Games are considered a suitable and standard benchmark for checking the performance of artificial intelligence-based algorithms in terms of training, evaluating, and comparing the performance of AI agents. In this research, an application of the Intrinsic Curiosity Module (ICM) and the Asynchronous Advantage Actor–Critic (A3C) algorithm is explored using action games. Having been proven successful in several gaming environments, its effectiveness in action games is rarely explored. Providing efficient learning and adaptation facilities, this research aims to assess whether integrating ICM with A3C promotes curiosity-driven explorations and adaptive learning in action games. Using the MAME Toolkit library, we interface with the game environments, preprocess game screens to focus on relevant visual elements, and create diverse game episodes for training. The A3C policy is optimized using the Proximal Policy Optimization (PPO) algorithm with tuned hyperparameters. Comparisons are made with baseline methods, including vanilla A3C, ICM with pixel-based predictions, and state-of-the-art exploration techniques. Additionally, we evaluate the agent’s generalization capability in separate environments. The results demonstrate that ICM and A3C effectively promote curiosity-driven exploration in action games, with the agent learning exploration behaviors without relying solely on external rewards. Notably, we also observed an improved efficiency and learning speed compared to baseline approaches. This research contributes to curiosity-driven exploration in reinforcement learning-based virtual environments and provides insights into the exploration of complex action games. Successfully applying ICM and A3C in action games presents exciting opportunities for adaptive learning and efficient exploration in challenging real-world environments.

Automated seizure detection from long-term scalp electroencephalography (EEG) remains challenging because seizure windows are sparse, channel configurations vary across patients, and clinically useful systems must maintain strict control of false alarms. This study presents a patient-independent hybrid generative-discriminative framework evaluated on the CHB-MIT cohort under leave-one-patient-out cross-validation. In each outer fold, one patient was reserved exclusively for final testing, whereas channel harmonization, representation learning, feature refinement, model training, and operating-point calibration were performed using only the remaining patients. A fold-specific convolutional β-variational autoencoder ([Formula: see text]) was trained on interictal windows from the training patients only. For each window, latent statistics and reconstruction error were combined with handcrafted seizure-sensitive descriptors, followed by train-only sanitization and supervised feature selection. Candidate boosted-tree classifiers, namely LightGBM, CatBoost, and XGBoost, were trained within the training pool, and operating parameters were selected on validation data only. Final performance was then computed on the held-out patient using window-level metrics with record-wise temporal postprocessing. Across the 24 held-out evaluation cases, corresponding to 23 distinct individuals in the public CHB-MIT release, the framework achieved mean accuracy of 0.9964, mean balanced accuracy of 0.9411, mean specificity of 0.9967, and mean false alarms per hour of 4.7154. Across seizure-positive held-out cases, the mean sensitivity, F1-score, ROC-AUC, and average precision were 0.5505, 0.3543, 0.9496, and 0.2990, respectively. These findings show that leakage-safe hybrid representation learning can support seizure detection while explicitly handling heterogeneous channels, severe class imbalance, and false-alarm control in long-term scalp EEG.

The center of pressure (CoP) is a key biomechanical indicator for assessing balance and fall risk; however, force plates, the gold standard for CoP measurement, are costly and impractical for widespread use. Low-cost alternatives such as inertial units or pressure sensors are limited by drift, sparse sensor coverage, and directional performance imbalances, with previous supervised learning approaches reporting ML-AP NRMSE differences of 3.2–4.7% using 1D time-series models on sparse sensor arrays. Therefore, we propose a tactile sensor-based CoP estimation system using deep learning models that can extract 2D spatial features from each pressure distribution image with CNN/ResNet encoders followed by a Bi-LSTM for temporal patterns. Using data from 23 healthy adults performing four balance protocols, we compared ResNet-Bi-LSTM and CNN-Bi-LSTM with baseline CNN-LSTM and Bi-LSTM models used in previous studies. Model performance was validated using leave-one-out cross-validation (LOOCV) and evaluated with RMSE, NRMSE, and R². The ResNet-Bi-LSTM with angular features achieved the best performance, with RMSE values of 18.63 ± 4.57 mm in the mediolateral (ML) direction and 17.65 ± 3.48 mm in the anteroposterior (AP) direction, while reducing the ML/AP NRMSE difference to 1.3% compared to 3.2–4.7% in previous studies. Under dynamic protocols, ResNet-Bi-LSTM maintained the lowest RMSE across models. These findings suggest that tactile sensor-based systems may provide a cost-effective alternative to force plates and hold potential for applications in gait analysis and real-time balance monitoring. Future work will validate clinical applicability in patient populations and explore real-time implementation.

Peen forming is a dieless manufacturing process used to shape large, thin aerospace panels through controlled shot impacts that induce residual stresses and curvature. Despite long-standing industrial use, process monitoring still depends largely on indirect proxies such as Almen intensity and coverage, limiting spatially resolved deformation assessment and hindering closed-loop control. In parallel, vision-based artificial intelligence (AI) has enabled adaptive monitoring and feedback in smart-manufacturing domains such as welding, additive manufacturing, and sheet forming. This review examines how such sensing and learning strategies can be transferred to adaptive peening forming. We compare six vision sensing modalities and assess major AI model families for surface mapping, temporal prediction, robustness, and deployment maturity. The synthesis shows that progress is primarily constrained by limited validated datasets, harsh in-cabinet sensing conditions, scarce closed-loop demonstrations, and weak validation on curved aerospace geometries. We conclude that the sensing and AI foundations for adaptive peen forming are already emerging, but industrial translation now depends on stronger experimental validation, standardized benchmarking, robust multi-sensor integration, and edge-capable feedback pipelines.