Self-supervised pretraining has been observed to be effective at improving feature representations for transfer learning, leveraging large amounts of unlabelled data. This review summarizes recent research into its usage in X-ray, computed tomography, magnetic resonance, and ultrasound imaging, concentrating on studies that compare self-supervised pretraining to fully supervised learning for diagnostic tasks such as classification and segmentation. The most pertinent finding is that self-supervised pretraining generally improves downstream task performance compared to full supervision, most prominently when unlabelled examples greatly outnumber labelled examples. Based on the aggregate evidence, recommendations are provided for practitioners considering using self-supervised learning. Motivated by limitations identified in current research, directions and practices for future study are suggested, such as integrating clinical knowledge with theoretically justified self-supervised learning methods, evaluating on public datasets, growing the modest body of evidence for ultrasound, and characterizing the impact of self-supervised pretraining on generalization.
Estimating the transferability of publicly available pretrained models to a target task has assumed an important place for transfer learning tasks in recent years. Existing efforts propose metrics that allow a user to choose one model from a pool of pre-trained models without having to fine-tune each model individually and identify one explicitly. With the growth in the number of available pre-trained models and the popularity of model ensembles, it also becomes essential to study the transferability of multiple-source models for a given target task. The few existing efforts study transferability in such multi-source ensemble settings using just the outputs of the classification layer and neglect possible domain or task mismatch. Moreover, they overlook the most important factor while selecting the source models, viz., the cohesiveness factor between them, which can impact the performance and confidence in the prediction of the ensemble. To address these gaps, we propose a novel Optimal tranSport-based suBmOdular tRaNsferability metric (OSBORN) to estimate the transferability of an ensemble of models to a downstream task. OSBORN collectively accounts for image domain difference, task difference, and cohesiveness of models in the ensemble to provide reliable estimates of transferability. We gauge the performance of OSBORN on both image classification and semantic segmentation tasks. Our setup includes 28 source datasets, 11 target datasets, 5 model architectures, and 2 pre-training methods. We benchmark our method against current state-of-the-art metrics MS-LEEP and E-LEEP, and outperform them consistently using the proposed approach.
Graph contrastive learning (GCL) has recently emerged as a promising approach for graph representation learning. Some existing methods adopt the 1-vs-K scheme to construct one positive and K negative samples for each graph, but it is difficult to set K. For those methods that do not use negative samples, it is often necessary to add additional strategies to avoid model collapse, which could only alleviate the problem to some extent. All these drawbacks will undoubtedly have an adverse impact on the generalizability and efficiency of the model. In this paper, to address these issues, we propose a novel graph self-contrast framework GraphSC, which only uses one positive and one negative sample, and chooses triplet loss as the objective. Specifically, self-contrast has two implications. First, GraphSC generates both positive and negative views of a graph sample from the graph itself via graph augmentation functions of various intensities, and use them for self-contrast. Second, GraphSC uses Hilbert-Schmidt Independence Criterion (HSIC) to factorize the representations into multiple factors and proposes a masked self-contrast mechanism to better separate positive and negative samples. Further, Since the triplet loss only optimizes the relative distance between the anchor and its positive/negative samples, it is difficult to ensure the absolute distance between the anchor and positive sample. Therefore, we explicitly reduced the absolute distance between the anchor and positive sample to accelerate convergence. Finally, we conduct extensive experiments to evaluate the performance of GraphSC against 19 other state-of-the-art methods in both unsupervised and transfer learning settings.
This paper proposes a novel semi-supervised TTS framework, QS-TTS, to improve TTS quality with lower supervised data requirements via Vector-Quantized Self-Supervised Speech Representation Learning (VQ-S3RL) utilizing more unlabeled speech audio. This framework comprises two VQ-S3R learners: first, the principal learner aims to provide a generative Multi-Stage Multi-Codebook (MSMC) VQ-S3R via the MSMC-VQ-GAN combined with the contrastive S3RL, while decoding it back to the high-quality audio; then, the associate learner further abstracts the MSMC representation into a highly-compact VQ representation through a VQ-VAE. These two generative VQ-S3R learners provide profitable speech representations and pre-trained models for TTS, significantly improving synthesis quality with the lower requirement for supervised data. QS-TTS is evaluated comprehensively under various scenarios via subjective and objective tests in experiments. The results powerfully demonstrate the superior performance of QS-TTS, winning the highest MOS over supervised or semi-supervised baseline TTS approaches, especially in low-resource scenarios. Moreover, comparing various speech representations and transfer learning methods in TTS further validates the notable improvement of the proposed VQ-S3RL to TTS, showing the best audio quality and intelligibility metrics. The trend of slower decay in the synthesis quality of QS-TTS with decreasing supervised data further highlights its lower requirements for supervised data, indicating its great potential in low-resource scenarios.
The assessment of the well-being of the peripheral auditory nerve system in individuals experiencing hearing impairment is conducted through auditory brainstem response (ABR) testing. Audiologists assess and document the results of the ABR test. They interpret the findings and assign labels to them using reference-based markers like peak latency, waveform morphology, amplitude, and other relevant factors. Inaccurate assessment of ABR tests may lead to incorrect judgments regarding the integrity of the auditory nerve system; therefore, proper Hearing Loss (HL) diagnosis and analysis are essential. To identify and assess ABR automation while decreasing the possibility of human error, machine learning methods, notably deep learning, may be an appropriate option. To address these issues, this study proposed deep-learning models using the transfer-learning (TL) approach to extract features from ABR testing and diagnose HL using support vector machines (SVM). Pre-trained convolutional neural network (CNN) architectures like AlexNet, DenseNet, GoogleNet, InceptionResNetV2, InceptionV3, MobileNetV2, NASNetMobile, ResNet18, ResNet50, ResNet101, ShuffleNet, and SqueezeNet are used to extract features from the collected ABR reported images dataset in the proposed model. It has been decided to use six measures accuracy, precision, recall, geometric mean (GM), standard deviation (SD), and area under the ROC curve to measure the effectiveness of the proposed model. According to experimental findings, the ShuffleNet and ResNet50 models' TL is effective for ABR to diagnose HL using an SVM classifier, with a high accuracy rate of 95% when using the 5-fold cross-validation method.
This paper develops a novel method to estimate a latent factor model for a large target panel with missing observations by optimally using the information from auxiliary panel data sets. We refer to our estimator as target-PCA. Transfer learning from auxiliary panel data allows us to deal with a large fraction of missing observations and weak signals in the target panel. We show that our estimator is more efficient and can consistently estimate weak factors, which are not identifiable with conventional methods. We provide the asymptotic inferential theory for target-PCA under very general assumptions on the approximate factor model and missing patterns. In an empirical study of imputing data in a mixed-frequency macroeconomic panel, we demonstrate that target-PCA significantly outperforms all benchmark methods.
Detection of tumors in metastatic colorectal cancer (mCRC) plays an essential role in the early diagnosis and treatment of liver cancer. Deep learning models backboned by fully convolutional neural networks (FCNNs) have become the dominant model for segmenting 3D computerized tomography (CT) scans. However, since their convolution layers suffer from limited kernel size, they are not able to capture long-range dependencies and global context. To tackle this restriction, vision transformers have been introduced to solve FCNN's locality of receptive fields. Although transformers can capture long-range features, their segmentation performance decreases with various tumor sizes due to the model sensitivity to the input patch size. While finding an optimal patch size improves the performance of vision transformer-based models on segmentation tasks, it is a time-consuming and challenging procedure. This paper proposes a technique to select the vision transformer's optimal input multi-resolution image patch size based on the average volume size of metastasis lesions. We further validated our suggested framework using a transfer-learning technique, demonstrating that the highest Dice similarity coefficient (DSC) performance was obtained by pre-training on training data with a larger tumour volume using the suggested ideal patch size and then training with a smaller one. We experimentally evaluate this idea through pre-training our model on a multi-resolution public dataset. Our model showed consistent and improved results when applied to our private multi-resolution mCRC dataset with a smaller average tumor volume. This study lays the groundwork for optimizing semantic segmentation of small objects using vision transformers. The implementation source code is available at:https://github.com/Ramtin-Mojtahedi/OVTPS.
Warning: this paper contains content that may be offensive or upsetting. Most hate speech datasets neglect the cultural diversity within a single language, resulting in a critical shortcoming in hate speech detection. To address this, we introduce CREHate, a CRoss-cultural English Hate speech dataset. To construct CREHate, we follow a two-step procedure: 1) cultural post collection and 2) cross-cultural annotation. We sample posts from the SBIC dataset, which predominantly represents North America, and collect posts from four geographically diverse English-speaking countries (Australia, United Kingdom, Singapore, and South Africa) using culturally hateful keywords we retrieve from our survey. Annotations are collected from the four countries plus the United States to establish representative labels for each country. Our analysis highlights statistically significant disparities across countries in hate speech annotations. Only 56.2% of the posts in CREHate achieve consensus among all countries, with the highest pairwise label difference rate of 26%. Qualitative analysis shows that label disagreement occurs mostly due to different interpretations of sarcasm and the personal bias of annotators on divisive topics. Lastly, we evaluate large language models (LLMs) under a zero-shot setting and show that current LLMs tend to show higher accuracies on Anglosphere country labels in CREHate. Our dataset and codes are available at: https://github.com/nlee0212/CREHate
Document AI aims to automatically analyze documents by leveraging natural language processing and computer vision techniques. One of the major tasks of Document AI is document layout analysis, which structures document pages by interpreting the content and spatial relationships of layout, image, and text. This task can be image-centric, wherein the aim is to identify and label various regions such as authors and paragraphs, or text-centric, where the focus is on classifying individual words in a document. Although there are increasingly sophisticated methods for improving layout analysis, doubts remain about the extent to which their findings can be generalized to a broader context. Specifically, prior work developed systems based on very different architectures, such as transformer-based, graph-based, and CNNs. However, no work has mentioned the effectiveness of these models in a comparative analysis. Moreover, while language-independent Document AI models capable of knowledge transfer have been developed, it remains to be investigated to what degree they can effectively transfer knowledge. In this study, we aim to fill these gaps by conducting a comparative evaluation of state-of-the-art models in document layout analysis and investigating the potential of cross-lingual layout analysis by utilizing machine translation techniques.
Machine learning and deep learning models are potential vectors for various attack scenarios. For example, previous research has shown that malware can be hidden in deep learning models. Hiding information in a learning model can be viewed as a form of steganography. In this research, we consider the general question of the steganographic capacity of learning models. Specifically, for a wide range of models, we determine the number of low-order bits of the trained parameters that can be overwritten, without adversely affecting model performance. For each model considered, we graph the accuracy as a function of the number of low-order bits that have been overwritten, and for selected models, we also analyze the steganographic capacity of individual layers. The models that we test include the classic machine learning techniques of Linear Regression (LR) and Support Vector Machine (SVM); the popular general deep learning models of Multilayer Perceptron (MLP) and Convolutional Neural Network (CNN); the highly-successful Recurrent Neural Network (RNN) architecture of Long Short-Term Memory (LSTM); the pre-trained transfer learning-based models VGG16, DenseNet121, InceptionV3, and Xception; and, finally, an Auxiliary Classifier Generative Adversarial Network (ACGAN). In all cases, we find that a majority of the bits of each trained parameter can be overwritten before the accuracy degrades. Of the models tested, the steganographic capacity ranges from 7.04 KB for our LR experiments, to 44.74 MB for InceptionV3. We discuss the implications of our results and consider possible avenues for further research.
Transfer learning has become crucial in computer vision tasks due to the vast availability of pre-trained deep learning models. However, selecting the optimal pre-trained model from a diverse pool for a specific downstream task remains a challenge. Existing methods for measuring the transferability of pre-trained models rely on statistical correlations between encoded static features and task labels, but they overlook the impact of underlying representation dynamics during fine-tuning, leading to unreliable results, especially for self-supervised models. In this paper, we present an insightful physics-inspired approach named PED to address these challenges. We reframe the challenge of model selection through the lens of potential energy and directly model the interaction forces that influence fine-tuning dynamics. By capturing the motion of dynamic representations to decline the potential energy within a force-driven physical model, we can acquire an enhanced and more stable observation for estimating transferability. The experimental results on 10 downstream tasks and 12 self-supervised models demonstrate that our approach can seamlessly integrate into existing ranking techniques and enhance their performances, revealing its effectiveness for the model selection task and its potential for understanding the mechanism in transfer learning. Code will be available at https://github.com/lixiaotong97/PED.
In an age dominated by resource-intensive foundation models, the ability to efficiently adapt to downstream tasks is crucial. Visual Prompting (VP), drawing inspiration from the prompting techniques employed in Large Language Models (LLMs), has emerged as a pivotal method for transfer learning in the realm of computer vision. As the importance of efficiency continues to rise, research into model compression has become indispensable in alleviating the computational burdens associated with training and deploying over-parameterized neural networks. A primary objective in model compression is to develop sparse and/or quantized models capable of matching or even surpassing the performance of their over-parameterized, full-precision counterparts. Although previous studies have explored the effects of model compression on transfer learning, its impact on visual prompting-based transfer remains unclear. This study aims to bridge this gap, shedding light on the fact that model compression detrimentally impacts the performance of visual prompting-based transfer, particularly evident in scenarios with low data volume. Furthermore, our findings underscore the adverse influence of sparsity on the calibration of downstream visual-prompted models. However, intriguingly, we also illustrate that such negative effects on calibration are not present when models are compressed via quantization. This empirical investigation underscores the need for a nuanced understanding beyond mere accuracy in sparse and quantized settings, thereby paving the way for further exploration in Visual Prompting techniques tailored for sparse and quantized models.
Tissue phenotyping is a fundamental computational pathology (CPath) task in learning objective characterizations of histopathologic biomarkers in anatomic pathology. However, whole-slide imaging (WSI) poses a complex computer vision problem in which the large-scale image resolutions of WSIs and the enormous diversity of morphological phenotypes preclude large-scale data annotation. Current efforts have proposed using pretrained image encoders with either transfer learning from natural image datasets or self-supervised pretraining on publicly-available histopathology datasets, but have not been extensively developed and evaluated across diverse tissue types at scale. We introduce UNI, a general-purpose self-supervised model for pathology, pretrained using over 100 million tissue patches from over 100,000 diagnostic haematoxylin and eosin-stained WSIs across 20 major tissue types, and evaluated on 33 representative CPath clinical tasks in CPath of varying diagnostic difficulties. In addition to outperforming previous state-of-the-art models, we demonstrate new modeling capabilities in CPath such as resolution-agnostic tissue classification, slide classification using few-shot class prototypes, and disease subtyping generalization in classifying up to 108 cancer types in the OncoTree code classification system. UNI advances unsupervised representation learning at scale in CPath in terms of both pretraining data and downstream evaluation, enabling data-efficient AI models that can generalize and transfer to a gamut of diagnostically-challenging tasks and clinical workflows in anatomic pathology.
Human activity recognition (HAR) using machine learning has shown tremendous promise in detecting construction workers' activities. HAR has many applications in human-robot interaction research to enable robots' understanding of human counterparts' activities. However, many existing HAR approaches lack robustness, generalizability, and adaptability. This paper proposes a transfer learning methodology for activity recognition of construction workers that requires orders of magnitude less data and compute time for comparable or better classification accuracy. The developed algorithm transfers features from a model pre-trained by the original authors and fine-tunes them for the downstream task of activity recognition in construction. The model was pre-trained on Kinetics-400, a large-scale video-based human activity recognition dataset with 400 distinct classes. The model was fine-tuned and tested using videos captured from manual material handling (MMH) activities found on YouTube. Results indicate that the fine-tuned model can recognize distinct MMH tasks in a robust and adaptive manner which is crucial for the widespread deployment of collaborative robots in construction.
Battery state of health (SOH), which informs the maximal available capacity of the battery, is a key indicator of battery aging failure. Accurately estimating battery SOH is a vital function of the battery management system that remains to be addressed. In this study, a physics-informed Gaussian process regression (GPR) model is developed for battery SOH estimation, with the performance being systematically compared with that of different types of features and machine learning (ML) methods. The method performance is validated based on 58826 cycling data units of 118 cells. Experimental results show that the physics-driven ML generally estimates more accurate SOH than other non-physical features under different scenarios. The physical features-based GPR predicts battery SOH with the errors being less than 1.1% based on 10 to 20 mins' relaxation data. And the high robustness and generalization capability of the methodology is also validated against different ratios of training and test data under unseen conditions. Results also highlight the more effective capability of knowledge transfer between different types of batteries with the physical features and GPR. This study demonstrates the excellence of physical features in indicating the state evolution of complex systems, and the improved indication performance of these features by combining a suitable ML method.
Multi-task learning (MTL) has gained significant popularity in recommender systems as it enables simultaneous optimization of multiple objectives. A key challenge in MTL is negative transfer, but existing studies explored negative transfer on all samples, overlooking the inherent complexities within them. We split the samples according to the relative amount of positive feedback among tasks. Surprisingly, negative transfer still occurs in existing MTL methods on samples that receive comparable feedback across tasks. Existing work commonly employs a shared-embedding paradigm, limiting the ability of modeling diverse user preferences on different tasks. In this paper, we introduce a novel Shared and Task-specific EMbeddings (STEM) paradigm that aims to incorporate both shared and task-specific embeddings to effectively capture task-specific user preferences. Under this paradigm, we propose a simple model STEM-Net, which is equipped with an All Forward Task-specific Backward gating network to facilitate the learning of task-specific embeddings and direct knowledge transfer across tasks. Remarkably, STEM-Net demonstrates exceptional performance on comparable samples, achieving positive transfer. Comprehensive evaluation on three public MTL recommendation datasets demonstrates that STEM-Net outperforms state-of-the-art models by a substantial margin. Our code is released at https://github.com/LiangcaiSu/STEM.
Completing a sentence, phrase or word after typing few words / characters is very helpful for Intuit financial experts, while taking notes or having a live chat with users, since they need to write complex financial concepts more efficiently and accurately many times in a day. In this paper, we tie together different approaches like large language models, traditional Markov Models and char level models to create an end-to-end system to provide personalised sentence/word auto-complete suggestions to experts, under strict latency constraints. Proposed system can auto-complete sentences, phrases or words while writing with personalisation and can be trained with very less data and resources with good efficiency. Our proposed system is not only efficient and personalized but also robust as it leverages multiple machine learning techniques along with transfer learning approach to fine tune large language model with Intuit specific data. This ensures that even in cases of rare or unusual phrases, the system can provide relevant auto-complete suggestions in near real time. Survey has showed that this system saves expert note-taking time and boosts expert confidence in their communication with teammates and clients. Since enabling this predictive writing feature for QBLive experts, more than a million keystrokes have been saved based on these suggestions. We have done comparative study for our ensemble choice. Moreover this feature can be integrated with any product which has writing facility within a very short period of time.
Transfer learning improves the performance of deep learning models by initializing them with parameters pre-trained on larger datasets. Intuitively, transfer learning is more effective when pre-training is on the in-domain datasets. A recent study by NASA has demonstrated that the microstructure segmentation with encoder-decoder algorithms benefits more from CNN encoders pre-trained on microscopy images than from those pre-trained on natural images. However, CNN models only capture the local spatial relations in images. In recent years, attention networks such as Transformers are increasingly used in image analysis to capture the long-range relations between pixels. In this study, we compare the segmentation performance of Transformer and CNN models pre-trained on microscopy images with those pre-trained on natural images. Our result partially confirms the NASA study that the segmentation performance of out-of-distribution images (taken under different imaging and sample conditions) is significantly improved when pre-training on microscopy images. However, the performance gain for one-shot and few-shot learning is more modest with Transformers. We also find that for image segmentation, the combination of pre-trained Transformers and CNN encoders are consistently better than pre-trained CNN encoders alone. Our dataset (of about 50,000 images) combines the public portion of the NASA dataset with additional images we collected. Even with much less training data, our pre-trained models have significantly better performance for image segmentation. This result suggests that Transformers and CNN complement each other and when pre-trained on microscopy images, they are more beneficial to the downstream tasks.
Skeleton-based action segmentation requires recognizing composable actions in untrimmed videos. Current approaches decouple this problem by first extracting local visual features from skeleton sequences and then processing them by a temporal model to classify frame-wise actions. However, their performances remain limited as the visual features cannot sufficiently express composable actions. In this context, we propose Latent Action Composition (LAC), a novel self-supervised framework aiming at learning from synthesized composable motions for skeleton-based action segmentation. LAC is composed of a novel generation module towards synthesizing new sequences. Specifically, we design a linear latent space in the generator to represent primitive motion. New composed motions can be synthesized by simply performing arithmetic operations on latent representations of multiple input skeleton sequences. LAC leverages such synthesized sequences, which have large diversity and complexity, for learning visual representations of skeletons in both sequence and frame spaces via contrastive learning. The resulting visual encoder has a high expressive power and can be effectively transferred onto action segmentation tasks by end-to-end fine-tuning without the need for additional temporal models. We conduct a study focusing on transfer-learning and we show that representations learned from pre-trained LAC outperform the state-of-the-art by a large margin on TSU, Charades, PKU-MMD datasets.
LiDAR odometry estimation and 3D semantic segmentation are crucial for autonomous driving, which has achieved remarkable advances recently. However, these tasks are challenging due to the imbalance of points in different semantic categories for 3D semantic segmentation and the influence of dynamic objects for LiDAR odometry estimation, which increases the importance of using representative/salient landmarks as reference points for robust feature learning. To address these challenges, we propose a saliency-guided approach that leverages attention information to improve the performance of LiDAR odometry estimation and semantic segmentation models. Unlike in the image domain, only a few studies have addressed point cloud saliency information due to the lack of annotated training data. To alleviate this, we first present a universal framework to transfer saliency distribution knowledge from color images to point clouds, and use this to construct a pseudo-saliency dataset (i.e. FordSaliency) for point clouds. Then, we adopt point cloud-based backbones to learn saliency distribution from pseudo-saliency labels, which is followed by our proposed SalLiDAR module. SalLiDAR is a saliency-guided 3D semantic segmentation model that integrates saliency information to improve segmentation performance. Finally, we introduce SalLONet, a self-supervised saliency-guided LiDAR odometry network that uses the semantic and saliency predictions of SalLiDAR to achieve better odometry estimation. Our extensive experiments on benchmark datasets demonstrate that the proposed SalLiDAR and SalLONet models achieve state-of-the-art performance against existing methods, highlighting the effectiveness of image-to-LiDAR saliency knowledge transfer. Source code will be available at https://github.com/nevrez/SalLONet.