This work presents a novel algorithm for progressively adapting neural network architecture along the depth. In particular, we attempt to address the following questions in a mathematically principled way: i) Where to add a new capacity (layer) during the training process? ii) How to initialize the new capacity? At the heart of our approach are two key ingredients: i) the introduction of a ``shape functional" to be minimized, which depends on neural network topology, and ii) the introduction of a topological derivative of the shape functional with respect to the neural network topology. Using an optimal control viewpoint, we show that the network topological derivative exists under certain conditions, and its closed-form expression is derived. In particular, we explore, for the first time, the connection between the topological derivative from a topology optimization framework with the Hamiltonian from optimal control theory. Further, we show that the optimality condition for the shape functional leads to an eigenvalue problem for deep neural architecture adaptation. Our approach thus determines the most sensitive location along the depth where a new layer needs to be inserted during the training phase and the associated parametric initialization for the newly added layer. We also demonstrate that our layer insertion strategy can be derived from an optimal transport viewpoint as a solution to maximizing a topological derivative in $p$-Wasserstein space, where $p>= 1$. Numerical investigations with fully connected network, convolutional neural network, and vision transformer on various regression and classification problems demonstrate that our proposed approach can outperform an ad-hoc baseline network and other architecture adaptation strategies. Further, we also demonstrate other applications of topological derivative in fields such as transfer learning.
In recent years, Graph Neural Network (GNN) based models have shown promising results in simulating physics of complex systems. However, training dedicated graph network based physics simulators can be costly, as most models are confined to fully supervised training, which requires extensive data generated from traditional physics simulators. To date, how transfer learning could improve the model performance and training efficiency has remained unexplored. In this work, we introduce a pre-training and transfer learning paradigm for graph network simulators. We propose the scalable graph U-net (SGUNET). Incorporating an innovative depth-first search (DFS) pooling, the SGUNET is adaptable to different mesh sizes and resolutions for various simulation tasks. To enable the transfer learning between differently configured SGUNETs, we propose a set of mapping functions to align the parameters between the pre-trained model and the target model. An extra normalization term is also added into the loss to constrain the difference between the pre-trained weights and target model weights for better generalization performance. To pre-train our physics simulator we created a dataset which includes 20,000 physical simulations of randomly selected 3D shapes from the open source A Big CAD (ABC) dataset. We show that our proposed transfer learning methods allow the model to perform even better when fine-tuned with small amounts of training data than when it is trained from scratch with full extensive dataset. On the 2D Deformable Plate benchmark dataset, our pre-trained model fine-tuned on 1/16 of the training data achieved an 11.05\% improvement in position RMSE compared to the model trained from scratch.
Data-driven soft sensors help in process operations by providing real-time estimates of otherwise hard- to-measure process quantities, e.g., viscosities or product concentrations. Currently, soft sensors need to be developed individually per plant. Using transfer learning, machine learning-based soft sensors could be reused and fine-tuned across plants and applications. However, transferring data-driven soft sensor models is in practice often not possible, because the fixed input structure of standard soft sensor models prohibits transfer if, e.g., the sensor information is not identical in all plants. We propose a topology-aware graph neural network approach for transfer learning of soft sensor models across multiple plants. In our method, plants are modeled as graphs: Unit operations are nodes, streams are edges, and sensors are embedded as attributes. Our approach brings two advantages for transfer learning: First, we not only include sensor data but also crucial information on the plant topology. Second, the graph neural network algorithm is flexible with respect to its sensor inputs. This allows us to model data from different plants with different sensor networks. We test the transfer learning capabilities of our modeling approach on ammonia synthesis loops with different process topologies. We build a soft sensor predicting the ammonia concentration in the product. After training on data from one process, we successfully transfer our soft sensor model to a previously unseen process with a different topology. Our approach promises to extend the data-driven soft sensors to cases to leverage data from multiple plants.
Recent advancement in code understanding and generation demonstrates that code LLMs fine-tuned on a high-quality instruction dataset can gain powerful capabilities to address wide-ranging code-related tasks. However, most previous existing methods mainly view each programming language in isolation and ignore the knowledge transfer among different programming languages. To bridge the gap among different programming languages, we introduce a novel multi-agent collaboration framework to enhance multilingual instruction tuning for code LLMs, where multiple language-specific intelligent agent components with generation memory work together to transfer knowledge from one language to another efficiently and effectively. Specifically, we first generate the language-specific instruction data from the code snippets and then provide the generated data as the seed data for language-specific agents. Multiple language-specific agents discuss and collaborate to formulate a new instruction and its corresponding solution (A new programming language or existing programming language), To further encourage the cross-lingual transfer, each agent stores its generation history as memory and then summarizes its merits and faults. Finally, the high-quality multilingual instruction data is used to encourage knowledge transfer among different programming languages to train Qwen2.5-xCoder. Experimental results on multilingual programming benchmarks demonstrate the superior performance of Qwen2.5-xCoder in sharing common knowledge, highlighting its potential to reduce the cross-lingual gap.
Artificial intelligence-assisted imaging analysis has made substantial strides in tumor diagnosis and management. Here we present PASTA, a pan-tumor CT foundation model that achieves state-of-the-art performance on 45 of 46 representative oncology tasks -- including lesion segmentation, tumor detection in plain CT, tumor staging, survival prediction, structured report generation, and cross-modality transfer learning, significantly outperforming the second-best models on 35 tasks. This remarkable advancement is driven by our development of PASTA-Gen, an innovative synthetic tumor generation framework that produces a comprehensive dataset of 30,000 CT scans with pixel-level annotated lesions and paired structured reports, encompassing malignancies across ten organs and five benign lesion types. By leveraging this rich, high-quality synthetic data, we overcome a longstanding bottleneck in the development of CT foundation models -- specifically, the scarcity of publicly available, high-quality annotated datasets due to privacy constraints and the substantial labor required for scaling precise data annotation. Encouragingly, PASTA demonstrates exceptional data efficiency with promising practical value, markedly improving performance on various tasks with only a small amount of real-world data. The open release of both the synthetic dataset and PASTA foundation model effectively addresses the challenge of data scarcity, thereby advancing oncological research and clinical translation.
Lipreading is an important technique for facilitating human-computer interaction in noisy environments. Our previously developed self-supervised learning method, AV2vec, which leverages multimodal self-distillation, has demonstrated promising performance in speaker-independent lipreading on the English LRS3 dataset. However, AV2vec faces challenges such as high training costs and a potential scarcity of audio-visual data for lipreading in languages other than English, such as Chinese. Additionally, most studies concentrate on speakerindependent lipreading models, which struggle to account for the substantial variation in speaking styles across di?erent speakers. To address these issues, we propose a comprehensive approach. First, we investigate cross-lingual transfer learning, adapting a pre-trained AV2vec model from a source language and optimizing it for the lipreading task in a target language. Second, we enhance the accuracy of lipreading for specific target speakers through a speaker adaptation strategy, which is not extensively explored in previous research. Third, after analyzing the complementary performance of lipreading with lip region-of-interest (ROI) and face inputs, we introduce a model ensembling strategy that integrates both, signi?cantly boosting model performance. Our method achieved a character error rate (CER) of 77.3% on the evaluation set of the ChatCLR dataset, which is lower than the top result from the 2024 Chat-scenario Chinese Lipreading Challenge.
Fingerprint localization has gained significant attention due to its cost-effective deployment, low complexity, and high efficacy. However, traditional methods, while effective for static data, often struggle in dynamic environments where data distributions and feature spaces evolve-a common occurrence in real-world scenarios. To address the challenges of robustness and adaptability in fingerprint localization for dynamic indoor environments, this paper proposes DF-Loc, an end-to-end dynamic fingerprint localization system based on multi-source unsupervised domain adaptation (MUDA). DF-Loc leverages historical data from multiple time scales to facilitate knowledge transfer in specific feature spaces, thereby enhancing generalization capabilities in the target domain and reducing reliance on labeled data. Specifically, the system incorporates a Quality Control (QC) module for CSI data preprocessing and employs image processing techniques for CSI fingerprint feature reconstruction. Additionally, a multi-scale attention-based feature fusion backbone network is designed to extract multi-level transferable fingerprint features. Finally, a dual-stage alignment model aligns the distributions of multiple source-target domain pairs, improving regression characteristics in the target domain. Extensive experiments conducted in office and classroom environments demonstrate that DF-Loc outperforms comparative methods in terms of both localization accuracy and robustness. With 60% of reference points used for training, DF-Loc achieves average localization errors of 0.79m and 3.72m in "same-test" scenarios, and 0.94m and 4.39m in "different-test" scenarios, respectively. This work pioneers an end-to-end multi-source transfer learning approach for fingerprint localization, providing valuable insights for future research in dynamic environments.
The advent of Music-Language Models has greatly enhanced the automatic music generation capability of AI systems, but they are also limited in their coverage of the musical genres and cultures of the world. We present a study of the datasets and research papers for music generation and quantify the bias and under-representation of genres. We find that only 5.7% of the total hours of existing music datasets come from non-Western genres, which naturally leads to disparate performance of the models across genres. We then investigate the efficacy of Parameter-Efficient Fine-Tuning (PEFT) techniques in mitigating this bias. Our experiments with two popular models -- MusicGen and Mustango, for two underrepresented non-Western music traditions -- Hindustani Classical and Turkish Makam music, highlight the promises as well as the non-triviality of cross-genre adaptation of music through small datasets, implying the need for more equitable baseline music-language models that are designed for cross-cultural transfer learning.
This paper proposes a Curriculum-Transfer-Learning based physics-informed neural network (CTL-PINN) for long-term simulation of physical and mechanical behaviors. The main innovation of CTL-PINN lies in decomposing long-term problems into a sequence of short-term subproblems. Initially, the standard PINN is employed to solve the first sub-problem. As the simulation progresses, subsequent time-domain problems are addressed using a curriculum learning approach that integrates information from previous steps. Furthermore, transfer learning techniques are incorporated, allowing the model to effectively utilize prior training data and solve sequential time domain transfer problems. CTL-PINN combines the strengths of curriculum learning and transfer learning, overcoming the limitations of standard PINNs, such as local optimization issues, and addressing the inaccuracies over extended time domains encountered in CL-PINN and the low computational efficiency of TL-PINN. The efficacy and robustness of CTL-PINN are demonstrated through applications to nonlinear wave propagation, Kirchhoff plate dynamic response, and the hydrodynamic model of the Three Gorges Reservoir Area, showcasing its superior capability in addressing long-term computational challenges.
In machine learning practice, early stopping has been widely used to regularize models and can save computational costs by halting the training process when the model's performance on a validation set stops improving. However, conventional early stopping applies the same stopping criterion to all instances without considering their individual learning statuses, which leads to redundant computations on instances that are already well-learned. To further improve the efficiency, we propose an Instance-dependent Early Stopping (IES) method that adapts the early stopping mechanism from the entire training set to the instance level, based on the core principle that once the model has mastered an instance, the training on it should stop. IES considers an instance as mastered if the second-order differences of its loss value remain within a small range around zero. This offers a more consistent measure of an instance's learning status compared with directly using the loss value, and thus allows for a unified threshold to determine when an instance can be excluded from further backpropagation. We show that excluding mastered instances from backpropagation can increase the gradient norms, thereby accelerating the decrease of the training loss and speeding up the training process. Extensive experiments on benchmarks demonstrate that IES method can reduce backpropagation instances by 10%-50% while maintaining or even slightly improving the test accuracy and transfer learning performance of a model.
EEG-based neural networks, pivotal in medical diagnosis and brain-computer interfaces, face significant intellectual property (IP) risks due to their reliance on sensitive neurophysiological data and resource-intensive development. Current watermarking methods, particularly those using abstract trigger sets, lack robust authentication and fail to address the unique challenges of EEG models. This paper introduces a cryptographic wonder filter-based watermarking framework tailored for EEG-based neural networks. Leveraging collision-resistant hashing and public-key encryption, the wonder filter embeds the watermark during training, ensuring minimal distortion ($\leq 5\%$ drop in EEG task accuracy) and high reliability (100\% watermark detection). The framework is rigorously evaluated against adversarial attacks, including fine-tuning, transfer learning, and neuron pruning. Results demonstrate persistent watermark retention, with classification accuracy for watermarked states remaining above 90\% even after aggressive pruning, while primary task performance degrades faster, deterring removal attempts. Piracy resistance is validated by the inability to embed secondary watermarks without severe accuracy loss ( $>10\%$ in EEGNet and CCNN models). Cryptographic hashing ensures authentication, reducing brute-force attack success probabilities. Evaluated on the DEAP dataset across models (CCNN, EEGNet, TSception), the method achieves $>99.4\%$ null-embedding accuracy, effectively eliminating false positives. By integrating wonder filters with EEG-specific adaptations, this work bridges a critical gap in IP protection for neurophysiological models, offering a secure, tamper-proof solution for healthcare and biometric applications. The framework's robustness against adversarial modifications underscores its potential to safeguard sensitive EEG models while maintaining diagnostic utility.
Surrogate models are widely used to approximate complex systems across science and engineering to reduce computational costs. Despite their widespread adoption, the field lacks standardisation across key stages of the modelling pipeline, including data sampling, model selection, evaluation, and downstream analysis. This fragmentation limits reproducibility and cross-domain utility -- a challenge further exacerbated by the rapid proliferation of AI-driven surrogate models. We argue for the urgent need to establish a structured reporting standard, the Surrogate Model Reporting Specification (SMRS), that systematically captures essential design and evaluation choices while remaining agnostic to implementation specifics. By promoting a standardised yet flexible framework, we aim to improve the reliability of surrogate modelling, foster interdisciplinary knowledge transfer, and, as a result, accelerate scientific progress in the AI era.
Multi-source transfer learning provides an effective solution to data scarcity in real-world supervised learning scenarios by leveraging multiple source tasks. In this field, existing works typically use all available samples from sources in training, which constrains their training efficiency and may lead to suboptimal results. To address this, we propose a theoretical framework that answers the question: what is the optimal quantity of source samples needed from each source task to jointly train the target model? Specifically, we introduce a generalization error measure that aligns with cross-entropy loss, and minimize it based on the Cram\'er-Rao Bound to determine the optimal transfer quantity for each source task. Additionally, we develop an architecture-agnostic and data-efficient algorithm OTQMS to implement our theoretical results for training deep multi-source transfer learning models. Experimental studies on diverse architectures and two real-world benchmark datasets show that our proposed algorithm significantly outperforms state-of-the-art approaches in both accuracy and data efficiency. The code and supplementary materials are available in https://anonymous.4open.science/r/Materials.
Brain-Computer Interfaces (BCIs) can decode imagined speech from neural activity. However, these systems typically require extensive training sessions where participants imaginedly repeat words, leading to mental fatigue and difficulties identifying the onset of words, especially when imagining sequences of words. This paper addresses these challenges by transferring a classifier trained in overt speech data to covert speech classification. We used electroencephalogram (EEG) features derived from the Hilbert envelope and temporal fine structure, and used them to train a bidirectional long-short-term memory (BiLSTM) model for classification. Our method reduces the burden of extensive training and achieves state-of-the-art classification accuracy: 86.44% for overt speech and 79.82% for covert speech using the overt speech classifier.
Preference-based reinforcement learning (RL) offers a promising approach for aligning policies with human intent but is often constrained by the high cost of human feedback. In this work, we introduce PrefVLM, a framework that integrates Vision-Language Models (VLMs) with selective human feedback to significantly reduce annotation requirements while maintaining performance. Our method leverages VLMs to generate initial preference labels, which are then filtered to identify uncertain cases for targeted human annotation. Additionally, we adapt VLMs using a self-supervised inverse dynamics loss to improve alignment with evolving policies. Experiments on Meta-World manipulation tasks demonstrate that PrefVLM achieves comparable or superior success rates to state-of-the-art methods while using up to 2 x fewer human annotations. Furthermore, we show that adapted VLMs enable efficient knowledge transfer across tasks, further minimizing feedback needs. Our results highlight the potential of combining VLMs with selective human supervision to make preference-based RL more scalable and practical.
Multilingual language models have significantly advanced due to rapid progress in natural language processing. Models like BLOOM 1.7B, trained on diverse multilingual datasets, aim to bridge linguistic gaps. However, their effectiveness in capturing linguistic knowledge, particularly for low-resource languages, remains an open question. This study critically examines MLMs capabilities in multilingual understanding, semantic representation, and cross-lingual knowledge transfer. While these models perform well for high-resource languages, they struggle with less-represented ones. Additionally, traditional evaluation methods often overlook their internal syntactic and semantic encoding. This research addresses key limitations through three objectives. First, it assesses semantic similarity by analyzing multilingual word embeddings for consistency using cosine similarity. Second, it examines BLOOM-1.7B and Qwen2 through Named Entity Recognition and sentence similarity tasks to understand their linguistic structures. Third, it explores cross-lingual knowledge transfer by evaluating generalization from high-resource to low-resource languages in sentiment analysis and text classification. By leveraging linguistic probing, performance metrics, and visualizations, this study provides insights into the strengths and limitations of MLMs. The findings aim to enhance multilingual NLP models, ensuring better support for both high- and low-resource languages, thereby promoting inclusivity in language technologies.
Universal time series representation learning is challenging but valuable in real-world applications such as classification, anomaly detection, and forecasting. Recently, contrastive learning (CL) has been actively explored to tackle time series representation. However, a key challenge is that the data augmentation process in CL can distort seasonal patterns or temporal dependencies, inevitably leading to a loss of semantic information. To address this challenge, we propose Topological Contrastive Learning for time series (TopoCL). TopoCL mitigates such information loss by incorporating persistent homology, which captures the topological characteristics of data that remain invariant under transformations. In this paper, we treat the temporal and topological properties of time series data as distinct modalities. Specifically, we compute persistent homology to construct topological features of time series data, representing them in persistence diagrams. We then design a neural network to encode these persistent diagrams. Our approach jointly optimizes CL within the time modality and time-topology correspondence, promoting a comprehensive understanding of both temporal semantics and topological properties of time series. We conduct extensive experiments on four downstream tasks-classification, anomaly detection, forecasting, and transfer learning. The results demonstrate that TopoCL achieves state-of-the-art performance.
Deep learning techniques are increasingly being adopted in diagnostic medical imaging. However, the limited availability of high-quality, large-scale medical datasets presents a significant challenge, often necessitating the use of transfer learning approaches. This study investigates self-supervised learning methods for pre-training capsule networks in polyp diagnostics for colon cancer. We used the PICCOLO dataset, comprising 3,433 samples, which exemplifies typical challenges in medical datasets: small size, class imbalance, and distribution shifts between data splits. Capsule networks offer inherent interpretability due to their architecture and inter-layer information routing mechanism. However, their limited native implementation in mainstream deep learning frameworks and the lack of pre-trained versions pose a significant challenge. This is particularly true if aiming to train them on small medical datasets, where leveraging pre-trained weights as initial parameters would be beneficial. We explored two auxiliary self-supervised learning tasks, colourisation and contrastive learning, for capsule network pre-training. We compared self-supervised pre-trained models against alternative initialisation strategies. Our findings suggest that contrastive learning and in-painting techniques are suitable auxiliary tasks for self-supervised learning in the medical domain. These techniques helped guide the model to capture important visual features that are beneficial for the downstream task of polyp classification, increasing its accuracy by 5.26% compared to other weight initialisation methods.
Human Activity Recognition (HAR) such as fall detection has become increasingly critical due to the aging population, necessitating effective monitoring systems to prevent serious injuries and fatalities associated with falls. This study focuses on fine-tuning the Vision Transformer (ViT) model specifically for HAR using radar-based Time-Doppler signatures. Unlike traditional image datasets, these signals present unique challenges due to their non-visual nature and the high degree of similarity among various activities. Directly fine-tuning the ViT with all parameters proves suboptimal for this application. To address this challenge, we propose a novel approach that employs Low-Rank Adaptation (LoRA) fine-tuning in the weight space to facilitate knowledge transfer from pre-trained ViT models. Additionally, to extract fine-grained features, we enhance feature representation through the integration of a serial-parallel adapter in the feature space. Our innovative joint fine-tuning method, tailored for radar-based Time-Doppler signatures, significantly improves HAR accuracy, surpassing existing state-of-the-art methodologies in this domain. Our code is released at https://github.com/wangyijunlyy/SelaFD.
Fingerprint recognition remains one of the most reliable biometric technologies due to its high accuracy and uniqueness. Traditional systems rely on contact-based scanners, which are prone to issues such as image degradation from surface contamination and inconsistent user interaction. To address these limitations, contactless fingerprint recognition has emerged as a promising alternative, providing non-intrusive and hygienic authentication. This study evaluates the impact of image enhancement tech-niques on the performance of pre-trained deep learning models using transfer learning for touchless fingerprint recognition. The IIT-Bombay Touchless and Touch-Based Fingerprint Database, containing data from 200 subjects, was employed to test the per-formance of deep learning architectures such as VGG-16, VGG-19, Inception-V3, and ResNet-50. Experimental results reveal that transfer learning methods with fingerprint image enhance-ment (indirect method) significantly outperform those without enhancement (direct method). Specifically, VGG-16 achieved an accuracy of 98% in training and 93% in testing when using the enhanced images, demonstrating superior performance compared to the direct method. This paper provides a detailed comparison of the effectiveness of image enhancement in improving the accuracy of transfer learning models for touchless fingerprint recognition, offering key insights for developing more efficient biometric systems.