While conditional diffusion models have achieved remarkable success in various applications, they require abundant data to train from scratch, which is often infeasible in practice. To address this issue, transfer learning has emerged as an essential paradigm in small data regimes. Despite its empirical success, the theoretical underpinnings of transfer learning conditional diffusion models remain unexplored. In this paper, we take the first step towards understanding the sample efficiency of transfer learning conditional diffusion models through the lens of representation learning. Inspired by practical training procedures, we assume that there exists a low-dimensional representation of conditions shared across all tasks. Our analysis shows that with a well-learned representation from source tasks, the samplecomplexity of target tasks can be reduced substantially. In addition, we investigate the practical implications of our theoretical results in several real-world applications of conditional diffusion models. Numerical experiments are also conducted to verify our results.
This paper investigates the graph neural network (GNN)-enabled beamforming design for interference channels. We propose a model termed interference channel GNN (ICGNN) to solve a quality-of-service constrained energy efficiency maximization problem. The ICGNN is two-stage, where the direction and power parts of beamforming vectors are learned separately but trained jointly via unsupervised learning. By formulating the dimensionality of features independent of the transceiver pairs, the ICGNN is scalable with the number of transceiver pairs. Besides, to improve the performance of the ICGNN, the hybrid maximum ratio transmission and zero-forcing scheme reduces the output ports, the feature enhancement module unifies the two types of links into one type, the subgraph representation enhances the message passing efficiency, and the multi-head attention and residual connection facilitate the feature extracting. Furthermore, we present the over-the-air distributed implementation of the ICGNN. Ablation studies validate the effectiveness of key components in the ICGNN. Numerical results also demonstrate the capability of ICGNN in achieving near-optimal performance with an average inference time less than 0.1 ms. The scalability of ICGNN for unseen problem sizes is evaluated and enhanced by transfer learning with limited fine-tuning cost. The results of the centralized and distributed implementations of ICGNN are illustrated.
This paper investigates the deep learning based approaches for simultaneous wireless information and power transfer (SWIPT). The quality-of-service (QoS) constrained sum-rate maximization problems are, respectively, formulated for power-splitting (PS) receivers and time-switching (TS) receivers and solved by a unified graph neural network (GNN) based model termed SWIPT net (SWIPTNet). To improve the performance of SWIPTNet, we first propose a single-type output method to reduce the learning complexity and facilitate the satisfaction of QoS constraints, and then, utilize the Laplace transform to enhance input features with the structural information. Besides, we adopt the multi-head attention and layer connection to enhance feature extracting. Furthermore, we present the implementation of transfer learning to the SWIPTNet between PS and TS receivers. Ablation studies show the effectiveness of key components in the SWIPTNet. Numerical results also demonstrate the capability of SWIPTNet in achieving near-optimal performance with millisecond-level inference speed which is much faster than the traditional optimization algorithms. We also show the effectiveness of transfer learning via fast convergence and expressive capability improvement.
Automatic Speech Recognition (ASR) performance for low-resource languages is still far behind that of higher-resource languages such as English, due to a lack of sufficient labeled data. State-of-the-art methods deploy self-supervised transfer learning where a model pre-trained on large amounts of data is fine-tuned using little labeled data in a target low-resource language. In this paper, we present and examine a method for fine-tuning an SSL-based model in order to improve the performance for Frisian and its regional dialects (Clay Frisian, Wood Frisian, and South Frisian). We show that Frisian ASR performance can be improved by using multilingual (Frisian, Dutch, English and German) fine-tuning data and an auxiliary language identification task. In addition, our findings show that performance on dialectal speech suffers substantially, and, importantly, that this effect is moderated by the elicitation approach used to collect the dialectal data. Our findings also particularly suggest that relying solely on standard language data for ASR evaluation may underestimate real-world performance, particularly in languages with substantial dialectal variation.
Multimodal Federated Learning (MFL) enables multiple clients to collaboratively train models on multimodal data while ensuring clients' privacy. However, modality and task heterogeneity hinder clients from learning a unified representation, weakening local model generalization, especially in MFL with mixed modalities where only some clients have multimodal data. In this work, we propose an Adaptive prototype-based Multimodal Federated Learning (AproMFL) framework for mixed modalities and heterogeneous tasks to address the aforementioned issues. Our AproMFL transfers knowledge through adaptively-constructed prototypes without a prior public dataset. Clients adaptively select prototype construction methods in line with tasks; server converts client prototypes into unified multimodal prototypes and aggregates them to form global prototypes, avoid clients keeping unified labels. We divide the model into various modules and only aggregate mapping modules to reduce communication and computation overhead. To address aggregation issues in heterogeneity, we develop a client relationship graph-based scheme to dynamically adjust aggregation weights. Extensive experiments on representative datasets evidence effectiveness of AproMFL.
The design of analog circuits is a cornerstone of integrated circuit (IC) development, requiring the optimization of complex, interconnected sub-structures such as amplifiers, comparators, and buffers. Traditionally, this process relies heavily on expert human knowledge to refine design objectives by carefully tuning sub-components while accounting for their interdependencies. Existing methods, such as Bayesian Optimization (BO), offer a mathematically driven approach for efficiently navigating large design spaces. However, these methods fall short in two critical areas compared to human expertise: (i) they lack the semantic understanding of the sizing solution space and its direct correlation with design objectives before optimization, and (ii) they fail to reuse knowledge gained from optimizing similar sub-structures across different circuits. To overcome these limitations, we propose the Large Language Model-based Universal Sizing Optimizer (LLM-USO), which introduces a novel method for knowledge representation to encode circuit design knowledge in a structured text format. This representation enables the systematic reuse of optimization insights for circuits with similar sub-structures. LLM-USO employs a hybrid framework that integrates BO with large language models (LLMs) and a learning summary module. This approach serves to: (i) infuse domain-specific knowledge into the BO process and (ii) facilitate knowledge transfer across circuits, mirroring the cognitive strategies of expert designers. Specifically, LLM-USO constructs a knowledge summary mechanism to distill and apply design insights from one circuit to related ones. It also incorporates a knowledge summary critiquing mechanism to ensure the accuracy and quality of the summaries and employs BO-guided suggestion filtering to identify optimal design points efficiently.
How to mitigate negative transfer in transfer learning is a long-standing and challenging issue, especially in the application of medical image segmentation. Existing methods for reducing negative transfer focus on classification or regression tasks, ignoring the non-uniform negative transfer risk in different image regions. In this work, we propose a simple yet effective weighted fine-tuning method that directs the model's attention towards regions with significant transfer risk for medical semantic segmentation. Specifically, we compute a transferability-guided transfer risk map to quantify the transfer hardness for each pixel and the potential risks of negative transfer. During the fine-tuning phase, we introduce a map-weighted loss function, normalized with image foreground size to counter class imbalance. Extensive experiments on brain segmentation datasets show our method significantly improves the target task performance, with gains of 4.37% on FeTS2021 and 1.81% on iSeg2019, avoiding negative transfer across modalities and tasks. Meanwhile, a 2.9% gain under a few-shot scenario validates the robustness of our approach.
When training large flexible models, practitioners often rely on grid search to select hyperparameters that control over-fitting. This grid search has several disadvantages: the search is computationally expensive, requires carving out a validation set that reduces the available data for training, and requires users to specify candidate values. In this paper, we propose an alternative: directly learning regularization hyperparameters on the full training set via the evidence lower bound ("ELBo") objective from variational methods. For deep neural networks with millions of parameters, we recommend a modified ELBo that upweights the influence of the data likelihood relative to the prior. Our proposed technique overcomes all three disadvantages of grid search. In a case study on transfer learning of image classifiers, we show how our method reduces the 88+ hour grid search of past work to under 3 hours while delivering comparable accuracy. We further demonstrate how our approach enables efficient yet accurate approximations of Gaussian processes with learnable length-scale kernels.
3D cell segmentation methods are often hindered by \emph{oversegmentation}, where a single cell is incorrectly split into multiple fragments. This degrades the final segmentation quality and is notoriously difficult to resolve, as oversegmentation errors often resemble \emph{natural gaps} between adjacent cells. Our work makes two key contributions. First, for 3D cell segmentation, we are the first work to formulate oversegmentation as a concrete problem and propose a geometric framework to identify and correct these errors. Our approach builds a pre-trained classifier using both 2D geometric and 3D topological features extracted from flawed 3D segmentation results. Second, we introduce a novel metric, \emph{Geo-Wasserstein} divergence, to quantify changes in 2D geometries. This captures the evolving trends in cell mask shape changes in a geometry-aware manner. We validate our method through extensive experiments on in-domain plant datasets, including both synthesized and real cases, as well as on out-of-domain animal datasets to demonstrate transfer learning performance. An ablation study further highlights the contribution of the \emph{Geo-Wasserstein} divergence. A clear pipeline is provided for end-users to build pre-trained models to any labeled dataset.
Self-supervised learning (SSL) algorithms have emerged as powerful tools that can leverage large quantities of unlabeled audio data to pre-train robust representations that support strong performance on diverse downstream tasks. Up to now these have mostly been developed separately for speech and non-speech applications. Here, we explored the domain specificity of a convolutional model's pre-training data relative to different downstream speech and non-speech tasks using a self-supervised pre-training approach (BYOL-A). We found that these pre-trained models (regardless of whether they were pre-trained on speech data, non-speech data or both) enabled good performance on nearly all downstream tasks, beating or nearly matching the performance of popular domain-specific models. Only small domain-specificity advantages were observed between the different pre-training datasets. The popular domain-specific models used as baselines performed very well in their target domains, but generally faltered outside of them. Together, these results demonstrate that SSL methods can be a powerful way to learn flexible representations for domain specific data without labels. These models can be a powerful resource for later transfer learning, fine-tuning or data exploration applications when the downstream data are similar, but also perhaps when there may be a domain mismatch.
Resting-state functional magnetic resonance imaging (rs-fMRI) and its derived functional connectivity networks (FCNs) have become critical for understanding neurological disorders. However, collaborative analyses and the generalizability of models still face significant challenges due to privacy regulations and the non-IID (non-independent and identically distributed) property of multiple data sources. To mitigate these difficulties, we propose Domain Adversarial Federated Learning (DAFed), a novel federated deep learning framework specifically designed for non-IID fMRI data analysis in multi-site settings. DAFed addresses these challenges through feature disentanglement, decomposing the latent feature space into domain-invariant and domain-specific components, to ensure robust global learning while preserving local data specificity. Furthermore, adversarial training facilitates effective knowledge transfer between labeled and unlabeled datasets, while a contrastive learning module enhances the global representation of domain-invariant features. We evaluated DAFed on the diagnosis of ASD and further validated its generalizability in the classification of AD, demonstrating its superior classification accuracy compared to state-of-the-art methods. Additionally, an enhanced Score-CAM module identifies key brain regions and functional connectivity significantly associated with ASD and MCI, respectively, uncovering shared neurobiological patterns across sites. These findings highlight the potential of DAFed to advance multi-site collaborative research in neuroimaging while protecting data confidentiality.
Lensless cameras disregard the conventional design that imaging should mimic the human eye. This is done by replacing the lens with a thin mask, and moving image formation to the digital post-processing. State-of-the-art lensless imaging techniques use learned approaches that combine physical modeling and neural networks. However, these approaches make simplifying modeling assumptions for ease of calibration and computation. Moreover, the generalizability of learned approaches to lensless measurements of new masks has not been studied. To this end, we utilize a modular learned reconstruction in which a key component is a pre-processor prior to image recovery. We theoretically demonstrate the pre-processor's necessity for standard image recovery techniques (Wiener filtering and iterative algorithms), and through extensive experiments show its effectiveness for multiple lensless imaging approaches and across datasets of different mask types (amplitude and phase). We also perform the first generalization benchmark across mask types to evaluate how well reconstructions trained with one system generalize to others. Our modular reconstruction enables us to use pre-trained components and transfer learning on new systems to cut down weeks of tedious measurements and training. As part of our work, we open-source four datasets, and software for measuring datasets and for training our modular reconstruction.
The analysis of wearable sensor data has enabled many successes in several applications. To represent the high-sampling rate time-series with sufficient detail, the use of topological data analysis (TDA) has been considered, and it is found that TDA can complement other time-series features. Nonetheless, due to the large time consumption and high computational resource requirements of extracting topological features through TDA, it is difficult to deploy topological knowledge in various applications. To tackle this problem, knowledge distillation (KD) can be adopted, which is a technique facilitating model compression and transfer learning to generate a smaller model by transferring knowledge from a larger network. By leveraging multiple teachers in KD, both time-series and topological features can be transferred, and finally, a superior student using only time-series data is distilled. On the other hand, mixup has been popularly used as a robust data augmentation technique to enhance model performance during training. Mixup and KD employ similar learning strategies. In KD, the student model learns from the smoothed distribution generated by the teacher model, while mixup creates smoothed labels by blending two labels. Hence, this common smoothness serves as the connecting link that establishes a connection between these two methods. In this paper, we analyze the role of mixup in KD with time-series as well as topological persistence, employing multiple teachers. We present a comprehensive analysis of various methods in KD and mixup on wearable sensor data.
MRI quality control (QC) is challenging due to unbalanced and limited datasets, as well as subjective scoring, which hinder the development of reliable automated QC systems. To address these issues, we introduce an approach that pretrains a model on synthetically generated motion artifacts before applying transfer learning for QC classification. This method not only improves the accuracy in identifying poor-quality scans but also reduces training time and resource requirements compared to training from scratch. By leveraging synthetic data, we provide a more robust and resource-efficient solution for QC automation in MRI, paving the way for broader adoption in diverse research settings.
This study explores computational approaches for measuring moral foundations (MFs) in non-English corpora. Since most resources are developed primarily for English, cross-linguistic applications of moral foundation theory remain limited. Using Chinese as a case study, this paper evaluates the effectiveness of applying English resources to machine translated text, local language lexicons, multilingual language models, and large language models (LLMs) in measuring MFs in non-English texts. The results indicate that machine translation and local lexicon approaches are insufficient for complex moral assessments, frequently resulting in a substantial loss of cultural information. In contrast, multilingual models and LLMs demonstrate reliable cross-language performance with transfer learning, with LLMs excelling in terms of data efficiency. Importantly, this study also underscores the need for human-in-the-loop validation of automated MF assessment, as the most advanced models may overlook cultural nuances in cross-language measurements. The findings highlight the potential of LLMs for cross-language MF measurements and other complex multilingual deductive coding tasks.
Recent advances in CV and NLP have inspired researchers to develop general-purpose graph foundation models through pre-training across diverse domains. However, a fundamental challenge arises from the substantial differences in graph topologies across domains. Additionally, real-world graphs are often sparse and prone to noisy connections and adversarial attacks. To address these issues, we propose the Multi-Domain Graph Foundation Model (MDGFM), a unified framework that aligns and leverages cross-domain topological information to facilitate robust knowledge transfer. MDGFM bridges different domains by adaptively balancing features and topology while refining original graphs to eliminate noise and align topological structures. To further enhance knowledge transfer, we introduce an efficient prompt-tuning approach. By aligning topologies, MDGFM not only improves multi-domain pre-training but also enables robust knowledge transfer to unseen domains. Theoretical analyses provide guarantees of MDGFM's effectiveness and domain generalization capabilities. Extensive experiments on both homophilic and heterophilic graph datasets validate the robustness and efficacy of our method.
Air pollution (AP) poses a great threat to human health, and people are paying more attention than ever to its prediction. Accurate prediction of AP helps people to plan for their outdoor activities and aids protecting human health. In this paper, long-short term memory (LSTM) recurrent neural networks (RNNs) have been used to predict the future concentration of air pollutants (APS) in Macau. Additionally, meteorological data and data on the concentration of APS have been utilized. Moreover, in Macau, some air quality monitoring stations (AQMSs) have less observed data in quantity, and, at the same time, some AQMSs recorded less observed data of certain types of APS. Therefore, the transfer learning and pre-trained neural networks have been employed to assist AQMSs with less observed data to build a neural network with high prediction accuracy. The experimental sample covers a period longer than 12-year and includes daily measurements from several APS as well as other more classical meteorological values. Records from five stations, four out of them are AQMSs and the remaining one is an automatic weather station, have been prepared from the aforesaid period and eventually underwent to computational intelligence techniques to build and extract a prediction knowledge-based system. As shown by experimentation, LSTM RNNs initialized with transfer learning methods have higher prediction accuracy; it incurred shorter training time than randomly initialized recurrent neural networks.
The aging sewer infrastructure in the U.S., covering 2.1 million kilometers, encounters increasing structural issues, resulting in around 75,000 yearly sanitary sewer overflows that present serious economic, environmental, and public health hazards. Conventional inspection techniques and deterministic models do not account for the unpredictable nature of sewer decline, whereas probabilistic methods depend on extensive historical data, which is frequently lacking or incomplete. This research intends to enhance predictive accuracy for the condition of sewer pipelines through machine learning models artificial neural networks (ANNs) and multiple linear regression (MLR) by integrating factors such as pipe age, material, diameter, environmental influences, and PACP ratings. ANNs utilized ReLU activation functions and Adam optimization, whereas MLR applied regularization to address multicollinearity, with both models assessed through metrics like RMSE, MAE, and R2. The findings indicated that ANNs surpassed MLR, attaining an R2 of 0.9066 compared to MLRs 0.8474, successfully modeling nonlinear relationships while preserving generalization. MLR, on the other hand, offered enhanced interpretability by pinpointing significant predictors such as residual buildup. As a result, pipeline degradation is driven by pipe length, age, and pipe diameter as key predictors, while depth, soil type, and segment show minimal influence in this analysis. Future studies ought to prioritize hybrid models that merge the accuracy of ANNs with the interpretability of MLR, incorporating advanced methods such as SHAP analysis and transfer learning to improve scalability in managing infrastructure and promoting environmental sustainability.
Self Supervised Representation Learning (SSRepL) can capture meaningful and robust representations of the Attention Deficit Hyperactivity Disorder (ADHD) data and have the potential to improve the model's performance on also downstream different types of Neurodevelopmental disorder (NDD) detection. In this paper, a novel SSRepL and Transfer Learning (TL)-based framework that incorporates a Long Short-Term Memory (LSTM) and a Gated Recurrent Units (GRU) model is proposed to detect children with potential symptoms of ADHD. This model uses Electroencephalogram (EEG) signals extracted during visual attention tasks to accurately detect ADHD by preprocessing EEG signal quality through normalization, filtering, and data balancing. For the experimental analysis, we use three different models: 1) SSRepL and TL-based LSTM-GRU model named as SSRepL-ADHD, which integrates LSTM and GRU layers to capture temporal dependencies in the data, 2) lightweight SSRepL-based DNN model (LSSRepL-DNN), and 3) Random Forest (RF). In the study, these models are thoroughly evaluated using well-known performance metrics (i.e., accuracy, precision, recall, and F1-score). The results show that the proposed SSRepL-ADHD model achieves the maximum accuracy of 81.11% while admitting the difficulties associated with dataset imbalance and feature selection.
Global Coconut (Cocos nucifera (L.)) cultivation faces significant challenges, including yield loss, due to pest and disease outbreaks. In particular, Weligama Coconut Leaf Wilt Disease (WCWLD) and Coconut Caterpillar Infestation (CCI) damage coconut trees, causing severe coconut production loss in Sri Lanka and nearby coconut-producing countries. Currently, both WCWLD and CCI are detected through on-field human observations, a process that is not only time-consuming but also limits the early detection of infections. This paper presents a study conducted in Sri Lanka, demonstrating the effectiveness of employing transfer learning-based Convolutional Neural Network (CNN) and Mask Region-based-CNN (Mask R-CNN) to identify WCWLD and CCI at their early stages and to assess disease progression. Further, this paper presents the use of the You Only Look Once (YOLO) object detection model to count the number of caterpillars distributed on leaves with CCI. The introduced methods were tested and validated using datasets collected from Matara, Puttalam, and Makandura, Sri Lanka. The results show that the proposed methods identify WCWLD and CCI with an accuracy of 90% and 95%, respectively. In addition, the proposed WCWLD disease severity identification method classifies the severity with an accuracy of 97%. Furthermore, the accuracies of the object detection models for calculating the number of caterpillars in the leaflets were: YOLOv5-96.87%, YOLOv8-96.1%, and YOLO11-95.9%.