Informatica

Credit card fraud (CCF) poses a growing threat to the global economy, with payment card fraud losses reaching approximately USD 33.8 billion in 2023. While deep learning approaches such as CNNs, autoencoders, and Transformer-based models have improved detection accuracy, their “black box” nature restricts adoption in high-stakes financial domains where transparent and auditable reasoning is essential for trust and regulatory compliance. To address this challenge, we propose MTLFraudDetect, a multi-task framework that combines fraud detection with complementary model-analysis techniques. The model combines fraud classification with two auxiliary tasks—data reconstruction and amount prediction—while employing curriculum learning and PCGrad to stabilize multi-task optimization. For model analysis, the framework combines reconstruction-based anomaly-sensitivity signals with SHAP-based decision attribution. Reconstruction errors indicate transaction-level and feature-wise deviations from patterns learned by the reconstruction branch, whereas SHAP examines how the classifier’s shared latent representation and transaction amount contribute to its final prediction. Under an explicitly separated development-test protocol with training-fold-only preprocessing and out-of-fold hyperparameter selection, MTLFraudDetect achieves competitive F1-score and AUC-PR performance while providing complementary evidence at the levels of global anomaly detection, feature-wise error localization, and post-hoc decision attribution. These results do not by themselves establish formal explanatory faithfulness or complete interpretability.

Focusing on the problems of failing to make full use of spatial context information and limited local receptive field when U-Net is utilized to solve MRI brain tumour segmentation, a novel 3D multi-scale attention U-Net method, i.e. MAU-Net, is proposed in this paper. Firstly, a Mixed Depth-wise Convolution (MDConv) module is introduced in the encoder and decoder, which leverages various convolution kernels to extract the multi-scale features of brain tumour images, and effectively strengthens the feature expression of the brain tumour lesion region in the up and down sampling. Secondly, a Context Pyramid Module (CPM) combining multi-scale and attention is embedded in the skip connection position to achieve the combination of local feature enhancement at multi-scale with global feature correlation. Finally, MAU-Net adopts Self-ensemble in the decoding process to achieve complementary detailed features of sampled brain tumour images at different scales, thereby further improving segmentation performance. Ablation and comparison experiment results on the publicly available BraTS 2019/2020 datasets well validate its effectiveness. It respectively achieves the Dice Similarity Coefficients (DSC) of 90.6%/90.2%, 82.7%/82.8%, and 77.9%/78.5% on the whole tumour (WT), tumour core (TC) and enhanced tumour (ET) segmentation. Additionally, on the BraTS 2021 training set, the DSC for WT, TC, and ET reached 93.7%, 93.2%, and 88.9%, respectively.

We attempted to determine the most common localizations of epileptogenic foci by using common functional (EEG and PET/CT) and structural (MRI) imaging methods. Also, we compared the number of epileptogenic foci detected with all diagnostic methods and determined the success rate of surgery in the operated patients when the epileptogenic foci coincided on all three imaging methods. 35 patients (including children) with clinically proven refractory epilepsy were included into the study. All patients underwent an MRI scan with epilepsy protocol, Fluorodeoxyglucose-18-PET/CT scan, and an EEG prior to a PET study. 14 patients underwent neurosurgery for removal of epileptogenic foci. We found a statistically significant difference between the number of epileptogenic foci which were found in PET/CT and EEG studies but there was no significant difference between MRI and PET/CT lesion numbers. The most common localization of epileptogenic activity on EEG was right temporal lobe (54.3%); the most common lobe with structural changes on MRI was right temporal lobe (42.9%); the most common hypometabolism zone on PET/CT was in right temporal lobe (45.7%). 10 out of 14 patients who underwent surgery demonstrated excellent postsurgical outcomes, with no epileptic seizures one year or more after the operation; 3/14 patients had 1–2 seizures after surgery and one patient had the same count or more epileptic seizures in duration of one year or more. The measure of Agreement Kappa between PET/CT and EEG value was 0.613 $(p<0.05)$. Between PET/CT and MRI the value was 0.035 $(p>0.05)$. Surgical treatment may offer hope for patients with intractable epileptic seizures. PET/CT was an extremely useful imaging method to assist in the localization of epileptogenic zones. The dynamic functional information that brain PET/CT provides is complementary to anatomical imaging of MRI and functional information of EEG.

Identification problems of linear dynamic systems in the class of parametric mathematical models are considered. A method of calculating of guaranteed estimates of indefinite parameters is proposed. The method is based on the specific semiinfinite extremal problems solution.

The observation problem along with the certain independent value plays a great role while carrying out control of dynamic systems in the conditions of uncertainty (Kalman, 1957, Krasovski, 1985, Leondes, 1976). A new approach on connection between the problems of control and observation is presented in (Gabasov, 1991). Developing it, we justify the solution of observation problem in the given paper that arises, at optimization of linear dynamic systems. The paper consists of the two parts. In the part I the linear discrete system is investigated. In the part II the linear dynamic continuous system is considered.