Linear Classifier Research Articles

Understanding College Students’ Satisfaction With ChatGPT: An Exploratory And Predictive Machine Learning Approach Using Feature Engineering

Artificial Intelligence (AI) technologies are continually improving and becoming more pervasive in many facets of our lives. ChatGPT is one such cutting-edge artificial intelligence application, and it has received a lot of worldwide media attention, specifically from educationists, technologists, and learners. It is imperative to understand and evaluate the impact of ChatGPT on computer science students as it directly and holistically influences them. A quantitative instrumental case study explores ChatGPT’s impact on early adopters in education. A survey of undergraduate computer science students at a state university of Delhi was conducted to get insight into their opinion on adopting this revolutionising technology for their education, career, and overall satisfaction. An end-to-end data science approach is applied to encompass exploratory and predictive modelling with feature engineering solutions. Results reveal the most influential features contributing to students’ satisfaction in adopting ChatGPT for their day-to-day chores concerning their social life, education, and career. The Linear Support Vector classifier, a machine learning algorithm for predicting the satisfaction or dissatisfaction in students’ shows an accuracy score of 72.73% and 97.72%, respectively. The AUC for this multiclass prediction model is convincing and is 0.74, 0.71, and 0.96 for satisfied, neutral, and dissatisfied classes, respectively.

A multimodal transfer learning framework for the classification of disaster-related social media images

Rapidly classifying disaster-related social media (SM) images during a catastrophe event is critical for enhancing disaster response efforts. However, the biggest challenge lies in acquiring labeled data for an ongoing (target) disaster to train supervised learning-based models, given that the labeling process is both time-consuming and costly. In this study, we address this challenge by proposing a new multimodal transfer learning framework for the real-time classification of SM images of the target disaster. The proposed framework is based on Contrastive Language-Image Pretraining (CLIP) model, jointly pretrained on a dataset of image-text pairs via contrastive learning. We propose two distinct methods to design our classification framework (1) Zero-Shot CLIP: it learns visual representations from images paired with natural language descriptions of classes. By utilizing the vision and language capabilities of CLIP, we extract meaningful features from unlabeled target disaster images and map them to semantically related textual class descriptions, enabling image classification without training on disaster-specific data. (2) Linear-Probe CLIP: it further enhances the performance and involves training a linear classifier on top of the pretrained CLIP model’s features, specifically tailored to the disaster image classification task. By optimizing the linear-probe classifier, we improve the model’s ability to discriminate between different classes and achieve higher performance without the need for labeled data of the target disaster. Both methods are evaluated on a benchmark X (formerly Twitter) dataset comprising images of seven real-world disaster events. The experimental outcomes showcase the efficacy of the proposed methods, with Linear-Probe CLIP achieving a remarkable 7% improvement in average F1-score relative to the state-of-the-art methods.

Self-Supervised Simple Siamese Framework for Fault Diagnosis of Rotating Machinery With Unlabeled Samples.

Fault diagnosis is vital to ensuring the security of rotating machinery operations. While fault data obtained from mechanical equipment for this issue are often insufficient and of no labels. In this case, supervised algorithms cannot come into play. Hence, this article proposes a self-supervised simple Siamese framework (SSF) for bearing fault diagnosis based on the contrastive learning algorithm SimSiam which uses a simplified Siamese network to find the distinguishable features of different fault categories. SSF consists of a weight-sharing encoder applied on two inputs, a nonlinear predictor and a linear classifier. SSF learns invariant characteristics of fault samples via maximizing the similarity between two views of each inputted sample. Several data augmentation (DA) methods for vibration signals, which provide different sample views for the model, are also studied, for it is crucial for contrastive learning. After fine-tuning the learned encoder and a linear layer classifier with a small subset of labeled data (1%-5% of the total samples), the network achieves satisfactory performance for bearing fault diagnosis. A series of experiments based on the data from three different scenarios are used to verify the proposed methods, getting 100%, 99.38%, and 98.87% accuracy separately.

Improving Robustness of Intent Detection Under Adversarial Attacks: A Geometric Constraint Perspective.

Deep neural networks (DNNs)-based natural language processing (NLP) systems are vulnerable to being fooled by adversarial examples presented in recent studies. Intent detection tasks in dialog systems are no exception, however, relatively few works have been attempted on the defense side. The combination of linear classifier and softmax is widely used in most defense methods for other NLP tasks. Unfortunately, it does not encourage the model to learn well-separated feature representations. Thus, it is easy to induce adversarial examples. In this article, we propose a simple, yet efficient defense method from the geometric constraint perspective. Specifically, we first propose an M-similarity metric to shrink variances of intraclass features. Intuitively, better geometric conditions of feature space can bring lower misclassification probability (MP). Therefore, we derive the optimal geometric constraints of anchors within each category from the overall MP (OMP) with theoretical guarantees. Due to the nonconvex characteristic of the optimal geometric condition, it is hard to satisfy the traditional optimization process. To this end, we regard such geometric constraints as manifold optimization processes in the Stiefel manifold, thus naturally avoiding the above challenges. Experimental results demonstrate that our method can significantly improve robustness compared with baselines, while retaining the excellent performance on normal examples.

Contrastive visual clustering for improving instance-level contrastive learning as a plugin

Contrastive learning has achieved remarkable success in computer vision, however it is built on instance-level discrimination which leaves the valuable intra-class correlation in dataset unexploited. Current semantic clustering methods are proven to be helpful but they would suffer from the error accumulated in the iteration process without ground-truth guidance. In an attempt to remedy the clustering error accumulation when utilizing intra-class correlation for contrastive learning, we propose an online Contrastive Visual Clustering (CVC) method with two actions: gathering instances with highly similar feature embeddings, and penalizing instances being clustered with low confidence. CVC can integrate with not only contrastive learning but also arbitrary self-supervised learning frameworks simply as a plugin. Under various experiment settings, we show that CVC improves the linear classification performance by a large margin for models pre-trained with self-supervised representation learning, in both image and video scenarios. The code is available at https://github.com/yliu1229/CVC.

Bacterial Disease Detection of Cherry Plant Using Deep Features

Although the cherry plant is widely grown in the world and Turkey, it is a fruit tree that is difficult to grow and maintain. It can be exposed to various pesticide diseases, especially during fruiting. Today, approaches based on expert reviews and analyses are used for the identification of these diseases. In addition, cherry producers are trying to detect diseases with their knowledge based on experience. Computer-aided agricultural analysis systems are also being developed depending on the rapid developments in technology. These systems help to monitor all processes from planting, cultivation, and harvesting of agricultural products and to make decisions to grow the products healthily. One of the most important issues to be detected and monitored with these systems is plant diseases. The features of the cherry plant disease will be determined by using a pre-trained convolutional neural network (CNN) model which is DarkNet-19, within the scope of this study. These machine learning-based features have been used for the detection of bacteria-based diseases commonly seen on the leaves of cherry plants. The acquired features are classified with Linear Discriminant Analysis, K-Nearest Neighbor, and Support Vector Machine classifiers to solve the multi-class problem including diseased (less and very) and healthy plants. The experimental results show that a success rate of 88.1% was obtained in the detection of the disease.

Energy optimization for wireless sensor network using minimum redundancy maximum relevance feature selection and classification techniques

In wireless sensor networks (WSN), conserving energy is usually a basic issue, and several approaches are applied to optimize energy consumption. In this article, we adopt feature selection approaches by using minimum redundancy maximum relevance (MRMR) as a feature selection technique to minimize the number of sensors thereby conserving energy. MRMR ranks the sensors according to their significance. The selected features are then classified by different types of classifiers; SVM with linear kernel classifier, naïve Bayes classifier, and k-nearest neighbors classifier (KNN) to compare accuracy values. The simulation results illustrated an improvement in the lifetime extension factor of sensors and showed that the KNN classifier gives better results than the naïve Bayes and SVM classifier.

Indirect functional connectivity does not predict overall survival in glioblastoma

BackgroundLesion network mapping (LNM) is a popular framework to assess clinical syndromes following brain injury. The classical approach involves embedding lesions from patients into a normative functional connectome and using the corresponding functional maps as proxies for disconnections. However, previous studies indicated limited predictive power of this approach in behavioral deficits. We hypothesized similarly low predictiveness for overall survival (OS) in glioblastoma (GBM). MethodsA retrospective dataset of patients with GBM was included (n = 99). Lesion masks were registered in the normative space to compute disconnectivity maps. The brain functional normative connectome consisted in data from 173 healthy subjects obtained from the Human Connectome Project. A modified version of the LNM was then applied to core regions of GBM masks. Linear regression, classification, and principal component (PCA) analyses were conducted to explore the relationship between disconnectivity and OS. OS was considered both as continuous and categorical (low, intermediate, and high survival) variable. ResultsThe results revealed no significant associations between OS and network disconnection strength when analyzed at both voxel-wise and classification levels. Moreover, patients stratified into different OS groups did not exhibit significant differences in network connectivity patterns. The spatial similarity among the first PCA of network maps for each OS group suggested a lack of distinctive network patterns associated with survival duration. ConclusionsCompared with indirect structural measures, functional indirect mapping does not provide significant predictive power for OS in patients with GBM. These findings are consistent with previous research that demonstrated the limitations of indirect functional measures in predicting clinical outcomes, underscoring the need for more comprehensive methodologies and a deeper understanding of the factors influencing clinical outcomes in this challenging disease.

Crash severity analysis: A data-enhanced double layer stacking model using semantic understanding

The crash severity analysis is of significant importance in traffic crash prevention and emergency resource allocation. A range of innovations offers potential traffic crash severity prediction models to improve road safety. However, the semantic information inherent in traffic crash data, which is crucial in enabling a deeper understanding of its underlying factors and impacts, has yet to be fully utilized. Moreover, traffic crash data are commonly characterized by a small sample size, which leads to sample imbalance problem resulting in prediction performance decline. To tackle these problems, we propose a semantic understanding-based data-enhanced double-layer stacking model, named EnLKtreeGBDT, for crash severity prediction. Specifically, to fully leverage the inherent semantic information within traffic crash data and analyze the factors influencing crashes, we design a semantic enhancement module for multi-dimensional feature extraction. This module aims to enhance the understanding of crash semantics and improve prediction accuracy. Then we introduce a data enhancement module that utilizes data denoising and migration techniques to address the challenge of data imbalance, reducing the prediction model's dependence on large sample crash data. Furthermore, we construct a two-layer stacking model that combines multiple linear and nonlinear classifiers. This model is designed to augment the capability of learning linear and nonlinear mixed relationships, thereby improving the accuracy of predicting the severity of crashes on complex urban roads. Experiments on historical datasets of UK road safety crashes validate the effectiveness of the proposed model, and superior performance of prediction precision is achieved compared with the state-of-the-arts. The ablation experiments on both semantic and data enhancement modules further confirm the indispensability of each module in the proposed model.

EResNet-SVM: an overfitting-relieved deep learning model for recognition of plant diseases and pests.

The accurate recognition and early warning for plant diseases and pests are a prerequisite of intelligent prevention and control for plant diseases and pests. As a result of the phenotype similarity of the hazarded plant after plant diseases and pests occur, as well as the interference of the external environment, traditional deep learning models often face the overfitting problem in phenotype recognition of plant diseases and pests, which leads to not only the slow convergence speed of the network, but also low recognition accuracy. Motivated by the above problems, the present study proposes a deep learning model EResNet-support vector machine (SVM) to alleviate the overfitting for the recognition and classification of plant diseases and pests. First, the feature extraction capability of the model is improved by increasing feature extraction layers in the convolutional neural network. Second, the order-reduced modules are embedded and a sparsely activated function is introduced to reduce model complexity and alleviate overfitting. Finally, a classifier fused by SVM and fully connected layers are introduced to transforms the original non-linear classification problem into a linear classification problem in high-dimensional space to further alleviate the overfitting and improve the recognition accuracy of plant diseases and pests. The ablation experiments further demonstrate that the fused structure can effectively alleviate the overfitting and improve the recognition accuracy. The experimental recognition results for typical plant diseases and pests show that the proposed EResNet-SVM model has 99.30% test accuracy for eight conditions (seven plant diseases and one normal), which is 5.90% higher than the original ResNet18. Compared with the classic AlexNet, GoogLeNet, Xception, SqueezeNet and DenseNet201 models, the accuracy of the EResNet-SVM model has improved by 5.10%, 7%, 8.10%, 6.20% and 1.90%, respectively. The testing accuracy of the EResNet-SVM model for 6 insect pests is 100%, which is 3.90% higher than that of the original ResNet18 model. This research provides not only useful references for alleviating the overfitting problem in deep learning, but also a theoretical and technical support for the intelligent detection and control of plant diseases and pests. © 2024 Society of Chemical Industry.

Carrier-Free Ultra-Wideband Sensor Target Recognition in the Jungle Environment

Carrier-free ultra-wideband sensors have high penetrability anti-jamming solid ability, which is not easily affected by the external environment, such as weather. Also, it has good performance in the complex jungle environment. In this paper, we propose a jungle vehicle identification system based on a carrier-free ultra-wideband sensor. Firstly, a composite jungle environment with the target vehicle is modeled. From this model, the simulation obtains time-domain echoes under the excitation of carrier-free ultra-wideband sensor signals in different orientations. Secondly, the time-domain signals are transformed into MTF images through the Markov transfer field to show the statistical characteristics of the time-domain echoes. At the same time, we propose an improved RepVGG network. The structure of the RepVGG network contains five stages, which consist of several RepVGG Blocks. Each RepVGG Block is created by combining convolutional kernels of different sizes using a weighted sum. We add the self-attention module to the output of stage 0 to improve the ability to extract the features of the MTF map and better capture the complex relationship between characteristics during training. In addition, a self-attention module is added before the linear layer classification output in stage 4 to improve the classification accuracy of the network. Moreover, a combined cross-entropy loss and sparsity penalty loss function helps enhance the performance and accuracy of the network. The experimental results show that the system can recognize jungle vehicle targets well.

Feature Importance Analysis and Machine Learning for Alzheimer’s Disease Early Detection: Feature Fusion of the Hippocampus, Entorhinal Cortex, and Standardized Uptake Value Ratio

Introduction: Alzheimer’s disease (AD) is a progressive neurological disorder characterized by mild memory loss and ranks as a leading cause of mortality in the USA, accounting for approximately 120,000 deaths per year. It is also the primary form of dementia. Early detection is critical for timely intervention as the neurodegenerative process often starts 15–20 years before cognitive symptoms manifest. This study focuses on determining feature importance in AD classification using fused texture features from 3D magnetic resonance imaging hippocampal and entorhinal cortex and standardized uptake value ratio (SUVR) derived from positron emission tomography (PET) images. Methods: To achieve this objective, we employed four distinct classifiers (Linear Support Vector Classification, Linear Discriminant Analysis, Logistic Regression, and Logistic Regression Classifier with Stochastic Gradient Descent Learning). These classifiers were used to derive both average and top-ranked importance scores for each feature based on their outputs. Our framework is designed to distinguish between two classes, AD-negative (or mild cognitive impairment stable [MCIs]) and AD-positive (or MCI conversion [MCIc]), using a probabilistic neural network classifier and the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Results: The findings from the feature importance highlight the crucial role of the GLCM texture features extracted from the hippocampus and entorhinal cortex, demonstrating their superior performance compared to the volume and SUVR. GLCM texture AD classification achieved approximately 90% sensitivity in identifying MCIc cases while maintaining low false positives (below 30%) when fused with other features. Moreover, the receiver operating characteristic curves validate the GLCMs’ superior performance in distinguishing between MCIs and MCIc. Additionally, fusing different types of features improved classification performance compared to relying solely on any single feature category. Conclusion: Our study emphasizes the pivotal role of GLCM texture features in early Alzheimer’s detection.

Advancing Pulmonary Nodule Diagnosis by Integrating Engineered and Deep Features Extracted from CT Scans

Enhancing lung cancer diagnosis requires precise early detection methods. This study introduces an automated diagnostic system leveraging computed tomography (CT) scans for early lung cancer identification. The main approach is the integration of three distinct feature analyses: the novel 3D-Local Octal Pattern (LOP) descriptor for texture analysis, the 3D-Convolutional Neural Network (CNN) for extracting deep features, and geometric feature analysis to characterize pulmonary nodules. The 3D-LOP method innovatively captures nodule texture by analyzing the orientation and magnitude of voxel relationships, enabling the distinction of discriminative features. Simultaneously, the 3D-CNN extracts deep features from raw CT scans, providing comprehensive insights into nodule characteristics. Geometric features and assessing nodule shape further augment this analysis, offering a holistic view of potential malignancies. By amalgamating these analyses, our system employs a probability-based linear classifier to deliver a final diagnostic output. Validated on 822 Lung Image Database Consortium (LIDC) cases, the system’s performance was exceptional, with measures of 97.84%, 98.11%, 94.73%, and 0.9912 for accuracy, sensitivity, specificity, and Area Under the ROC Curve (AUC), respectively. These results highlight the system’s potential as a significant advancement in clinical diagnostics, offering a reliable, non-invasive tool for lung cancer detection that promises to improve patient outcomes through early diagnosis.

Immersive Emotion Analysis in VR Environments: A Sensor-Based Approach to Prevent Distortion

As virtual reality (VR) technology advances, research has focused on enhancing VR content for a more realistic user experience. Traditional emotion analysis relies on surveys, but they suffer from delayed responses and decreased immersion, leading to distorted results. To overcome these limitations, we propose an emotion analysis method using sensor data in the VR environment. Our approach can take advantage of the user’s immediate response and not reduce immersion. Linear regression, classification analysis, and tree-based methods were applied to electrocardiogram and galvanic skin response (GSR) sensor data to measure valence and arousal values. We introduced a novel emotional dimension model by analyzing correlations between emotions and the valence and arousal values. Experimental results demonstrated the highest accuracy of 77% and 92.3% for valence and arousal prediction, respectively, using GSR sensor data. Furthermore, an accuracy of 80.25% was achieved in predicting valence and arousal using nine emotions. Our proposed model improves VR content through more accurate emotion analysis in a VR environment, which can be useful for targeting customers in various industries, such as marketing, gaming, education, and healthcare.

Sentiment Analysis - An optimized Weighted Horizontal Ensemble approach

Sentiment Analysis has gained authority as one of the primary means of analyzing feedbacks and opinion by individuals, organizations and governments. The result of sentiment analysis informs an organization on areas to improve and how best to manage customers. While sentiment analysis may be misleading as no algorithm has been considered 100% efficient, the choice of algorithms can optimize the result based on the dataset in question. This paper aims at studying various algorithms and implementing a weighted horizontal ensemble algorithm as a panacea to low confidence level in the results of sentiment analysis. We designed a system that implements the original Naive Bayes algorithm, Multinomial Naïve Bayes algorithm, Bernoulli Native Bayes algorithm, Logistic Regression algorithm, Linear Support Vector Classifier algorithm and the Stochastic Gradient Descent algorithm. Our dataset was sourced from the Stanford University. It contains fifty thousand (50,000) movie reviews. Dataset from the Nigerian movie review was used to test the models. The reviews were encoded as a sequence of word indices. An accuracy of over 91% was achieved. The Ensemble technique delivered an F1-measure of 90%. Ensemble technique provides a more reliable confidence level on sentiment analysis. The researchers also discovered that change in writing style can affect the performance of sentiment analysis

Concept-drifts adaptation for machine learning EEG epilepsy seizure prediction.

Seizure prediction remains a challenge, with approximately 30% of patients unresponsive to conventional treatments. Addressing this issue is crucial for improving patients' quality of life, as timely intervention can mitigate the impact of seizures. In this research field, it is critical to identify the preictal interval, the transition from regular brain activity to a seizure. While previous studies have explored various Electroencephalogram (EEG) based methodologies for prediction, few have been clinically applicable. Recent studies have underlined the dynamic nature of EEG data, characterised by data changes with time, known as concept drifts, highlighting the need for automated methods to detect and adapt to these changes. In this study, we investigate the effectiveness of automatic concept drift adaptation methods in seizure prediction. Three patient-specific seizure prediction approaches with a 10-minute prediction horizon are compared: a seizure prediction algorithm incorporating a window adjustment method by optimising performance with Support Vector Machines (Backwards-Landmark Window), a seizure prediction algorithm incorporating a data-batch (seizures) selection method using a logistic regression (Seizure-batch Regression), and a seizure prediction algorithm with a dynamic integration of classifiers (Dynamic Weighted Ensemble). These methods incorporate a retraining process after each seizure and use a combination of univariate linear features and SVM classifiers. The Firing Power was used as a post-processing technique to generate alarms before seizures. These methodologies were compared with a control approach based on the typical machine learning pipeline, considering a group of 37 patients with Temporal Lobe Epilepsy from the EPILEPSIAE database. The best-performing approach (Backwards-Landmark Window) achieved results of 0.75 ± 0.33 for sensitivity and 1.03 ± 1.00 for false positive rate per hour. This new strategy performed above chance for 89% of patients with the surrogate predictor, whereas the control approach only validated 46%.

Stock Price Prediction and Stock Behaviour using GA and LR with NIFTY50 index Based on Daily Charts

The price of a company's stock, which can increase in lockstep with the price of a single share, is the one of the indicator to measure its performance. Clients or stockholding companies find it challenging to make long-term projections regarding the value of specific stocks due to the unpredictable nature of stock prices. Consequently, there is no business-related subject more talked about than stock market predictions. It is crucial to resolve this issue in a way that benefits buyers and investors because they frequently experience investment losses. Machine learning is useful in developing model for stock value predictions. We are utilizing Python and Linear Regression, one of the Machine Learning statistical techniques for predictive analysis, to create a stock price prediction website in order to address this issue. Our study primarily focuses on the NIFTY50 index's performance in distributed lag with the purpose of predicting stock prices in the Indian stock market. Several useful characteristics of the NIFTY50 lag index were extracted by means of a genetic algorithm. After that, we uncovered hidden correlations between the stock index and a given stock's trend by using the linear regression classifier. For the purpose of testing our approach, we used it to forecast the future of three distinct equities. In comparison to state-of-the-art forecasting methodologies, our experimental results demonstrated an accuracy of 82.55%. For the purpose of predicting daily changes in stock prices, the NIFTY50 stock index proved to be useful.

Analyzing lower body movements using machine learning to classify autistic children

Autism spectrum disorder is a neurological disorder that affects children at an early age and its symptoms appear in varying degrees. Early detection of ASD and follow-up of treatment by specialists help improve autistic people's lives and reduce the symptoms they suffer from. In the past few years, a group of studies conducted gait analysis to diagnose children with autism, and although the results of these studies showed weak motor skills of Autism Spectrum Disorder children (ASD) compared to their peers of Typically Developing children (TD), there are some challenges facing researchers in this field. The first is the ability to create A safe, effective, friendly environment that reduces tension in ASD children and helps classify children into ASD/TD effectively. And the second is the ability to create an automated ASD diagnosis system. In this work, a system based on machine learning and augmented reality technology has been proposed to create an effective and friendly interactive environment that helps diagnose ASD and classify children to ASD / TD. The movement of the lower limbs of 15 ASD children and 15 TD children are tracked via a marker-less technique through a Kinect v2 sensor when playing a specific game based on kicking the ball task, that is using augmented reality technology. Then the kinematics features are extracted through Linear Discriminant Analysis technology. To determine these features' ability in the classification process, three machine learning algorithms are used, namely Support Vector Machine, Linear Discriminant Analysis, and Decision Tree. The results showed that the obtained kinematics features can classify children into ASD/TD, where the Support Vector Machine classifier obtained a high accuracy of 96.7% with sensitivity equal to 93.3% and specificity equal to 100%. In addition, the Linear Discriminant Analysis classifier obtained less accuracy of 93.3% with sensitivity equal to 93.3% and specificity equal to 93.3% compared with the Support Vector Machine classifier. Where the Decision Tree classifier obtained the lowest accuracy of 80.0% with sensitivity equal to 73.3% and specificity equal to 86.7% compared with the Support Vector Machine and Linear Discriminant Analysis classifiers. In addition to that, when using Linear Discriminant Analysis to reduce dimensions and extract features, the accuracy, sensitivity, and specificity of the Support Vector Machine classifier were increased and achieved by 100%. Based on these results, it has been proven that the proposed method is effective in classifying children to ASD/TD.

The effect of harmonization on the variability of PET radiomic features extracted using various segmentation methods

PurposeThis study aimed to examine the robustness of positron emission tomography (PET) radiomic features extracted via different segmentation methods before and after ComBat harmonization in patients with non-small cell lung cancer (NSCLC).MethodsWe included 120 patients (positive recurrence = 46 and negative recurrence = 74) referred for PET scanning as a routine part of their care. All patients had a biopsy-proven NSCLC. Nine segmentation methods were applied to each image, including manual delineation, K-means (KM), watershed, fuzzy-C-mean, region-growing, local active contour (LAC), and iterative thresholding (IT) with 40, 45, and 50% thresholds. Diverse image discretizations, both without a filter and with different wavelet decompositions, were applied to PET images. Overall, 6741 radiomic features were extracted from each image (749 radiomic features from each segmented area). Non-parametric empirical Bayes (NPEB) ComBat harmonization was used to harmonize the features. Linear Support Vector Classifier (LinearSVC) with L1 regularization For feature selection and Support Vector Machine classifier (SVM) with fivefold nested cross-validation was performed using StratifiedKFold with ‘n_splits’ set to 5 to predict recurrence in NSCLC patients and assess the impact of ComBat harmonization on the outcome.ResultsFrom 749 extracted radiomic features, 206 (27%) and 389 (51%) features showed excellent reliability (ICC ≥ 0.90) against segmentation method variation before and after NPEB ComBat harmonization, respectively. Among all, 39 features demonstrated poor reliability, which declined to 10 after ComBat harmonization. The 64 fixed bin widths (without any filter) and wavelets (LLL)-based radiomic features set achieved the best performance in terms of robustness against diverse segmentation techniques before and after ComBat harmonization. The first-order and GLRLM and also first-order and NGTDM feature families showed the largest number of robust features before and after ComBat harmonization, respectively. In terms of predicting recurrence in NSCLC, our findings indicate that using ComBat harmonization can significantly enhance machine learning outcomes, particularly improving the accuracy of watershed segmentation, which initially had fewer reliable features than manual contouring. Following the application of ComBat harmonization, the majority of cases saw substantial increase in sensitivity and specificity.ConclusionRadiomic features are vulnerable to different segmentation methods. ComBat harmonization might be considered a solution to overcome the poor reliability of radiomic features.

Cross-modal de-deviation for enhancing few-shot classification

Few-shot learning poses a critical challenge due to the deviation problem caused by the scarcity of available samples. In this work, we aim to address deviations in both feature representations and prototypes. To achieve this, we propose a cross-modal de-deviation framework that leverages class semantic information to provide robust prior knowledge for the samples. This framework begins with a visual-to-semantic autoencoder trained on the labeled samples to predict semantic features for the unlabeled samples. Then, we devise a binary linear programming model to match the initial prototypes with the cluster centers of the unlabeled samples. To circumvent potential mismatches between the cluster centers and the initial prototypes, we perform the label assignment process in the semantic space by transforming the cluster centers into semantic representations and utilizing the class ground truth semantic features as reference points. Moreover, we model a linear classifier with the concatenation of the refined prototypes and the class ground truth semantic features serving as the initial weights. Then we propose a novel optimization strategy based on the alternating least squares (ALS) model. From the ALS model, we can derive two closed-form solutions regarding to the features and weights, facilitating alternative optimization of them. Extensive experiments conducted on few-shot learning benchmarks demonstrate the competitive advantages of our CMDD method over the state-of-the-art approaches, confirming its effectiveness in reducing deviation. The code is available at: https://github.com/pmhDL/CMDD.git.