Affecting almost 10 million people globally, Parkinson’s disease (PD) is identified as the second most prevalent neuro-degenerative disease by 2025, and 25.2 million people are expected to live with PD worldwide. Current diagnostic techniques often fail to detect early, as symptoms usually start to show after significant neuronal loss. To address the limitations of unimodal diagnostic approaches, we propose a novel multimodal deep learning framework that integrates spiral kinematics, acoustic features, and neuroimaging data to achieve an early stage and accurate detection of Parkinson’s Disease. The framework captures motor dysfunction through spiral drawing tests, speech impairments through acoustic features, and neurodegenerative brain changes via neuroimaging. Dedicated deep learning models, CNNs for spiral image analysis, voice data processed through RNNs, and spatial-temporal features captured via 3D CNNs for MRI data processed each modality. Combining these modalities with an attention-based fusion approach improved diagnosis performance.
The paper proposes a hybrid AI framework that combines temporal and graph measures to measure financial risk in dynamic nonstationary markets. The structure has a temporal transformer encoder, a relation graph neural network (GNN) and multi-task probabilistic prediction heads to jointly score the probability of default (PD), value at risk (VaR), conditional value at risk (CVaR), and expected losses. The system uses many kinds of information. This involves market signals, economic information, firm information, randomly generated news-based features, and specific exposure networks. Our preprocessing pipeline aligns different time series at various resolutions. We use four concept-drift handling mechanisms namely online adaptation, divergence detection, ensembles and stress simulation for augmenting. This bolsters strength as market circumstances shift. The proposed model surpasses statistical baselines including CR, deep-learning baselines such as LSTM and GNN baselines like PGNN on three datasets (1,500 global firms over crisis regimes). The architecture improves the performance of traditional models by enhancing the PD AUC by 12.6 % as well as reducing the forecast errors of VaR and CVaR by 28-50 % and generating large expected-loss improvements at the portfolio level. There are various methods for explaining GNN outputs including SHAP feature attributions, GNN edge-level interpretability, and rule-based surrogate governance models that satisfy auditability requirements. According to the results, the novel temporalgraph multi-tasking system for systematic financial risk assessment is more flexible, interpretable and accurate approaches real-world volatility and systemic interdependence in comparison to existing methods.
The classical approach to detecting depression from vision emphasizes interpretable features, such as facial expression, and classifiers such as the Support Vector Machine (SVM). With the advent of deep learning, there has been a shift in feature representations and classification approaches. Contemporary approaches use learnt features from general-purpose vision models such as VGGNet to train machine learning models. Little is known about how classical and deep approaches compare in depression detection with respect to accuracy, fairness, and generalizability, especially across contexts. To address these questions, we compared classical and deep approaches to the detection of depression in the visual modality in two different contexts: Mother-child interactions in the TPOT database and patient-clinician interviews in the Pitt database. In the former, depression was operationalized as a history of depression per the DSM and current or recent clinically significant symptoms. In the latter, all participants met initial criteria for depression per DSM, and depression was reassessed over the course of treatment. The classical approach included handcrafted features with SVM classifiers. Learnt features were turn-level embeddings from the FMAE-IAT that were combined with Multi-Layer Perceptron classifiers. The classical approach achieved higher accuracy in both contexts. It was also significantly fairer than the deep approach in the patient-clinician context. Cross-context generalizability was modest at best for both approaches, which suggests that depression may be context-specific.
In visual object recognition problems, low light exposure and low-quality images present significant challenges in navigation, surveillance, and image retrieval applications, where reliable feature detection is critical. Although recent deep learning-based image enhancement methods improve visual quality in the pixel domain, these improvements often do not translate to downstream machine vision performance, as important local gradient structures required for stable key point detection are frequently suppressed. In this work, we propose IllumiSIFT, a task-driven dark image enhancement framework that focuses on preserving Scale-Invariant Feature Transform (SIFT) key points by directly learning the Difference-of-Gaussian (DoG) pyramid from low-light image inputs. Unlike conventional pixel-level recovery approaches, the proposed method employs a cascaded residual learning architecture to predict Gaussian-blurred representations at multiple scales, enabling the generation of enhanced DoG images that are inherently aligned with the SIFT detection process. Extensive experiments conducted on the CDVS, Oxford Buildings, and Paris datasets demonstrate that the proposed approach consistently outperforms state-of-the-art enhancement methods in downstream SIFT matching performance under severe low-light conditions. These results confirm that gradient-domain, task-aligned enhancement provides a more effective and practical solution for recognition-centric low-light imaging applications.
Cyber threat is a critical challenge in today's technological era where Cloud computing plays a vital role in securing the network infrastructure against cyber threats. IDS has two key functions: detecting and stopping malicious activity, but traditional techniques still struggle a lot with unevenly distributed network traffic. In this work, a CNN-MLP hybrid model is applied to the collected dataset CICIDS-2017 to accurately detect cloud intrusions. To do the analysis, a preprocessing step has been performed with data cleaning, feature selection, one-hot encoding, class balance via SMOTE, Min-Max normalisation, etc. After this, the dataset was divided into training and testing subsets. The proposed hybrid model consists of convolutional layers for hierarchical feature extraction and multilayer perceptrons for classification, which obtained precision (98.90 %), accuracy (99.52 %), recall (99.60 %), and F1-score (99.00 %). Such outcomes proved the model's robustness and generalizability in cloud intrusion detection, which is a huge advantage for modern-day cybersecurity.
