The high-resolution seismic method can provide acoustic reflectivity images of shallow marine geology structures. In South Vietnam, the demand for construction materials like sand is high; therefore, the exploration of its deposits is considered in this study. This study investigated an around 200-square-kilometer area offshore Bac Lieu using 2D seismic sub-bottom lines. We employed the processed seismic amplitude and its seismic attributes as mean and variance textures to interpret the data. The processed seismic amplitude and its attributes can represent the young Holocene sediments (i.e., sand, silt, clay, and their mixtures) thanks to their different seismic patterns. Our interpretation result consists of 3D horizons of the seabed, Holocene silt, and sand sediments, which are compatible with the prior geological information, including three nearby drill holes. The seabed gradually descends from 10.0 m to 19.0 m over a horizontal distance of around 11 km. Moreover, the interpreted results show that the sand sediments reside in the center of survey area, with a maximum thickness of around 12.0 m. Interestingly, a fill sediment channel effectively separates two different zones of young Holocene sand and silt sediments. The findings provide valuable information for Vietnamese government officers to develop sustainable policies and regulations for marine mineral exploitation and exploration.

Modern deep neural network training is based on mini-batch stochastic gradient optimization. While using extensive mini-batches improves the computational parallelism, the small batch training proved that it delivers improved generalization performance and allows a significantly smaller memory, which might also improve machine throughput. However, mini-batch size and characteristics, a key factor for training deep neural networks, has not been sufficiently investigated in training correlated group features and looping with highly complex ones. In addition, the unsupervised learning method clusters the data into different groups with similar properties to make the training process more stable and faster. Then, the supervised learning algorithm was applied with the cluster repeated mini-batch training (CRMT) methods. The CRMT algorithm changed the random minibatch characteristics in the training step into training in order of clusters. Specifically, the self-organizing maps (SOM) were used to cluster the information into n groups based on the dataset's labels Then, neural network models (ANN) were trained with each cluster using the cluster repeated mini-batch training method. Experiments conducted on EEG datasets demonstrate the survey of the proposed method and optimize it. In addition, the results in our research outperform other state-of-the-art methods.

Objective: Advanced deep-learning techniques have been employed to enhance the field of brain-computer interfaces through the processing of electroencephalogram (EEG) signals. Recent studies have shown application of convolutional neural networks (CNNs) with manually extracted EEG features to be a promising approach to improve classification accuracy. However, these models require complex input features to generate output decisions. This study aimed to improve EEG recognition by developing novel CNN architectures based on branch CNNs: the step CNN (the base model) and the input- and layer-additive CNNs, which incorporate connected layers. The models exhibit exceptional performance across established WAY-EEG-GAL benchmark datasets. Additionally, the proposed method was evaluated and optimized through EEG dataset experiments, and surpassed other state-of-the-art techniques. Using connected layers in the models increased the effectiveness in recognizing EEG signals during grab-and-lift classification.