end domain generalization methods under varying conditions, specifically in variable-speed and variable-load scenarios, using multiple experiments on a real-world gearbox. Additionally, a two-stage diagnostic framework is proposed, aiming to improve fault diagnosis performance under scenarios with significant operating condition impacts. By incorporating a domain-generalized encoder with a retraining strategy, the framework is able to extract condition-invariant fault features while simultaneously alleviating potential overfitting to the source domain. Several experiments on a real-world gearbox dataset are conducted to validate the effectiveness of the proposed approach.
Large-scale foundation models, including neural network interatomic potentials (NIPs) in computational materials science, have demonstrated significant potential. However, despite their success in accelerating atomistic simulations, NIPs face challenges in directly predicting electronic properties and often require coupling to higher-scale models or extensive simulations for macroscopic properties. Machine learning (ML) offers alternatives for structure-to-property mapping but faces trade-offs: feature-based methods often lack generalizability, while deep neural networks require significant data and computational power. To address these trade-offs, we introduce HackNIP, a two-stage pipeline that leverages pretrained NIPs. This method first extracts fixed-length feature vectors (embeddings) from NIP foundation models and then uses these embeddings to train shallow ML models for downstream structure-to-property predictions. This study investigates whether such a hybridization approach, by ``hacking" the NIP, can outperform end-to-end deep neural networks, determines the dataset size at which this transfer learning approach surpasses direct fine-tuning of the NIP, and identifies which NIP embedding depths yield the most informative features. HackNIP is benchmarked on Matbench, evaluated for data efficiency, and tested on diverse tasks including \textit{ab initio}, experimental, and molecular properties. We also analyze how embedding depth impacts performance. This work demonstrates a hybridization strategy to overcome ML trade-offs in materials science, aiming to democratize high-performance predictive modeling.
A Bayesian knowledge transfer mechanism that leverages external information to improve the performance of the Kalman fixed-lag interval smoother (FLIS) is proposed. Exact knowledge of the external observation model is assumed to be missing, which hinders the direct application of Bayes' rule in traditional transfer learning approaches. This limitation is overcome by the fully probabilistic design, conditioning the targeted task of state estimation on external information. To mitigate the negative impact of inaccurate external data while leveraging precise information, a latent variable is introduced. Favorably, in contrast to a filter, FLIS retrospectively refines past decisions up to a fixed time horizon, reducing the accumulation of estimation error and consequently improving the performance of state inference. Simulations indicate that the proposed algorithm better exploits precise external knowledge compared to a similar technique and achieves comparable results when the information is imprecise.
Accurate graph similarity is critical for knowledge transfer in VLSI design, enabling the reuse of prior solutions to reduce engineering effort and turnaround time. We propose Pieceformer, a scalable, self-supervised similarity assessment framework, equipped with a hybrid message-passing and graph transformer encoder. To address transformer scalability, we incorporate a linear transformer backbone and introduce a partitioned training pipeline for efficient memory and parallelism management. Evaluations on synthetic and real-world CircuitNet datasets show that Pieceformer reduces mean absolute error (MAE) by 24.9% over the baseline and is the only method to correctly cluster all real-world design groups. We further demonstrate the practical usage of our model through a case study on a partitioning task, achieving up to 89% runtime reduction. These results validate the framework's effectiveness for scalable, unbiased design reuse in modern VLSI systems.
In this paper, the physics informed neural networks (PINNs) is employed for the numerical simulation of heat transfer involving a moving source. To reduce the computational effort, a new training method is proposed that uses a continuous time-stepping through transfer learning. Within this, the time interval is divided into smaller intervals and a single network is initialized. On this single network each time interval is trained with the initial condition for (n+1)th as the solution obtained at nth time increment. Thus, this framework enables the computation of large temporal intervals without increasing the complexity of the network itself. The proposed framework is used to estimate the temperature distribution in a homogeneous medium with a moving heat source. The results from the proposed framework is compared with traditional finite element method and a good agreement is seen.
We propose co-creative learning as a novel paradigm where humans and AI, i.e., biological and artificial agents, mutually integrate their partial perceptual information and knowledge to construct shared external representations, a process we interpret as symbol emergence. Unlike traditional AI teaching based on unilateral knowledge transfer, this addresses the challenge of integrating information from inherently different modalities. We empirically test this framework using a human-AI interaction model based on the Metropolis-Hastings naming game (MHNG), a decentralized Bayesian inference mechanism. In an online experiment, 69 participants played a joint attention naming game (JA-NG) with one of three computer agent types (MH-based, always-accept, or always-reject) under partial observability. Results show that human-AI pairs with an MH-based agent significantly improved categorization accuracy through interaction and achieved stronger convergence toward a shared sign system. Furthermore, human acceptance behavior aligned closely with the MH-derived acceptance probability. These findings provide the first empirical evidence for co-creative learning emerging in human-AI dyads via MHNG-based interaction. This suggests a promising path toward symbiotic AI systems that learn with humans, rather than from them, by dynamically aligning perceptual experiences, opening a new venue for symbiotic AI alignment.
Transfer learning enhances the training of novel sensory and decision models by employing rich feature representations from large, pre-trained teacher models. Cognitive neuroscience shows that the human brain creates low-dimensional, abstract representations for efficient sensorimotor coding. Importantly, the brain can learn these representations with significantly fewer data points and less computational power than artificial models require. We introduce Brain2Model Transfer Learning (B2M), a framework where neural activity from human sensory and decision-making tasks acts as the teacher model for training artificial neural networks. We propose two B2M strategies: (1) Brain Contrastive Transfer, which aligns brain activity and network activations through a contrastive objective; and (2) Brain Latent Transfer, which projects latent dynamics from similar cognitive tasks onto student networks via supervised regression of brain-derived features. We validate B2M in memory-based decision-making with a recurrent neural network and scene reconstruction for autonomous driving with a variational autoencoder. The results show that student networks benefiting from brain-based transfer converge faster and achieve higher predictive accuracy than networks trained in isolation. Our findings indicate that the brain's representations are valuable for artificial learners, paving the way for more efficient learning of complex decision-making representations, which would be costly or slow through purely artificial training.
Pragmatics, the ability to infer meaning beyond literal interpretation, is crucial for social cognition and communication. While LLMs have been benchmarked for their pragmatic understanding, improving their performance remains underexplored. Existing methods rely on annotated labels but overlook the reasoning process humans naturally use to interpret implicit meaning. To bridge this gap, we introduce a novel pragmatic dataset, ImpliedMeaningPreference, that includes explicit reasoning (thoughts) for both correct and incorrect interpretations. Through preference-tuning and supervised fine-tuning, we demonstrate that thought-based learning significantly enhances LLMs' pragmatic understanding, improving accuracy by 11.12% across model families. We further discuss a transfer-learning study where we evaluate the performance of thought-based training for the other tasks of pragmatics (presupposition, deixis) that are not seen during the training time and observe an improvement of 16.10% compared to label-trained models.
Person re-identification (ReID) has evolved from handcrafted feature-based methods to deep learning approaches and, more recently, to models incorporating large language models (LLMs). Early methods struggled with variations in lighting, pose, and viewpoint, but deep learning addressed these issues by learning robust visual features. Building on this, LLMs now enable ReID systems to integrate semantic and contextual information through natural language. This survey traces that full evolution and offers one of the first comprehensive reviews of ReID approaches that leverage LLMs, where textual descriptions are used as privileged information to improve visual matching. A key contribution is the use of dynamic, identity-specific prompts generated by GPT-4o, which enhance the alignment between images and text in vision-language ReID systems. Experimental results show that these descriptions improve accuracy, especially in complex or ambiguous cases. To support further research, we release a large set of GPT-4o-generated descriptions for standard ReID datasets. By bridging computer vision and natural language processing, this survey offers a unified perspective on the field's development and outlines key future directions such as better prompt design, cross-modal transfer learning, and real-world adaptability.
Cross-domain time series imputation is an underexplored data-centric research task that presents significant challenges, particularly when the target domain suffers from high missing rates and domain shifts in temporal dynamics. Existing time series imputation approaches primarily focus on the single-domain setting, which cannot effectively adapt to a new domain with domain shifts. Meanwhile, conventional domain adaptation techniques struggle with data incompleteness, as they typically assume the data from both source and target domains are fully observed to enable adaptation. For the problem of cross-domain time series imputation, missing values introduce high uncertainty that hinders distribution alignment, making existing adaptation strategies ineffective. Specifically, our proposed solution tackles this problem from three perspectives: (i) Data: We introduce a frequency-based time series interpolation strategy that integrates shared spectral components from both domains while retaining domain-specific temporal structures, constructing informative priors for imputation. (ii) Model: We design a diffusion-based imputation model that effectively learns domain-shared representations and captures domain-specific temporal dependencies with dedicated denoising networks. (iii) Algorithm: We further propose a cross-domain consistency alignment strategy that selectively regularizes output-level domain discrepancies, enabling effective knowledge transfer while preserving domain-specific characteristics. Extensive experiments on three real-world datasets demonstrate the superiority of our proposed approach. Our code implementation is available here.