Ocular Toxoplasmosis (OT), is a common eye infection caused by T. gondii that can cause vision problems. Diagnosis is typically done through a clinical examination and imaging, but these methods can be complicated and costly, requiring trained personnel. To address this issue, we have created a benchmark study that evaluates the effectiveness of existing pre-trained networks using transfer learning techniques to detect OT from fundus images. Furthermore, we have also analysed the performance of transfer-learning based segmentation networks to segment lesions in the images. This research seeks to provide a guide for future researchers looking to utilise DL techniques and develop a cheap, automated, easy-to-use, and accurate diagnostic method. We have performed in-depth analysis of different feature extraction techniques in order to find the most optimal one for OT classification and segmentation of lesions. For classification tasks, we have evaluated pre-trained models such as VGG16, MobileNetV2, InceptionV3, ResNet50, and DenseNet121 models. Among them, MobileNetV2 outperformed all other models in terms of Accuracy (Acc), Recall, and F1 Score outperforming the second-best model, InceptionV3 by 0.7% higher Acc. However, DenseNet121 achieved the best result in terms of Precision, which was 0.1% higher than MobileNetv2. For the segmentation task, this work has exploited U-Net architecture. In order to utilize transfer learning the encoder block of the traditional U-Net was replaced by MobileNetV2, InceptionV3, ResNet34, and VGG16 to evaluate different architectures moreover two different two different loss functions (Dice loss and Jaccard loss) were exploited in order to find the most optimal one. The MobileNetV2/U-Net outperformed ResNet34 by 0.5% and 2.1% in terms of Acc and Dice Score, respectively when Jaccard loss function is employed during the training.
Time-Series Mining (TSM) is an important research area since it shows great potential in practical applications. Deep learning models that rely on massive labeled data have been utilized for TSM successfully. However, constructing a large-scale well-labeled dataset is difficult due to data annotation costs. Recently, pre-trained models have gradually attracted attention in the time series domain due to their remarkable performance in computer vision and natural language processing. In this survey, we provide a comprehensive review of Time-Series Pre-Trained Models (TS-PTMs), aiming to guide the understanding, applying, and studying TS-PTMs. Specifically, we first briefly introduce the typical deep learning models employed in TSM. Then, we give an overview of TS-PTMs according to the pre-training techniques. The main categories we explore include supervised, unsupervised, and self-supervised TS-PTMs. Further, extensive experiments involving 27 methods, 434 datasets, and 679 transfer learning scenarios are conducted to analyze the advantages and disadvantages of transfer learning strategies, Transformer-based models, and representative TS-PTMs. Finally, we point out some potential directions of TS-PTMs for future work.
Knowledge distillation has attracted a great deal of interest recently to compress pre-trained language models. However, existing knowledge distillation methods suffer from two limitations. First, the student model simply imitates the teacher's behavior while ignoring the underlying reasoning. Second, these methods usually focus on the transfer of sophisticated model-specific knowledge but overlook data-specific knowledge. In this paper, we present a novel attribution-driven knowledge distillation approach, which explores the token-level rationale behind the teacher model based on Integrated Gradients (IG) and transfers attribution knowledge to the student model. To enhance the knowledge transfer of model reasoning and generalization, we further explore multi-view attribution distillation on all potential decisions of the teacher. Comprehensive experiments are conducted with BERT on the GLUE benchmark. The experimental results demonstrate the superior performance of our approach to several state-of-the-art methods.
The field of audio captioning has seen significant advancements in recent years, driven by the availability of large-scale audio datasets and advancements in deep learning techniques. In this technical report, we present our approach to audio captioning, focusing on the use of a pretrained speech-to-text Whisper model and pretraining on synthetic captions. We discuss our training procedures and present our experiments' results, which include model size variations, dataset mixtures, and other hyperparameters. Our findings demonstrate the impact of different training strategies on the performance of the audio captioning model. Our code and trained models are publicly available on GitHub and Hugging Face Hub.
It has become a popular paradigm to transfer the knowledge of large-scale pre-trained models to various downstream tasks via fine-tuning the entire model parameters. However, with the growth of model scale and the rising number of downstream tasks, this paradigm inevitably meets the challenges in terms of computation consumption and memory footprint issues. Recently, Parameter-Efficient Fine-Tuning (PEFT) (e.g., Adapter, LoRA, BitFit) shows a promising paradigm to alleviate these concerns by updating only a portion of parameters. Despite these PEFTs having demonstrated satisfactory performance in natural language processing, it remains under-explored for the question of whether these techniques could be transferred to graph-based tasks with Graph Transformer Networks (GTNs). Therefore, in this paper, we fill this gap by providing extensive benchmarks with traditional PEFTs on a range of graph-based downstream tasks. Our empirical study shows that it is sub-optimal to directly transfer existing PEFTs to graph-based tasks due to the issue of feature distribution shift. To address this issue, we propose a novel structure-aware PEFT approach, named G-Adapter, which leverages graph convolution operation to introduce graph structure (e.g., graph adjacent matrix) as an inductive bias to guide the updating process. Besides, we propose Bregman proximal point optimization to further alleviate feature distribution shift by preventing the model from aggressive update. Extensive experiments demonstrate that G-Adapter obtains the state-of-the-art performance compared to the counterparts on nine graph benchmark datasets based on two pre-trained GTNs, and delivers tremendous memory footprint efficiency compared to the conventional paradigm.
This paper presents a generalized model for real-time detection of flying objects that can be used for transfer learning and further research, as well as a refined model that achieves state-of-the-art results for flying object detection. We achieve this by training our first (generalized) model on a data set containing 40 different classes of flying objects, forcing the model to extract abstract feature representations. We then perform transfer learning with these learned parameters on a data set more representative of real world environments (i.e. higher frequency of occlusion, very small spatial sizes, rotations, etc.) to generate our refined model. Object detection of flying objects remains challenging due to large variances of object spatial sizes/aspect ratios, rate of speed, occlusion, and clustered backgrounds. To address some of the presented challenges while simultaneously maximizing performance, we utilize the current state-of-the-art single-shot detector, YOLOv8, in an attempt to find the best trade-off between inference speed and mean average precision (mAP). While YOLOv8 is being regarded as the new state-of-the-art, an official paper has not been released as of yet. Thus, we provide an in-depth explanation of the new architecture and functionality that YOLOv8 has adapted. Our final generalized model achieves a mAP50 of 79.2%, mAP50-95 of 68.5%, and an average inference speed of 50 frames per second (fps) on 1080p videos. Our final refined model maintains this inference speed and achieves an improved mAP50 of 99.1% and mAP50-95 of 83.5%
A shared goal of several machine learning communities like continual learning, meta-learning and transfer learning, is to design algorithms and models that efficiently and robustly adapt to unseen tasks. An even more ambitious goal is to build models that never stop adapting, and that become increasingly more efficient through time by suitably transferring the accrued knowledge. Beyond the study of the actual learning algorithm and model architecture, there are several hurdles towards our quest to build such models, such as the choice of learning protocol, metric of success and data needed to validate research hypotheses. In this work, we introduce the Never-Ending VIsual-classification Stream (NEVIS'22), a benchmark consisting of a stream of over 100 visual classification tasks, sorted chronologically and extracted from papers sampled uniformly from computer vision proceedings spanning the last three decades. The resulting stream reflects what the research community thought was meaningful at any point in time, and it serves as an ideal test bed to assess how well models can adapt to new tasks, and do so better and more efficiently as time goes by. Despite being limited to classification, the resulting stream has a rich diversity of tasks from OCR, to texture analysis, scene recognition, and so forth. The diversity is also reflected in the wide range of dataset sizes, spanning over four orders of magnitude. Overall, NEVIS'22 poses an unprecedented challenge for current sequential learning approaches due to the scale and diversity of tasks, yet with a low entry barrier as it is limited to a single modality and well understood supervised learning problems. Moreover, we provide a reference implementation including strong baselines and an evaluation protocol to compare methods in terms of their trade-off between accuracy and compute.
Large foundation models, known for their strong zero-shot generalization, have excelled in visual and language applications. However, applying them to medical image segmentation, a domain with diverse imaging types and target labels, remains an open challenge. Current approaches, such as adapting interactive segmentation models like Segment Anything Model (SAM), require user prompts for each sample during inference. Alternatively, transfer learning methods like few/one-shot models demand labeled samples, leading to high costs. This paper introduces a new paradigm toward the universal medical image segmentation, termed 'One-Prompt Segmentation.' One-Prompt Segmentation combines the strengths of one-shot and interactive methods. In the inference stage, with just \textbf{one prompted sample}, it can adeptly handle the unseen task in a single forward pass. We train One-Prompt Model on 64 open-source medical datasets, accompanied by the collection of over 3,000 clinician-labeled prompts. Tested on 14 previously unseen datasets, the One-Prompt Model showcases superior zero-shot segmentation capabilities, outperforming a wide range of related methods. The code and data is released as https://github.com/KidsWithTokens/one-prompt.
Fine-grained classification is a challenging task that involves identifying subtle differences between objects within the same category. This task is particularly challenging in scenarios where data is scarce. Visual transformers (ViT) have recently emerged as a powerful tool for image classification, due to their ability to learn highly expressive representations of visual data using self-attention mechanisms. In this work, we explore Semi-ViT, a ViT model fine tuned using semi-supervised learning techniques, suitable for situations where we have lack of annotated data. This is particularly common in e-commerce, where images are readily available but labels are noisy, nonexistent, or expensive to obtain. Our results demonstrate that Semi-ViT outperforms traditional convolutional neural networks (CNN) and ViTs, even when fine-tuned with limited annotated data. These findings indicate that Semi-ViTs hold significant promise for applications that require precise and fine-grained classification of visual data.
End-to-end spoken language understanding (SLU) remains elusive even with current large pretrained language models on text and speech, especially in multilingual cases. Machine translation has been established as a powerful pretraining objective on text as it enables the model to capture high-level semantics of the input utterance and associations between different languages, which is desired for speech models that work on lower-level acoustic frames. Motivated particularly by the task of cross-lingual SLU, we demonstrate that the task of speech translation (ST) is a good means of pretraining speech models for end-to-end SLU on both intra- and cross-lingual scenarios. By introducing ST, our models reach higher performance over baselines on monolingual and multilingual intent classification as well as spoken question answering using SLURP, MINDS-14, and NMSQA benchmarks. To verify the effectiveness of our methods, we also create new benchmark datasets from both synthetic and real sources, for speech summarization and low-resource/zero-shot transfer from English to French or Spanish. We further show the value of preserving knowledge for the ST pretraining task for better downstream performance, possibly using Bayesian transfer regularizers.
It has been commonly observed that a teacher model with superior performance does not necessarily result in a stronger student, highlighting a discrepancy between current teacher training practices and effective knowledge transfer. In order to enhance the guidance of the teacher training process, we introduce the concept of distillation influence to determine the impact of distillation from each training sample on the student's generalization ability. In this paper, we propose Learning Good Teacher Matters (LGTM), an efficient training technique for incorporating distillation influence into the teacher's learning process. By prioritizing samples that are likely to enhance the student's generalization ability, our LGTM outperforms 10 common knowledge distillation baselines on 6 text classification tasks in the GLUE benchmark.
We present a Transfer Causal Learning (TCL) framework when target and source domains share the same covariate/feature spaces, aiming to improve causal effect estimation accuracy in limited data. Limited data is very common in medical applications, where some rare medical conditions, such as sepsis, are of interest. Our proposed method, named \texttt{$\ell_1$-TCL}, incorporates $\ell_1$ regularized TL for nuisance models (e.g., propensity score model); the TL estimator of the nuisance parameters is plugged into downstream average causal/treatment effect estimators (e.g., inverse probability weighted estimator). We establish non-asymptotic recovery guarantees for the \texttt{$\ell_1$-TCL} with generalized linear model (GLM) under the sparsity assumption in the high-dimensional setting, and demonstrate the empirical benefits of \texttt{$\ell_1$-TCL} through extensive numerical simulation for GLM and recent neural network nuisance models. Our method is subsequently extended to real data and generates meaningful insights consistent with medical literature, a case where all baseline methods fail.
In this work we introduce a self-supervised pretraining framework for transformers on functional Magnetic Resonance Imaging (fMRI) data. First, we pretrain our architecture on two self-supervised tasks simultaneously to teach the model a general understanding of the temporal and spatial dynamics of human auditory cortex during music listening. Our pretraining results are the first to suggest a synergistic effect of multitask training on fMRI data. Second, we finetune the pretrained models and train additional fresh models on a supervised fMRI classification task. We observe significantly improved accuracy on held-out runs with the finetuned models, which demonstrates the ability of our pretraining tasks to facilitate transfer learning. This work contributes to the growing body of literature on transformer architectures for pretraining and transfer learning with fMRI data, and serves as a proof of concept for our pretraining tasks and multitask pretraining on fMRI data.
We propose a fine-tuning algorithm for brain tumor segmentation that needs only a few data samples and helps networks not to forget the original tasks. Our approach is based on active learning and meta-learning. One of the difficulties in medical image segmentation is the lack of datasets with proper annotations, because it requires doctors to tag reliable annotation and there are many variants of a disease, such as glioma and brain metastasis, which are the different types of brain tumor and have different structural features in MR images. Therefore, it is impossible to produce the large-scale medical image datasets for all types of diseases. In this paper, we show a transfer learning method from high grade glioma to brain metastasis, and demonstrate that the proposed algorithm achieves balanced parameters for both glioma and brain metastasis domains within a few steps.
The field of clinical image analysis has been applying transfer learning models increasingly due to their less computational complexity, better accuracy etc. These are pre-trained models that don't require to be trained from scratch which eliminates the necessity of large datasets. Transfer learning models are mostly used for the analysis of brain, breast, or lung images but other sectors such as bone marrow cell detection or bone cancer detection can also benefit from using transfer learning models, especially considering the lack of available large datasets for these tasks. This paper studies the performance of several transfer learning models for osteosarcoma tumour detection. Osteosarcoma is a type of bone cancer mostly found in the cells of the long bones of the body. The dataset consists of H&E stained images divided into 4 categories- Viable Tumor, Non-viable Tumor, Non-Tumor and Viable Non-viable. Both datasets were randomly divided into train and test sets following an 80-20 ratio. 80% was used for training and 20\% for test. 4 models are considered for comparison- EfficientNetB7, InceptionResNetV2, NasNetLarge and ResNet50. All these models are pre-trained on ImageNet. According to the result, InceptionResNetV2 achieved the highest accuracy (93.29%), followed by NasNetLarge (90.91%), ResNet50 (89.83%) and EfficientNetB7 (62.77%). It also had the highest precision (0.8658) and recall (0.8658) values among the 4 models.
We propose a privacy-preserving ensemble infused enhanced Deep Neural Network (DNN) based learning framework in this paper for Internet-of-Things (IoT), edge, and cloud convergence in the context of healthcare. In the convergence, edge server is used for both storing IoT produced bioimage and hosting DNN algorithm for local model training. The cloud is used for ensembling local models. The DNN-based training process of a model with a local dataset suffers from low accuracy, which can be improved by the aforementioned convergence and Ensemble Learning. The ensemble learning allows multiple participants to outsource their local model for producing a generalized final model with high accuracy. Nevertheless, Ensemble Learning elevates the risk of leaking sensitive private data from the final model. The proposed framework presents a Differential Privacy-based privacy-preserving DNN with Transfer Learning for a local model generation to ensure minimal loss and higher efficiency at edge server. We conduct several experiments to evaluate the performance of our proposed framework.
Transformers, due to their ability to learn long range dependencies, have overcome the shortcomings of convolutional neural networks (CNNs) for global perspective learning. Therefore, they have gained the focus of researchers for several vision related tasks including medical diagnosis. However, their multi-head attention module only captures global level feature representations, which is insufficient for medical images. To address this issue, we propose a Channel Boosted Hybrid Vision Transformer (CB HVT) that uses transfer learning to generate boosted channels and employs both transformers and CNNs to analyse lymphocytes in histopathological images. The proposed CB HVT comprises five modules, including a channel generation module, channel exploitation module, channel merging module, region-aware module, and a detection and segmentation head, which work together to effectively identify lymphocytes. The channel generation module uses the idea of channel boosting through transfer learning to extract diverse channels from different auxiliary learners. In the CB HVT, these boosted channels are first concatenated and ranked using an attention mechanism in the channel exploitation module. A fusion block is then utilized in the channel merging module for a gradual and systematic merging of the diverse boosted channels to improve the network's learning representations. The CB HVT also employs a proposal network in its region aware module and a head to effectively identify objects, even in overlapping regions and with artifacts. We evaluated the proposed CB HVT on two publicly available datasets for lymphocyte assessment in histopathological images. The results show that CB HVT outperformed other state of the art detection models, and has good generalization ability, demonstrating its value as a tool for pathologists.
Blind image quality assessment is a challenging task particularly due to the unavailability of reference information. Training a deep neural network requires a large amount of training data which is not readily available for image quality. Transfer learning is usually opted to overcome this limitation and different deep architectures are used for this purpose as they learn features differently. After extensive experiments, we have designed a deep architecture containing two CNN architectures as its sub-units. Moreover, a self-collected image database BIQ2021 is proposed with 12,000 images having natural distortions. The self-collected database is subjectively scored and is used for model training and validation. It is demonstrated that synthetic distortion databases cannot provide generalization beyond the distortion types used in the database and they are not ideal candidates for general-purpose image quality assessment. Moreover, a large-scale database of 18.75 million images with synthetic distortions is used to pretrain the model and then retrain it on benchmark databases for evaluation. Experiments are conducted on six benchmark databases three of which are synthetic distortion databases (LIVE, CSIQ and TID2013) and three are natural distortion databases (LIVE Challenge Database, CID2013 and KonIQ-10 k). The proposed approach has provided a Pearson correlation coefficient of 0.8992, 0.8472 and 0.9452 subsequently and Spearman correlation coefficient of 0.8863, 0.8408 and 0.9421. Moreover, the performance is demonstrated using perceptually weighted rank correlation to indicate the perceptual superiority of the proposed approach. Multiple experiments are conducted to validate the generalization performance of the proposed model by training on different subsets of the databases and validating on the test subset of BIQ2021 database.
Collisions, crashes, and other incidents on road networks, if left unmitigated, can potentially cause cascading failures that can affect large parts of the system. Timely handling such extreme congestion scenarios is imperative to reduce emissions, enhance productivity, and improve the quality of urban living. In this work, we propose a Deep Reinforcement Learning (DRL) approach to reduce traffic congestion on multi-lane freeways during extreme congestion. The agent is trained to learn adaptive detouring strategies for congested freeway traffic such that the freeway lanes along with the local arterial network in proximity are utilized optimally, with rewards being congestion reduction and traffic speed improvement. The experimental setup is a 2.6-mile-long 4-lane freeway stretch in Shoreline, Washington, USA with two exits and associated arterial roads simulated on a microscopic and continuous multi-modal traffic simulator SUMO (Simulation of Urban MObility) while using parameterized traffic profiles generated using real-world traffic data. Our analysis indicates that DRL-based controllers can improve average traffic speed by 21\% when compared to no-action during steep congestion. The study further discusses the trade-offs involved in the choice of reward functions, the impact of human compliance on agent performance, and the feasibility of knowledge transfer from one agent to other to address data sparsity and scaling issues.
As the shortage of skilled workers continues to be a pressing issue, exacerbated by demographic change, it is becoming a critical challenge for organizations to preserve the knowledge of retiring experts and to pass it on to novices. While this knowledge transfer has traditionally taken place through personal interaction, it lacks scalability and requires significant resources and time. IT-based teaching systems have addressed this scalability issue, but their development is still tedious and time-consuming. In this work, we investigate the potential of machine learning (ML) models to facilitate knowledge transfer in an organizational context, leading to more cost-effective IT-based teaching systems. Through a systematic literature review, we examine key concepts, themes, and dimensions to better understand and design ML-based teaching systems. To do so, we capture and consolidate the capabilities of ML models in IT-based teaching systems, inductively analyze relevant concepts in this context, and determine their interrelationships. We present our findings in the form of a review of the key concepts, themes, and dimensions to understand and inform on ML-based teaching systems. Building on these results, our work contributes to research on computer-supported cooperative work by conceptualizing how ML-based teaching systems can preserve expert knowledge and facilitate its transfer from SMEs to human novices. In this way, we shed light on this emerging subfield of human-computer interaction and serve to build an interdisciplinary research agenda.