Recent advancements in machine learning have ushered in innovative techniques for augmenting datasets, particularly through contrastive learning in the computer vision domain. This study pioneers the application of contrastive learning for sentiment analysis, introducing a novel approach termed EmoConLearn. By fine-tuning contrastive learning embeddings, EmoConLearn significantly surpasses BERT-based embeddings in sentiment analysis accuracy, as evidenced by our evaluations on the DynaSent dataset. This research further delves into the efficacy of EmoConLearn across various domain-specific datasets, highlighting its versatility. Additionally, we investigate upsampling strategies to mitigate class imbalance, further enhancing EmoConLearn's performance in sentiment analysis benchmarks.

Recently, generative adversarial networks (GANs) can be used to generate the photo-realistic image from a low-dimension random noise. It is very dangerous that the synthesized or generated image is used on inappropriate contents in social media network. In order to successfully detect such fake image, an effective and efficient image forgery detector is desired. However, conventional image forgery detectors are failed to recognize the synthesized or generated images by using GAN-based generator since they are all generated but manipulation from the source. Therefore, we propose a deep learning-based approach to detect the fake image by combining the contrastive loss. First, several state-of-the-art GANs will be collected to generate the fake-real image pairs. Then, the contrastive will be used on the proposed common fake feature network (CFFN)to learn the discriminative feature between the fake image and real image (i.e., paired information). Finally, a smaller network will be concatenated to the CFFN to determine whether the feature of the input image is fake or real. Experimental results demonstrated that the proposed method significantly outperforms other state-of-the-art fake image detectors.

Attack investigation is an important research field in forensics analysis. Many existing supervised attack investigation methods rely on well-labeled data for effective training. While the unsupervised approach based on BERT can mitigate the issues, the high degree of similarity between certain real-world attack and normal behaviors makes it challenging to accurately identify disguised attacks. This paper proposes ConLBS, an attack investigation approach that combines the contrastive learning framework and multi-layer Transformer network to realize the classification of behavior sequences. Specifically, ConLBS constructs behavior sequences describing behavior patterns from audit logs, and a novel lemmatization strategy is proposed to map the semantics to the attack pattern layer. Four different augmentation strategies are explored to enhance the differentiation between attack and normal behavior sequences. Moreover, ConLBS can perform unsupervised representation learning on unlabeled sequences, and can be trained either supervised or unsupervised depending on the availability of labeled data. The performance of ConLBS is evaluated in two public datasets. The results show that ConLBS can effectively identify at-tack behavior sequences in the cases of unlabeled data or less labeled data to realize attack investigation, and achieve the superior effectiveness compared to existing methods and models.

This paper proposes a Grouped Contrastive Learning of self-supervised Sentence Representation (GCLSR), which can learn an effective and meaningful representation of sentences. Previous works maximize the similarity between two vectors to be the objective of contrastive learning, suffering from the high-dimensionality of the vectors. In addition, most previous works have adopted discrete data augmentation to obtain positive samples and directly employed contrastive framework of computer vision to perform contrastive training, which could hamper contrastive training because text data is discrete and sparse compared with image data. To address those issues, we propose a grouped contrastive learning framework, i.e., GCLSR, which divides the high-dimensional feature vector into several groups and respectively computes the groups’ contrastive losses to make use of more local information, eventually obtaining a more fine-grained sentence representation. In addition, in GCLSR, we design a new self-attention mechanism and a continuous as well as partial word vector augmentation (PWVA). For the discrete and sparse text data, the usage of self-attention could help model focus the informative words by measuring the importance of every word in a sentence. By using the PWVA, GCLSR can obtain high-quality positive samples used for contrastive learning. Experimental results demonstrate that our proposed GCLSR achieves an encouraging result on the challenging datasets of the standard semantic textual similarity (STS) task and transfer task.

Despite significant advances in tumor biology and clinical therapeutics, metastasis remains the primary cause of cancer-related deaths. While RNA-seq technology has been used extensively to study metastatic cancer characteristics, challenges persist in acquiring adequate transcriptomic data. To overcome this challenge, we propose MetGen, a generative contrastive learning based on deep learning model. MetGen generates synthetic metastatic cancer expression profiles using primary cancer and normal tissue expression data. Our results demonstrate that MetGen generates comparable samples to actual metastatic cancer samples, and we discuss the learning mechanism of MetGen. Additionally, we demonstrate MetGen's interpretability using metastatic prostate cancer and metastatic breast cancer. MetGen has learned highly relevant signatures in cancer, tissue, and tumor microenvironment, such as immune response and the metastasis process, which potentially fosters a more comprehensive understanding of metastatic cancer biology. The development of MetGen represents a significant step toward the study of metastatic cancer biology by providing a generative model that identifies candidate therapeutic targets for the treatment of metastatic cancer.

This study introduces SkinLiTE, a lightweight supervised contrastive learning model, tailored to enhance the detection and typification of skin lesions in dermoscopic images. The core of SkinLiTE lies in its unique integration of supervised and contrastive learning approaches, which leverages labeled data to learn generalizable representations. This approach is particularly adept at handling the complexities and imbalances inherent in skin lesions datasets. The methodology encompasses a two-phase learning process. In the first phase, SkinLiTE utilizes an encoder network and a projection head to transform and project dermoscopic images into a feature space where contrastive loss is applied, focusing on minimizing intra-class variations while maximizing inter-class differences. The second phase freezes the encoder's weights, leveraging the learned representations for classification through a series of dense and dropout layers. The model was evaluated using three datasets from Skin Cancer ISIC 2019-2020, covering a wide range of skin conditions. SkinLiTE demonstrated superior performance across various metrics—accuracy, AUC , and F1 score—particularly when compared with traditional supervised learning models. Notably, SkinLiTE achieved an accuracy of 0.9087 using AugMix augmentation for binary classification of skin lesions. It also showed comparable results with the state-of-the-art approaches of ISIC challenge without relying on external data, underscoring its efficacy and efficiency. The results highlight the potential of SkinLiTE as a significant step forward in the field of dermatological AI, offering a robust, efficient, and accurate tool for skin lesion detection and classification. Its lightweight architecture and ability to handle imbalanced datasets make it particularly suited for integration into Internet of Medical Things environments, paving the way for enhanced remote patient monitoring and diagnostic capabilities. This research contributes to the evolving landscape of AI in healthcare, demonstrating the impact of innovative learning methodologies in medical image analysis.

Recently, there has been considerable research on deepfake detection. However, most existing methods face challenges in adapting to the advancements in new generative models within unknown domains. In addition, the emergence of new generative models capable of producing and editing high-quality images, such as Diffusion, Consistency, and LCM, poses a challenge for traditional deepfake training models. These advancements highlight the need for adapting and evolving existing deepfake detection techniques to effectively counter the threats posed by sophisticated image manipulation technologies. In this paper, our objective is to detect deepfake videos in unknown domains using unlabeled data. Specifically, our proposed approach employs Meta Pseudo Labels (MPL) with Supervised Contrastive Learning, so-called SupCon-MPL, allowing the model to be trained on unlabeled images. MPL involves the simultaneous training of both a Teacher model and a Student model, where the Teacher model generates Pseudo Labels utilized to train the Student model. This method aims to enhance the adaptability and robustness of deepfake detection systems against emerging unknown domains. Supervised Contrastive Learning utilizes labels to compare samples within similar classes more intensively, while encouraging greater distinction from samples in dissimilar classes. This facilitates the learning of features in a diverse set of deepfake images by the model, consequently contributing to the performance of deepfake detection in unknown domains. When utilizing the ResNet50 model as the backbone, SupCon-MPL exhibited an improvement of 1.58% in accuracy compared to traditional MPL in known domain detection. Moreover, in the same generation of unknown domain detection, there was a 1.32% accuracy enhancement, while in the detection of post-generation unknown domains, there was an 8.74% increase in accuracy.

In this paper, we present the idea of Self Supervised learning on the Shape Completion and Classification of point clouds. Most 3D shape completion pipelines utilize autoencoders to extract features from point clouds used in downstream tasks such as Classification, Segmentation, Detection, and other related applications. Our idea is to add Contrastive Learning into Auto-Encoders to learn both global and local feature representations of point clouds. We use a combination of Triplet Loss and Chamfer distance to learn global and local feature representations. To evaluate the performance of embeddings for Classification, we utilize the PointNet classifier. We also extend the number of classes to evaluate our model from 4 to 10 to show the generalization ability of learned features. Based on our results, embedding generated from the Contrastive autoencoder enhances Shape Completion and Classification performance from 84.2% to 84.9% of point clouds achieving the state-of-the-art results with 10 classes.

Human Activity Recognition (HAR) is a process to automatically detect human activities based on stream data generated from various sensors, including inertial sensors, physiological sensors, location sensors, cameras, time, and many others. Unsupervised contrastive learning has been excellent, while the contrastive loss mechanism is less studied. In this paper, we provide a temperature (τ) variance study affecting the loss of SimCLR model and ultimately full HAR evaluation results. We focus on understanding the implications of unsupervised contrastive loss in context of HAR data. In this work, also regulation of the temperature(τ) coefficient is incorporated for improving the HAR feature qualities and overall performance for downstream tasks in healthcare setting. Performance boost of 1.3% is observed in experimentation.

Generative models have recently gained popularity in Remote Sensing, offering substantial benefits for interpreting and utilizing satellite imagery across diverse applications such as climate monitoring, urban planning, and wildfire detection. These models are particularly adept at addressing the challenges posed by satellite images, which often exhibit domain variability due to seasonal changes, sensor characteristics, and especially variations in spectral bands. Such variability can significantly impact model performance across various tasks. In response to these challenges, our work introduces a novel approach that harnesses the capabilities of Generative Adversarial Networks (GANs), augmented with Contrastive Learning, to generate target domain images that account for multi-spectral band variations effectively. By maximizing mutual information between corresponding patches and leveraging the power of GANs, our model aims to generate realistic-looking images across different multi-spectral domains. We present a comparative analysis of our model against other well-established generative models, demonstrating its efficacy in generating high-quality satellite images while effectively managing domain variations inherent to multi-spectral diversity.

Abstract: The main contribution of this paper is to introduce a new iterative training algorithm for restricted Boltzmann machines. The proposed learning path is inspired from online sequential extreme learning machine one of extreme learning machine variants which deals with time accumulated sequences of data with fixed or varied sizes. Recursive least squares rules are integrated for weights adaptation to avoid learning rate tuning and local minimum issues. The proposed approach is compared to one of the well known training algorithms for Boltzmann machines named “contrastive divergence”, in term of time, accuracy and algorithmic complexity under the same conditions. Results strongly encourage the new given rules during data reconstruction.

Decision support systems based on machine learning models should be able to help users identify opportunities and threats. Popular model-agnostic explanation models can identify factors that support various predictions, answering questions like “What factors affect sales?” or “Why did sales decline?”, but do not highlight what a person should or could do to get a more desirable outcome. Counterfactual explanation approaches address intervention, and some even consider feasibility, but there has been no evaluation of their suitability for real-time applications, such as question answering. Here we address this gap by introducing a novel model-agnostic method that provides specific, feasible changes that would impact the outcomes of a complex black-box AI model for a given instance and assess its real-world utility by measuring its real-time performance and ability to find achievable changes. The method uses the instance of concern to generate high-precision explanations and then applies a secondary method to find minimally contrastive and maximally probable high-precision counterfactual explanations, while limiting contrasted features to changes that are achievable. We demonstrate that using this method, it is possible to find such explanations quickly enough for use in real-time systems. High-precision achievable minimally contrastive explanations would be useful in applications where people seek remedial actions or question how effective a proposed remedy is likely to be.

To solve the problem that the common long-tailed classification method does not use the semantic features of the original label text of the image, and the difference between the classification accuracy of most classes and minority classes is large, the long-tailed image classification method based on enhanced contrast visual language trains the head class and tail class samples separately, uses text image to pre-train the information, and uses enhanced momentum contrast loss function and RandAugment enhancement to improve the learning of tail class samples. On the ImageNet-LT long-tailed dataset, the enhanced contrastive visual-language based long-tailed image classification method has improved all class accuracy, tail class accuracy, middle class accuracy, and F1 values by 3.4%, 7.6%, 3.5%, and 11.2%, respectively, compared to the BALLAD method. The difference in accuracy between the head class and tail class is reduced by 1.6% compared to the BALLAD method. The results of three comparative experiments indicate that the long-tailed image classification method based on enhanced contrastive visual-language has improved the performance of tail classes and reduced the accuracy difference between majority and minority classes.

An appearance model plays a crucial rule in multi-target tracking. In traditional approaches, the two steps of appearance modeling i.e visual representation and statistically similarity measure are modeled separately. Visual representation is achieved either through hand-crafted features or deep features and statically similarity is measure through a cross entropy loss function. A loss function based on cross-entropy (KL-divergence, mutual information) find closely related probability distribution for the targets. However, if the targets have similar visual representation, it ends up mixing the targets. To tackle this problem, we come up with a synergetic appearance model named Single Shot Appearance Model based on Siamese neural network. The network is trained with a contrastive loss function for finding the similarity between different targets in a single shot. The input to the network is two target patches and based on their similarity, a contrastive score is output by the network. The proposed model is evaluated on accumulative dissimilarity metric on three datasets. Quantitatively, promising results are achieved against three baseline methods.

A common challenge in real-world classification scenarios with sequentially appending target domain data is insufficient training datasets during the training phase. Therefore, conventional deep learning and transfer learning classifiers are not applicable especially when individual classes are not represented or are severely underrepresented at the outset. Domain Generalization approaches reach their limits when domain shifts become too large, making them occasionally unsuitable as well. In many (technical) domains, however, it is only the defect/ worn/ reject classes that are insufficiently represented, while the non-defect class is often available from the beginning. The proposed classification approach addresses such conditions and is based on a CNN encoder. Following a contrastive learning approach, it is trained with a modified triplet loss function using two datasets: Besides the non-defective target domain class (= 1st dataset), a state-of-the-art labeled source domain dataset that contains highly related classes (e.g., a related manufacturing error or wear defect) but originates from a (highly) different domain (e.g., different product, material, or appearance) (= 2nd dataset) is utilized. The approach learns the classification features from the source domain dataset while at the same time learning the differences between the source and the target domain in a single training step, aiming to transfer the relevant features to the target domain. The classifier becomes sensitive to the classification features and – by architecture – robust against the highly domain-specific context. The approach is benchmarked in a technical and a non-technical domain and shows convincing classification results. In particular, it is shown that the domain generalization capabilities and classification results are improved by the proposed architecture, allowing for larger domain shifts between source and target domains.

Domain generalization has led to remarkable achievements in Hyperspectral Image (HSI) classification. Inspired by contrastive language-image pre-training (CLIP), the language-aware domain generalization method has been explored to learn cross-domain-invariant representation. However, existing methods face some challenges: 1) The weak capacity to extract long-range contextual information and inter-class correlation. 2) Due to the inadequacies of the large-scale pre-training for HSI data, the spatial-spectral features of HSI and linguistic features can not be straightforwardly alignment. To address the above problems, a novel network has been proposed with a CLIP framework, which consists of an image encoder, based on an encoder-only transformer to obtain the global contextual information and inter-class correlation, a frozen text encoder, and a cross-attention mechanism, named Linguistic-Interact-with-Visual Engager (LIVE), enhances the interaction between two modalities. Extensive experiments demonstrating superior performance over state-of-the-art methods in HSI Domain Generalization with a CLIP framework.

Self-supervised learning is a method that learns general representation from unlabeled data. Masked image modeling (MIM), one of the generative self-supervised learning methods, has drawn attention showing state-of-the-art performance on various downstream tasks, though showing poor linear separability resulting from the token-level approach. In this paper, we propose a contrastive learning-based multi-view masked autoencoder for MIM, exploiting an image-level approach by learning common features from two different augmented views. We strengthen MIM by learning long-range global patterns from contrastive loss. Our framework adopts simple encoder-decoder architecture, learning rich and general representation by following a simple process: 1) two different views are generated from an input image with random masking and by contrastive loss, we can learn semantic distance of the representations generated by an encoder. By applying a high mask ratio, 80%, it works as strong augmentation and alleviates the representation collapse problem. 2) With reconstruction loss, decoder learns to reconstruct an original image from the masked image. We assess our framework by several experiments on benchmark datasets of image classification, object detection, and semantic segmentation. We achieve 84.3% fine-tuning accuracy on ImageNet-1K classification and 76.7% in linear probing, exceeding previous studies and show promising results on other downstream tasks. Experimental results demonstrate that our work can learn rich and general image representation by applying contrastive loss to masked image modeling.

In recent years, there has been a growing interest in remote sensing image-text cross-modal retrieval due to the rapid development of space information technology and the significant increase in remote sensing image data volume. One approach that has shown promising results in cross-modal retrieval of natural images is the multimodal fusion encoding method. However, remote sensing images have unique characteristics that make the retrieval task challenging. Firstly, the semantic features of remote sensing images are fine-grained, meaning they can be divided into multiple basic units of semantic expression. Additionally, these images exhibit variations in resolution, color, and perspective. Different combinations of basic units of semantic expression can generate diverse text descriptions. These characteristics pose considerable challenges for cross-modal retrieval. To address these challenges, this paper proposes a multi-task guided fusion encoder (MTGFE) based on the multimodal fusion encoding method. The model incorporates three tasks: image-text matching (ITM), masked language modeling (MLM), and the newly introduced multi-view joint representations contrast (MVJRC) task. By jointly training the model with these tasks, we aim to enhance its capability to capture fine-grained correlations between remote sensing images and texts. Specifically, the MVJRC task is designed to improve the model’s consistency in feature expression and fine-grained correlation, particularly for remote sensing images with significant differences in resolution, color, and angle. Furthermore, to address the computational complexity associated with large-scale fusion models and improve retrieval efficiency, this paper proposes a retrieval filtering method. This method achieves higher retrieval efficiency while minimizing accuracy loss. Extensive experiments were conducted on four public datasets to evaluate the proposed method, and the results validate its effectiveness. Overall, this study focuses on remote sensing image-text cross-modal retrieval and introduces the MTGFE model, which combines multimodal fusion encoding with multiple tasks to enhance the model’s ability to capture fine-grained correlations. Additionally, a retrieval filtering method is proposed to improve retrieval efficiency. Experimental results demonstrate the effectiveness of the proposed method.