We present a fast and robust algorithm for computing the minimum distance from a point in to a toroidal patch. Our approach replaces iterative numerical optimization with a topologically complete analytic classification that exhaustively enumerates all configurations on a toroidal patch. By reducing the problem to planar arc tests on circles obtained through simple radial normalization, the method identifies 4 regular regimes, which extend to 16 via Boolean combinations of mirrored-domain inclusion bits when the patch’s parameter domain overlaps the self-intersection interval. Binary decisions on these bits determine the nearest feature (surface, boundary curve, or corner) at constant cost and without trigonometric evaluations at query time. Each query is processed in about 2 microseconds on average, achieving over 99.99% agreement with a high-precision reference within a tolerance of , while significantly outperforming global optimizers. The algorithm also robustly handles patches derived from spindle tori with self-intersections, providing a consistent and efficient building block for geometry processing and rendering tasks, such as path tracing of surfaces of revolution represented as collections of toroidal patches. • Non-iterative algorithm for point–toroidal patch minimum distance computation. • Problem reformulated via topological relations between torus arcs and the point. • Closed-form solutions per topological case, enabling robust, efficient evaluation. • Demonstrates robust and efficient performance in large-scale rendering experiments.

Model-agnostic explanation methods are essential for interpreting machine learning models, but suffer from prohibitive computational costs that scale with the number of baselines. Existing acceleration approaches either lack a theoretical base or provide no principled guidance for baseline selection. To address this gap, we present ABSQR (Amortized Baseline Selection via Rank-Revealing QR). This framework exploits the low-rank structure of value matrices to accelerate multi-baseline attribution methods. Our approach combines deterministic baseline selection via SVD-guided QR decomposition with an amortized inference mechanism that utilizes cluster-based retrieval. We reduce computational complexity from O (m • 2d) to O (k • 2d), where k ≪ m. Experiments demonstrate that ABSQR achieves a 91.2% agreement rate with full baseline methods while providing 8.5× speedup across diverse datasets. As the first acceleration approach that preserves explanation error guarantees under computational speedup, ABSQR makes the practical deployment of interpretable AI systems feasible at scale.

Active noise control (ANC) algorithms have been developed within the adaptive algorithm framework. However, multichannel ANC systems, which include numerous reference sensors, control speakers, and error microphones, require a very long control filter converging time for control filter estimation. Traditional system identification methods, such as the Wiener filter method, are better suited for such systems because of their relatively shorter converging time. However, they require large amounts of data to achieve accurate statistical estimation. Therefore, this article proposes a control filter estimation method that requires only a short length of data. An iterative Wiener filter solution using Kronecker product decomposition for multichannel ANC systems converts the filter estimation process by breaking down the extensive control filter into multiple shorter control filters through Kronecker product decomposition. This decomposition effectively reduces the high‐dimensional system identification problem into manageable low‐dimensional ones. Numerical simulations demonstrate the superiority of the proposed method over conventional Wiener filter techniques, especially in scenarios when limited data are available for control filter estimation.

Long-range horizontal connections (LRCs) are conspicuous anatomical structures in the primary visual cortex (V1) of mammals, yet their detailed functions in relation to visual processing are not fully understood. Here, we show that LRCs are key components to organize a "small-world network" optimized for each size of the visual cortex, enabling the cost-efficient integration of visual information. Using computational simulations of a biologically inspired model neural network, we found that sparse LRCs added to networks, combined with dense local connections, compose a small-world network and significantly enhance image classification performance. We confirmed that the performance of the network appeared to be strongly correlated with the small-world coefficient of the model network under various conditions. Our theoretical model demonstrates that the amount of LRCs to build a small-world network depends on each size of cortex and that LRCs are beneficial only when the size of the network exceeds a certain threshold. Our model simulation of various sizes of cortices validates this prediction and provides an explanation of the species-specific existence of LRCs in animal data. Our results provide insight into a biological strategy of the brain to balance functional performance and resource cost.

The brain successfully performs visual object recognition with a limited number of hierarchical networks that are much shallower than artificial deep neural networks (DNNs) that perform similar tasks. Here, we show that long-range horizontal connections (LRCs), often observed in the visual cortex of mammalian species, enable such a cost-efficient visual object recognition in shallow neural networks. Using simulations of a model hierarchical network with convergent feedforward connections and LRCs, we found that the addition of LRCs to the shallow feedforward network significantly enhances the performance of networks for image classification, to a degree that is comparable to much deeper networks. We found that a combination of sparse LRCs and dense local connections dramatically increases performance per wiring cost. From network pruning with gradient-based optimization, we also confirmed that LRCs could emerge spontaneously by minimizing the total connection length while maintaining performance. Ablation of emerged LRCs led to a significant reduction of classification performance, which implies these LRCs are crucial for performing image classification. Taken together, our findings suggest a brain-inspired strategy for constructing a cost-efficient network architecture to implement parsimonious object recognition under physical constraints such as shallow hierarchical depth.

The serial-position effect in working memory is considered important for studying how a sequence of sensory information can be retained and manipulated simultaneously in neural memory circuits. Here, via a precise analysis of the primacy and recency effects in human psychophysical experiments, we propose that stable and flexible codings take distinct roles of retaining and updating information in working memory, and that their combination induces serial-position effects spontaneously. We found that stable encoding retains memory to induce the primacy effect, while flexible encoding used for learning new inputs induces the recency effect. A model simulation based on human data, confirmed that a neural network with both flexible and stable synapses could reproduce the major characteristics of serial-position effects. Our new prediction, that the control of resource allocation by flexible-stable coding balance can modulate memory performance in sequence-specific manner, was supported by pre-cued memory performance data in humans.

Active noise control (ANC) is a technology which lowers the noise level by using the principle of destructive interference of sound wave. Even though recent developments in digital signal processing (DSP) made it possible to implement ANC algorithms in real-time, insufficient computational power is still one of the challenges to solve. In the previous research, as a way of overcoming the lack of computational power, CPU-GPU architecture was proposed so that ANC algorithms utilize the massive computing power of GPU without suffering from the block data transfer between CPU and GPU memories. However, for the feasibility test of the proposed CPU-GPU architecture in the previous research, a conventional block ANC algorithm was used, and ANC algorithm which can fully utilize the massive computing power of GPU has not been developed. In this article, ANC algorithm, which directly derives blockwise least square solution through GPU computation while generating control signals through CPU computation, is proposed. Based on the observation about speaker saturation and increase of noise level after applying the conventional blockwise least square solution, a new cost function for preventing such problems is also proposed. Therefore, blockwise weighted least square ANC (BWLS-ANC) algorithm, which derives blockwise least square solution minimizing the proposed cost function through GPU computation while generating control signals through CPU computation, is proposed throughout this research. Problems of conventional blockwise least square solution upon ANC applications are observed through simulations and experiments. The feasibility of the proposed BWLS-ANC algorithm is verified through experiments.

This study analyzed the ongoing drought situation in South Korea using MODIS satellite data and AWS ground observation data. In order to analyze the drought, the Scaled Drought Condition Index (SDCI) was calculated utilizing concurrent precipitation, temperature and vegetation factors. Unlike conventional research that utilizes only satellite data, we utilized AWS ground-based observation data that has advantages in terms of data accuracy and collection stability. Precipitation and temperature inputs were sourced from AWS ground observation data while vegetation coverage was obtained from MODIS NDVI data. SDCI was applied to data on June 5, 2012 when the drought was severe. The results of the SDCI were compared with high-land vegetable harvest analysis, and showed a positive mutual correlation.

Satellite-derived rainfall can be a potential source for flood detection and monitoring in ungauged watersheds. This study investigated the hydrological utility of satellite-derived rainfall through rainfall-runoff modeling. The results showed that the COMS satellite rainfall led to large underestimation in runoff discharge indicating that a preprocessing scheme such as bias correction or adjustment prior to use of hydrological applications is required.

This study improves on the simple Landsat simulation imagery algorithm for target date and in-situ meteorological data. In order to improve the algorithm, a case-based reasoning (CBR) and a function of mask (Fmask) algorithm are applied. CBR is used to select the reference imagery, and Fmask is used to remove cloud and cloud shadows in each images. In similarity analyses of visual aspects, the improved simulation algorithm shows higher accuracy. In other words, the overall color in the simulated imagery was similar to the target imagery. In the quantitative analysis of similarity using the spectral correlation mapper (SCM) method, the percentage of pixels with correlation coefficients greater than 0.99 are about 66% for the improved simulation algorithm, about 16% higher than the results of the La et al. (2015) method.

Healthcare spending has been increasing in the last few decades. One of the main reasons for this increase is hospital readmissions, which is defined as a re-hospitalization of a patient after being discharged from a hospital within a short period of time. The excessive amount of money spent every year on hospital readmissions and the urge to enhance healthcare quality make reducing hospital readmissions a necessity. In this paper, we extract knowledge from a medical dataset and apply the concept of mining actionable rules to guide the health domain experts in their decision-making process. We present novel algorithms to increase the predictability of the patients' paths (the sequence of procedures that patients undertakes to reach a desired treatment) by clustering the patients according to their set of diagnoses. Moreover, we present a scoring metric to evaluate procedures in procedure graphs (the tree of all possible procedure paths) and a scoring metric to evaluate clusters of diagnoses which would allow us to anticipate the number of following readmissions for a new patient. Finally, we present an algorithm to evaluate the score (average number of following readmissions) for new patients prior to applying the action rules and after. The results presented in this paper show that our algorithms are able to reduce the average number of readmissions to a high degree.

Image super-resolution typically requires prior knowledge to overcome its under-determinedness. While recent generic prior models give impressive results, we note some common artifacts: blurry edges, jagged edges, staircase surfaces, and inconsistent colors. With understanding the underlying causes, we propose an improved generic prior model using isotropic Huber MRFs. In addition, to avoid the oversimplifying problem that generic models often encounter, we suggest applying them only to the structural component of an image, preserving the textural component. Experiments confirm that our framework successfully eliminates those artifacts while preserving textures. Showing competitive results, the GPU implementation allows near real-time application on video.

This paper applied an adaptive neuro-fuzzy inference system (ANFIS) based on a geographic information system (GIS) environment using landslide-related factors and location for landslide susceptibility mapping. Landslide-related factors such as slope, soil texture, wood type, lithology and density of lineament were extracted from topographic, soil, forest and lineament maps. Landslide locations were identified from interpretation of aerial photographs and field surveys. Landslide-susceptible areas were analyzed by the ANFIS method and mapped using occurrence factors. In particular, we applied various membership functions (MFs), and analysis results were verified by using the landslide location data. The predictive maps using triangular, trapezoidal, and polynomial MFs were the best individual MFs for modeling landslide susceptibility maps (84.96% accuracy), proving that ANFIS could be very effective in modeling landslide susceptibility mapping.

In this paper, we show the characteristics of the threshold-based IR-UWB location positioning system by experiment. We found that this system provides distance-dependent accuracy because the arrival time of the pulse is detected by threshold using 1-bit ADC (Analog to Digital Convertor). We also propose a simple compensation method based on measurement to improve positioning accuracy. The proposed method improves positioning error down to 52.7%.

We propose a selective detection scheme based on pulse repetition considering Bit Error Rate (BER) performance and complexity of noncoherent Ultra Wide Band (UWB) systems. To consider system complexity, the proposed scheme transmits the UWB signals by pulse repetition at the transmitter, like the conventional Pulse Repetition Coding (PRC). However, to effectively improve BER performance of the system, the proposed scheme performs selective detection by estimating the Signal-to-Noise Ratio (SNR) of the received pulse-repeated signal at the UWB receiver. Hence, the proposed scheme effectively improves BER performance of the noncoherent UWB systems without increase of the system complexity, as compared to the conventional PRC algorithm.

Many different methods for sound source localization have been developed. Most of them mainly depend on time difference of arrival (TDOA) or analytic head related transfer functions (HRTF). In real implementation, since the direct path between a source and a sensor is interrupted by obstacles as like a head or body of robot, it has to be considered the number of sensors as well as their positions. Therefore, in this paper, we present the methods, which are included sensor position problem, to localize the sound source with 4 microphones to cover the 3D space. Those are modified two-step TDOA methods. Our conclusion is that the different method has to be applied in case to be different microphone position on real robot platform.

We propose a novel design of artificial robot ear for sound direction estimation using measured two outputs only. The spectral features in head-related transfer functions and in interaural transfer functions (ITFs) are distinctive in the voice frequency band. Thus, these features provide effective sound cues to estimate the sound direction using two measured ear outputs only without input information. Especially, the direction of median source can be estimated by the bilateral asymmetry of microphone positions, whereas many previous studies, which depend on the interaural differences, cannot achieve. We propose a localization method to estimate the lateral and vertical angles simultaneously. The lateral angle is estimated using interaural time delay and Woodworth and Schlosberg's formula, and the front-back discrimination is achieved by the presence of the spectral features in ITF estimated from two measured outputs. The vertical angle of the frontal source is estimated by comparing the spectral features in the estimated ITF with those in the database built in an anechoic chamber.

This paper presents the KAIST interactive bicycle racing simulator system, which consists of a pair of bicycle simulators. The rider on the racing simulator experiences realistic sensations of motion, while being able to see the other bicycle simulator and having the audio-visual experience of riding in a velodrome or on the KAIST campus. The 2nd bicycle of the racing simulator system consists of a bicycle, a 4-DOF platform, a handlebar and a pedal resistance system to generate motion feelings; a real-time visual simulator a HMD and beam projection system; and a 3D sound system. The system has an integrating control network with an AOIM (Area Of Interest Management) based network structure for multiple simulators.

This paper presents key technologies and system integration issues of the KAIST interactive bicycle simulator. The rider on the bicycle feels the motion and has the visual experience as if he/she is riding in the campus of the Korea Advanced Institute of Science and Technology. The simulator consists of a bicycle, a Stewart platform, a magnetorheological handle, a pedal resistance system to generate motion feelings, a real-time visual simulator and projection system, sub-controllers and an integrating control network.

A robust controller design method is proposed to obtain a less conservative feedback controller, and applied to a single-link flexible manipulator. The objective is to maximize the control performance guaranteeing the robust stability when regulating the tip position of the flexible manipulator in the presence of a large time varying payload and parameter uncertainties such as stiffness, joint friction. Descriptor form representation, which allows separate treatment of payload uncertainty from other parametric uncertainties, is used to reduce the conservatism of the conventional robust control approaches. Uncertainty of the payload included in the left-hand side inertia matrix and the uncertain parameters included in the right-hand side damping and stiffness matrices are treated with polytopic and descaling techniques, respectively. Using the aforementioned techniques, the robust LQ controller design problem for a flexible manipulator is formulated based on the guaranteed cost approach. Then, the formulated problem has been solved by LMIs.