Small Amount Of Training Data Research Articles

KLASIFIKASI PENYAKIT PARU-PARU DENGAN MENGGUNAKAN METODE NAÏVE BAYES CLASSIFIER

The lungs are one of the organs of the human body that are very important in the process of respiration. There are several types of lung diseases, including Asthma, Bronchitis, Dyspnea, Pneumonia, COPD, and Tuberculosis. There are difficulties in the classification process, because the symptoms shown by sufferers have similarities between diseases. The purpose of this research is to classify lung disease using the Naive Bayes Classifier method. The choice of this method is because it only requires a small amount of training data to determine the estimated parameters needed in the classification process. This research was conducted at the Regional General Hospital Major General HA Thalib City of Sungai Full from August 3 2022 to September 3 2022. The data taken was in the form of medical records of lung disease patients from July to August as many as 134 patient data containing 19 symptoms disease and 6 disease diagnoses. From the test results using the Rapidminer application and data separation in the form of 34 testing data and 100 training data with a data comparison of 7:3, an accuracy value of 97.06 was obtained.

Implementation of the Naïve Bayes Algorithm for Divorce Prediction at the Tais Religious Court

Divorce is the breaking up of a husband and wife relationship, because there is no match between the two. Divorce is basically the release of the bond between husband and wife. In Tais district, based on the Tais Religious Court Annual Report from 2017, 2018 and 2019 it was found that there was an increase in the divorce rate. Based on the divorce rate which continues to increase every year, it is necessary to have data in the form of data to determine the factors that influence divorce. The use of Naive Bayes is that this method only requires a small amount of training data (Training Data) to determine the parameter estimates needed in the prediction process. In this case, it is assumed that the presence or absence of a certain event from a group is not related to the presence or absence of other events. With a system to predict divorce, it can help users who want to know the prediction of divorce. Based on the tests that have been carried out using the Black Box Method, the results show that the functionality of the divorce data prediction application has run as expected and the application is able to display the prediction results of Divorce Lawsuit and Divorce Divorce

A New Glycemic closed-loop control based on Dyna-Q for Type-1-Diabetes

Objective:Type 1 Diabetes Mellitus is an autoimmune disease which requires constant care from patients. Continuous Glucose Monitoring (CGM) devices allow to keep track of the glycemic trend for 24 h a day. Control algorithms are necessary to automate the therapy and to develop an artificial pancreas system. The objective of this study is to develop a fully-automated glycemic control system based on a Dyna-Q Reinforcement Learning algorithm that is able to automatically decide the insulin infusion without the need of carbohydrate information from the patient. Methods:A Dyna-Q Reinforcement Learning architecture is proposed to automate glycemic control while using only information on past CGM and insulin data, validated on data from 10 in silico patients. Results:The proposed glycemic predictor achieves an average RMSE and a MARD of 13.2 mg/dl and 6.9% on 10 virtual adults, and of 15.0 mg/dl and 7.2% on 12 real patients, while over 98.8% of the forecasts fall within the safe zones of the Clarke Error Grid. The controller is able to maintain the glucose levels of the virtual subjects in the target range for 60.7% of the simulation time on a 24 h scenario, without causing hypoglycemic events in 8 out of 10 patients. Conclusion:The proposed architecture is able to achieve good performance without exploiting information on carbohydrates and using a much smaller amount of training data compared to models in the literature. Significance:We proved that model-based Reinforcement Learning could be a valid approach for a human-safe fully-automated artificial pancreas.

Surrogate-assisted evolutionary neural architecture search with network embedding

To accelerate the performance estimation in neural architecture search, recently proposed algorithms adopt surrogate models to predict the performance of neural architectures instead of training the network from scratch. However, it is time-consuming to collect sufficient labeled architectures for surrogate model training. To enhance the capability of surrogate models using a small amount of training data, we propose a surrogate-assisted evolutionary algorithm with network embedding for neural architecture search (SAENAS-NE). Here, an unsupervised learning method is used to generate meaningful representation of each architecture and the architectures with more similar structures are closer in the embedding space, which considerably benefits the training of surrogate models. In addition, a new environmental selection based on a reference population is designed to keep diversity of the population in each generation and an infill criterion for handling the trade-off between convergence and model uncertainty is proposed for re-evaluation. Experimental results on three different NASBench and DARTS search space illustrate that network embedding makes the surrogate model achieve comparable or superior performance. The superiority of our proposed method SAENAS-NE over other state-of-the-art neural architecture algorithm has been verified in the experiments.

Muscular Human Cybertwin for Internet of Everything: A Pilot Study

The cybertwin-driven 6G that can obtain static and dynamic data stream of users provide an exciting potential for a novel muscular human cybertwin beyond traditonally used artificial neural networks (ANNs) and musculoskeletal models (MSMs). In this article, we propose the conceptual design of the muscular human cybertwin and construct a baseline model with an improved generalization ability over ANN and an easier adaptation to new data distributions over MSMs. In particular, we for the first time propose to combine ANN and MSM, which benefits from the combination of learning-based approaches and analytical approaches. We then experimentally compare different manners of the combination and demonstrate the better combining manner on our testing case. Finally, we evaluate our method on an open-sourced dataset and on data from wearable sensors from the aspects of joint moment prediction accuracy, data efficiency, generalization ability, and time efficiency of personalization. Our proposed method achieves accuracy similar with ANN and over 30 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> better than MSM with sufficient training data. Compared with ANN, the improved data efficiency is presented by the better accuracies with a small amount of training data, and the generalization ability to unseen walking conditions and new subjects are demonstrated by the over 70 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> accuracy improvements. Moreover, when fine-tuning the model, our algorithm is demonstrated by the time 75 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> shorter than calibrated MSM and the accuracy improvements.

Characterization of Subsurface Hydrogeological Structures With Convolutional Conditional Neural Processes on Limited Training Data

AbstractOne of the main issues in the application of statistical‐learning‐based methods to the characterization of hydrological phenomena is the complex parameterization of the high‐quality reconstruction. The difficulty of obtaining enough training data and generating multiple‐scale structures hinder the applications of deep‐learning‐based methods in hydrogeological modeling. Hydrogeological modeling problems can be regarded as stochastic processes, and we propose a novel hydrogeological modeling approach based on convolutional conditional neural processes (GM‐ConvCNPs), a meta‐learning approach for dealing with stochastic processes. In this work, GM‐ConvCNP is used to reconstruct the entire spatial structures of subsurface hydrological attributes and channels from a limited amount of conditioning data. To achieve 3‐D characterization of subsurface structures, 3‐D GM‐ConvCNP is proposed according to the spatial distribution characteristics of 3‐D conditioning data. Compared to other deep‐learning‐based methods, we use a small amount of training data and obtain positive model generalization capabilities. Four data sets of categorical and continuous hydrogeological structures are exploited in the experiments. Various validation tests including variograms, connectivity functions, and multi‐dimensional scaling (MDS) map are used to evaluate the quality of generated realizations. The proposed approach is able to significantly reduce training consumption and improve the performance of realizations compared to a set of different benchmark tests. The experiments confirm that the GM‐ConvCNP model can extract heterogeneous patterns by using meta‐learning from limited training data and reconstruct multiple‐scale hydrogeological structures. A case study demonstrates that our method can be applied to characterize multiple‐source hydrological variables.

Distilled and Contextualized Neural Models Benchmarked for Vulnerable Function Detection

Detecting vulnerabilities in programs is an important yet challenging problem in cybersecurity. The recent advancement in techniques of natural language understanding enables the data-driven research on automated code analysis to embrace Pre-trained Contextualized Models (PCMs). These models are pre-trained on the large corpus and can be fine-tuned for various downstream tasks, but their feasibility and effectiveness for software vulnerability detection have not been systematically studied. In this paper, we explore six prevalent PCMs and compare them with three mainstream Non-Contextualized Models (NCMs) in terms of generating effective function-level representations for vulnerability detection. We found that, although the detection performance of PCMs outperformed that of the NCMs, training and fine-tuning PCMs were computationally expensive. The budgets for deployment and inference are also considerable in practice, which may prevent the wide adoption of PCMs in the field of interest. However, we discover that, when the PCMs were compressed using the technique of knowledge distillation, they achieved similar detection performance but with significantly improved efficiency compared with their uncompressed counterparts when using 40,000 synthetic C functions for fine-tuning and approximately 79,200 real-world C functions for training. Among the distilled PCMs, the distilled CodeBERT achieved the most cost-effective performance. Therefore, we proposed a framework encapsulating the Distilled CodeBERT for an end-to-end Vulnerable function Detection (named DistilVD). To examine the performance of the proposed framework in real-world scenarios, DistilVD was tested on four open-source real-world projects with a small amount of training data. Results showed that DistilVD outperformed the five baseline approaches. Further evaluations on multi-class vulnerability detection also confirmed the effectiveness of DistilVD for detecting various vulnerability types.

Speech decoding from a small set of spatially segregated minimally invasive intracranial EEG electrodes with a compact and interpretable neural network

Objective. Speech decoding, one of the most intriguing brain-computer interface applications, opens up plentiful opportunities from rehabilitation of patients to direct and seamless communication between human species. Typical solutions rely on invasive recordings with a large number of distributed electrodes implanted through craniotomy. Here we explored the possibility of creating speech prosthesis in a minimally invasive setting with a small number of spatially segregated intracranial electrodes. Approach. We collected one hour of data (from two sessions) in two patients implanted with invasive electrodes. We then used only the contacts that pertained to a single stereotactic electroencephalographic (sEEG) shaft or an electrocorticographic (ECoG) stripe to decode neural activity into 26 words and one silence class. We employed a compact convolutional network-based architecture whose spatial and temporal filter weights allow for a physiologically plausible interpretation. Main results. We achieved on average 55% accuracy using only six channels of data recorded with a single minimally invasive sEEG electrode in the first patient and 70% accuracy using only eight channels of data recorded for a single ECoG strip in the second patient in classifying 26+1 overtly pronounced words. Our compact architecture did not require the use of pre-engineered features, learned fast and resulted in a stable, interpretable and physiologically meaningful decision rule successfully operating over a contiguous dataset collected during a different time interval than that used for training. Spatial characteristics of the pivotal neuronal populations corroborate with active and passive speech mapping results and exhibit the inverse space-frequency relationship characteristic of neural activity. Compared to other architectures our compact solution performed on par or better than those recently featured in neural speech decoding literature. Significance. We showcase the possibility of building a speech prosthesis with a small number of electrodes and based on a compact feature engineering free decoder derived from a small amount of training data.

Toward learning robust contrastive embeddings for binaural sound source localization.

Recent deep neural network based methods provide accurate binaural source localization performance. These data-driven models map measured binaural cues directly to source locations hence their performance highly depend on the training data distribution. In this paper, we propose a parametric embedding that maps the binaural cues to a low-dimensional space where localization can be done with a nearest-neighbor regression. We implement the embedding using a neural network, optimized to map points that are close to each other in the latent space (the space of source azimuths or elevations) to nearby points in the embedding space, thus the Euclidean distances between the embeddings reflect their source proximities, and the structure of the embeddings forms a manifold, which provides interpretability to the embeddings. We show that the proposed embedding generalizes well in various acoustic conditions (with reverberation) different from those encountered during training, and provides better performance than unsupervised embeddings previously used for binaural localization. In addition, the proposed method performs better than or equally well as a feed-forward neural network based model that directly estimates the source locations from the binaural cues, and it has better results than the feed-forward model when a small amount of training data is used. Moreover, we also compare the proposed embedding using both supervised and weakly supervised learning, and show that in both conditions, the resulting embeddings perform similarly well, but the weakly supervised embedding allows to estimate source azimuth and elevation simultaneously.

Small-data-driven fast seismic simulations for complex media using physics-informed Fourier neural operators

Deep learning (DL) seismic simulations have become a leading-edge field that could provide an effective alternative to traditional numerical solvers. We have developed a small-data-driven time-domain method for fast seismic simulations in complex media based on the physics-informed Fourier neural operator (FNO). Unlike most DL-based modeling schemes that either solve wave equations by embedding the physical constraints into the cost function or conduct physics-informed learning by incorporating the wave functions into convolutional neural networks (CNNs), the FNO uses a learning architecture similar to the structure of split-step Fourier wave propagators, which is composed of two CNNs formulated in the space and wavenumber domains, respectively. The space-domain CNN acts as a local trainable phase-screen compensation. The wavenumber-domain CNN represents a nonlocal spatial convolutional operator acting as a trainable wavenumber filter for the phase-shift process. The FNO method approximates the mathematical-physical behavior of wave equations through learning the mapping between seismic wavefields at different time/locations from training seismic data. That is, the learning process parameterizes the integral kernel directly in the Fourier space, so that we can establish an expressive and efficient architecture for a better balance between accuracy and performance than the traditional spatial CNNs. Applications to gradient, layered, and Marmousi velocity models demonstrate its performance in accuracy and efficiency. The FNO seismic simulation is a data-driven method that needs a small amount of training data, especially when using blended source training data. It is a discretization-independent method that is not subject to the limitation of spatial sampling and time steps imposed on traditional numerical solvers, implying that the training data can be discretized arbitrarily. It also is a model-independent method that can include absorption attenuation into seismic modeling without the need of viscoelastic wave equations.

로봇 디팔레타이징 산업용 인공지능 물체 검출 학습의 데이터 비용 절감을 위한 GAN 기반 데이터 증강 기법

The development of artificial intelligence (AI) has brought many changes to a variety of industrial fields. However, there are problems when applying AI due to the various environmental factors in industrial sites. In particular, the object detection model used for robot depalletizing often deteriorates over time due to the distribution of data changing every moment. While it is possible to maintain the performance of the model through additional data acquisition, continuous data acquisition is costly. Therefore, data augmentation techniques are used to reduce costs and maintain performance. However, data augmented by existing augmentation techniques do not significantly change in terms of the distribution of existing data, rendering it difficult to maintain the performance of the object detection model. In this paper, we develop a data augmentation pipeline based on generative adversarial networks, which is effective at maintaining performance and reducing the cost of object detection models. The proposed pipeline is composed of a data generator model and an object detection model. The generator model uses a small amount of training data to generate data with a new distribution, while the object detection model is trained with both training and generated data. The object detection model trained through the pipeline with 100 pieces of training data exhibited better performance on new data distribution by 9.9% AP compared to the model trained with 2000 pieces of training data. In addition, the results of the qualitative analysis confirmed that a representative error occurring in robot depalletizing could be improved by the proposed pipeline.

EcoTransLearn: an R-package to easily use transfer learning for ecological studies-a plankton case study.

In recent years, Deep Learning (DL) has been increasingly used in many fields, in particular in image recognition, due to its ability to solve problems where traditional machine learning algorithms fail. However, building an appropriate DL model from scratch, especially in the context of ecological studies, is a difficult task due to the dynamic nature and morphological variability of living organisms, as well as the high cost in terms of time, human resources and skills required to label a large number of training images. To overcome this problem, Transfer Learning (TL) can be used to improve a classifier by transferring information learnt from many domains thanks to a very large training set composed of various images, to another domain with a smaller amount of training data. To compensate the lack of 'easy-to-use' software optimized for ecological studies, we propose the EcoTransLearn R-package, which allows greater automation in the classification of images acquired with various devices (FlowCam, ZooScan, photographs, etc.), thanks to different TL methods pre-trained on the generic ImageNet dataset. EcoTransLearn is an open-source package. It is implemented in R and calls Python scripts for image classification step (using reticulate and tensorflow libraries). The source code, instruction manual and examples can be found at https://github.com/IFREMER-LERBL/EcoTransLearn. Supplementary data are available at Bioinformatics online.

Semi-supervised classification and clustering analysis for variable stars

ABSTRACT The immense amount of time series data produced by astronomical surveys has called for the use of machine learning algorithms to discover and classify several million celestial sources. In the case of variable stars, supervised learning approaches have become commonplace. However, this needs a considerable collection of expert-labelled light curves to achieve adequate performance, which is costly to construct. To solve this problem, we introduce two approaches. First, a semi-supervised hierarchical method, which requires substantially less trained data than supervised methods. Second, a clustering analysis procedure that finds groups that may correspond to classes or subclasses of variable stars. Both methods are primarily supported by dimensionality reduction of the data for visualization and to avoid the curse of dimensionality. We tested our methods with catalogues collected from the Optical Gravitational Lensing Experiment (OGLE), the Catalina Sky Survey (CSS), and the Gaia survey. The semi-supervised method reaches a performance of around 90 per cent for all of our three selected catalogues of variable stars using only $5{{\ \rm per\ cent}}$ of the data in the training. This method is suitable for classifying the main classes of variable stars when there is only a small amount of training data. Our clustering analysis confirms that most of the clusters found have a purity over 90 per cent with respect to classes and 80 per cent with respect to subclasses, suggesting that this type of analysis can be used in large-scale variability surveys as an initial step to identify which classes or subclasses of variable stars are present in the data and/or to build training sets, among many other possible applications.

A self-supervised COVID-19 CT recognition system with multiple regularizations

The diagnosis of Coronavirus Disease 2019 (COVID-19) exploiting machine learning algorithms based on chest computed tomography (CT) images has become an important technology. Though many excellent computer-aided methods leveraging CT images have been designed, they do not possess sufficiently high recognition accuracy. Besides, these methods entail vast amounts of training data, which might be difficult to be satisfied in some real-world applications. To address these two issues, this paper proposes a novel COVID-19 recognition system based on CT images, which has high recognition accuracy, while only requiring a small amount of training data. Specifically, the system possesses the following three improvements: 1) Data: a novel redesigned BCELoss that incorporates Label Smoothing, Focal Loss, and Label Weighting Regularization (LSFLLW-R) technique for optimizing the solution space and preventing overfitting, 2) Model: a backbone network processed by two-phase contrastive self-supervised learning for classifying multiple labels, and 3) Method: a decision-fusing ensemble learning method for getting a more stable system, with balanced metric values. Our proposed system is evaluated on the small-scale expanded COVID-CT dataset, achieving an accuracy of 94.3%, a precision of 94.1%, a recall (sensitivity) of 93.4%, an F1-score of 94.7%, and an Area Under the Curve (AUC) of 98.9%, for COVID-19 diagnosis, respectively. These experimental results verify that our system can not only identify pathological locations effectively, but also achieve better performance in terms of accuracy, generalizability, and stability, compared with several other state-of-the-art COVID-19 diagnosis methods.

Cloze-Style Data Augmentation for Few-Shot Intent Recognition

Intent recognition aims to identify users’ potential intents from their utterances, which is a key component in task-oriented dialog systems. A real challenge, however, is that the number of intent categories has grown faster than human-annotated data, resulting in only a small amount of data being available for many new intent categories. This lack of data leads to the overfitting of traditional deep neural networks on a small amount of training data, which seriously affects practical applications. Hence, some researchers have proposed few-shot learning should address the data-scarcity issue. One of the efficient methods is text augmentation, which always generates noisy or meaningless data. To address these issues, we propose leveraging the knowledge in pre-trained language models and constructed the cloze-style data augmentation (CDA) model. We employ unsupervised learning to force the augmented data to be semantically similar to the initial input sentences and contrastive learning to enhance the uniqueness of each category. Experimental results on CLINC-150 and BANKING-77 datasets show the effectiveness of our proposal by its beating of the competitive baselines. In addition, we conducted an ablation study to verify the function of each module in our models, and the results illustrate that the contrastive learning module plays the most important role in improving the recognition accuracy.

Dynamics-disentangled deep learning model for multi-cycle prediction of unsteady flow field

The prediction of an unsteady flow field inherently involving high-dimensional dynamics is challenging. The multi-cycle prediction is especially difficult due to the inevitably accumulated errors over time. A novel deep learning model is proposed in this paper to disentangle the high-dimensional dynamics into two separate attributes that, respectively, represent spatial and temporal dynamics. A continuous mapping of temporal dynamics is subsequently constructed, which alleviates the error accumulation and, thus, contributes to the long-term prediction of the unsteady flow field. The dynamics-disentangled deep learning model (D3LM) processes sequential image data of the unsteady flow field and is constituted by three sub-networks, an encoder introducing a stochastic latent variable to explicitly model the low-order temporal dynamics (called varying attribute herein) and extracting multi-level representations of spatial dynamics (called consistent attribute herein), a decoder integrating the disentangled attributes and generating a future flow field, and a discriminator improving the quality of the predicted flow field. The proposed model is evaluated by two simulated datasets of unsteady flows around a circular cylinder at divergent Reynolds numbers. Benefiting from modeling the continuous distribution of temporal dynamics with the stochastic latent variable, the proposal can give multi-cycle future predictions with high accuracy both spatially and temporally on the two datasets with a small amount of training data. Our work demonstrates the potential practicability of deep learning techniques for modeling the long-term nonlinear laws of unsteady flow.

Automatic depression score estimation with word embedding models

Depression is one of the most common mental health illnesses. The biggest obstacle lies in an efficient and early detection of the disorder. Self-report questionnaires are the instruments used by medical experts to elaborate a diagnosis. These questionnaires were designed by analyzing different depressive symptoms. However, factors such as social stigmas negatively affect the success of traditional methods. This paper presents a novel approach for automatically estimating the degree of depression in social media users. In this regard, we addressed the task Measuring the Severity of the Signs of Depression of eRisk 2020, an initiative in the CLEF Conference. We aimed to explore neural language models to exploit different aspects of the subject’s writings depending on the symptom to capture. We devised two distinct methods based on the symptoms’ sensitivity in terms of willingness on commenting about them publicly. The first exploits users’ general language based on their publications. The second seeks more direct evidence from publications that specifically mention the symptoms concerns. Both methods automatically estimate the Beck Depression Inventory (BDI-II) total score. For evaluating our proposals, we used benchmark Reddit data for depression severity estimation. Our findings showed that approaches based on neural language models are a feasible alternative for estimating depression rating scales, even when small amounts of training data are available.

FusionNet: A Convolution–Transformer Fusion Network for Hyperspectral Image Classification

In recent years, deep-learning-based hyperspectral image (HSI) classification networks have become one of the most dominant implementations in HSI classification tasks. Among these networks, convolutional neural networks (CNNs) and attention-based networks have prevailed over other HSI classification networks. While convolutional neural networks with perceptual fields can effectively extract local features in the spatial dimension of HSI, they are poor at capturing the global and sequential features of spectral–spatial information; networks based on attention mechanisms, for example, Transformer, usually have better ability to capture global features, but are relatively weak in discriminating local features. This paper proposes a fusion network of convolution and Transformer for HSI classification, known as FusionNet, in which convolution and Transformer are fused in both serial and parallel mechanisms to achieve the full utilization of HSI features. Experimental results demonstrate that the proposed network has superior classification results compared to previous similar networks, and performs relatively well even on a small amount of training data.

Identifying self-admitted technical debt in issue tracking systems using machine learning

Technical debt is a metaphor indicating sub-optimal solutions implemented for short-term benefits by sacrificing the long-term maintainability and evolvability of software. A special type of technical debt is explicitly admitted by software engineers (e.g. using a TODO comment); this is called Self-Admitted Technical Debt or SATD. Most work on automatically identifying SATD focuses on source code comments. In addition to source code comments, issue tracking systems have shown to be another rich source of SATD, but there are no approaches specifically for automatically identifying SATD in issues. In this paper, we first create a training dataset by collecting and manually analyzing 4,200 issues (that break down to 23,180 sections of issues) from seven open-source projects (i.e., Camel, Chromium, Gerrit, Hadoop, HBase, Impala, and Thrift) using two popular issue tracking systems (i.e., Jira and Google Monorail). We then propose and optimize an approach for automatically identifying SATD in issue tracking systems using machine learning. Our findings indicate that: 1) our approach outperforms baseline approaches by a wide margin with regard to the F1-score; 2) transferring knowledge from suitable datasets can improve the predictive performance of our approach; 3) extracted SATD keywords are intuitive and potentially indicating types and indicators of SATD; 4) projects using different issue tracking systems have less common SATD keywords compared to projects using the same issue tracking system; 5) a small amount of training data is needed to achieve good accuracy.

A novel few-shot classification framework for diabetic retinopathy detection and grading

Diabetes Retinopathy (DR) is a major microvascular complication of diabetes. Computer-Aided Diagnosis (CAD) tools for DR management are primarily developed using Artificial Intelligence (AI) methods, such as machine and deep learning algorithms. DR diagnostic tools have been developed in recent years using deep learning models. Thus, these models require large amounts of data for training. Consequently, these huge amounts of data are not balanced due to fewer cases in the dataset. To solve the problems associated with training models with small datasets, such as overfitting and poor approximation, this paper proposes a paradigm called Few-Shot Learning (FSL) which uses a relatively small amount of training data to train the models effectively. This paper proposes a novel prototype network, a type of FSL classification network capable of grading and detecting DR based on attention. The DRNet framework uses episodic learning to train its model on few-shot classification tasks. We developed a DRNet based on the APTOS2019 dataset for diabetic detection and grading. In the proposed network, aggregated transformations and gradient activations of classes are leveraged to design the attention mechanism to capture image representations. As a result, the system achieves 99.73 % accuracy, 99.82 % sensitivity, 99.63 % specificity in DR detection, 98.18 % accuracy, 97.41% sensitivity, and 99.55% specificity in DR grading. An analysis of objective performance metrics and model interpretation shows that the proposed model can detect DR more efficiently and grade the severity more accurately when using unseen fundus images than existing state-of-the-art methods.Therefore, this tool could help provide a second opinion to an ophthalmologist about the severity level of DR.