In industry deep learning application, our manually labeled dataset has a certain number of noisy data. In the previous works, we verify that manual re-labeling the training data that model prediction do not equal to the manual label, improve the accuracy at dev dataset and human evaluation. But manual re-labeling requires a lot of human labor to correct the noisy data in the training dataset. We think that the noisy data is the training data that model prediction and manual label is not equal. To solve this problem, we propose the randomly re-setting method that automaticly improve the accuracy at dev dataset and human evaluation. The core idea is that, we randomly re-set all the noisy data to model prediction or the manual label. All the noisy data means the noisy data in training dataset and dev dataset. We use the highest accuracy of dev dataset in the all possible randomly re-setted dev datasets, to judge the quality of randomly resetted training datasets. The motivation of randomly re-setting is that the best dataset must be in all randomly possible datasets.

Feedback delay networks (FDNs) are used in audio processing and synthesis. The modal shapes of the system describe the modal excitation by input and output signals. Previously, the Ehrlich-Aberth method was used to find modes in large FDNs. Here, the method is extended to the corresponding eigenvectors indicating the modal shape. In particular, the computational complexity of the proposed analysis method does not depend on the delay-line lengths and is thus suitable for large FDNs, such as artificial reverberators. We show the relation between the compact generalized eigenvectors and the spatially extended modal shapes. We illustrate this method with an example FDN in which the suggested modal excitation control does not increase the computational cost. The modal shapes can help optimize input and output gains. This letter teaches how selecting the input and output points along the delay lines of an FDN adjusts the spectral shape of the system output.

A document by Dylan Wheeler. Click on the document to view its contents.

Emotions are considered to convey much meaning in communication. Hence, artificial methods for emotion categorization are being developed to meet the increasing demand to introduce intelligent systems, such as robots, into shared workspaces. Deep learning algorithms have demonstrated limited competency in categorizing images from posed datasets with the main features of the face being visible. However, the use of sunglasses and face masks are common in our daily lives, especially with the outbreak of communicable diseases such as the recent coronavirus. Anecdotally, partial coverings of the face reduce the effectiveness of human communication, so would this have hampering effects on computer vision, and if so, would different emotion categories be affected equally? Here, we analyze the performance of emotion classification systems when faces are partially covered with simulated sunglasses and face masks. Deep neural networks consider all pixels in an image as equally important unlike the neuroscientific findings on how humans recognize emotions. Hence, we propose a method that considers different constituent parts (e.g. mouth, eyes, and jaw) separately, giving more attention to relevant (uncovered) regions of the face. The method is compared with three standard, partial coverings-based and attention-based methods. We found that face coverings worsen emotion categorization by up to 74% for the state-of-the-art methods, whereby emotion categories are affected differently by different coverings, e.g. clear mouth coverings have little effect on categorizing happiness, but sadness is affected badly. The proposed method (on average 60.43%) has significantly improved the performance over the standard deep learning (< 46% on average), partial coverings-based (< 47% on average), as well as the attention-based (< 51% on average) methods for both the CK+, KDEF, and RAF-DB datasets when faces were partially covered with sunglasses or different face masks.

Basketball is a dynamic sport that requires precise decision-making, often based on instantaneous assessments of player intent. Understanding a player's intention to shoot or pass the ball is crucial for analyzing game strategies, predicting player movements, and enhancing automated commentary. This paper explores the application of logistic regression technique for human intent detection in basketball, specifically focusing on the task of differentiating between shooting and passing movements based on various kinematic features.Human intent detection is a vital aspect of understanding and predicting human actions in various domains, and in sports like basketball, it plays a crucial role in enhancing player performance and game strategies. In basketball, detecting the intent behind a player’s movements and actions can provide valuable insights into their tactics, strategies, and overall game plan. This ability to discern intent from motion data can significantly impact coaching, player training, and real-time game analysis.

Within few-shot learning, in-context learning (ICL) has become a potential method for leveraging contextual information to improve model performance on small amounts of data or in resource-constrained environments where training models on large datasets is prohibitive. However, the quality of the selected sample in a few shots severely limits the usefulness of ICL. The primary goal of this paper is to enhance the performance of evaluation metrics for in-context learning by selecting high-quality samples in few-shot learning scenarios. We employ the chi-square test to identify high-quality samples and compare the results with those obtained using low-quality samples. Our findings demonstrate that utilizing high-quality samples leads to improved performance with respect to all evaluated metrics.

Quantum approximate optimization algorithm (QAOA) with layer depth p is promising near-optimum performance and low complexity for NP-hard maximum-likelihood (ML) detection in n × n multi-input multi-output (MIMO) systems. Experimental challenges for ML detection on Noisy Intermediate-Scale Quantum (NISQ) computers arise from accumulated errors with large p and n. Recursive QAOA (RQAOA) is promising with small p by reducing complexity over n steps. In this article, we modify RQAOA for p n with cost sorting and post-selection in m n steps, and then integrate it with majority voting (MV) and successive interference cancellation (SIC) into the QAOA-MVSIC algorithm to tackle experimental challenges. We truncate QAOA circuits to further improve experimental feasibility. Simulations with n = 24 and 12 for BPSK and QPSK modulations, respectively, show near-optimum bit-error rate (BER) with p = 1 and m ≤ 4. Truncated version requires O(m n p) quantum and O(m n2 ) classical operations with low complexity. We experimentally implement QAOA combined with MV (QAOA-MV) for n ∈ [17, 64] in IBM Eagle processor by observing superior performance of QAOA-MV over QAOA and reducing problem dimensions by at least n/4. We generalize QAOA as cost-restricted uniform sampling (CRUS) oracle and approximately simulate for n ≤ 128 to obtain comparison benchmark for future QAOA experiments.

Antenna synthesis is becoming increasingly challenging given the tight requirements for C-SWAP (cost, size, weight and power) reduction while maintaining stringent electromagnetic performance specifications. While recent reports have investigated developments in AI (primarily deep learning) to address this challenge, such methods have limited capability for multi-objective constrained inverse design and the ability to handle large shape sets. We present a multi-objective antenna synthesis methodology employing deep neural network surrogate-driven evolutionary algorithm. Specifically, we propose a branched-output convolutional neural network architecture for creating the surrogate model and showing its ability to predict multiple electromagnetic parameters over an operating frequency range. The proposed methodology and its synthesis capabilities are showcased by considering the set of polygonal-patch antennas for operation in the 2 to 3 GHz spectral band. Specifically, we report the design, fabrication and experimental characterization of three polygon-shaped patch antennas, each fulfilling different objectives, one being a single band with resonance at 2.45 GHz, the second antenna is dual-band with resonance at 2.2 GHz and 2.9 GHz, and the third one being wide-band centered around 2.7 GHz. The reported methodology enables rapid synthesis (in seconds), produces verifiable sound designs and is promising for furthering data-driven design methodologies for electromagnetic wave device synthesis.

This paper presents a novel data-driven approach to nonlinear system control using a behavioral systems framework. A Dynamic Latent Variable Autoencoder (DLVAE) is proposed to project the nonlinear physical variable space onto a linear latent variable space. A data-predictive control approach is developed to control the physical process variables through the latent variables. Based on the behavioral systems theory, the proposed data-driven control framework does not require the knowledge on the causality of the latent variables. The stability of the controlled system is ensured by utilizing the concept of trajectory-based dissipativity. The robustness of this control approach is achieved by incorporating the Lipschitz bounds between the latent and physical variable under dissipativity conditions.

Stress is a prevalent global concern impacting individuals across various life aspects. This paper investigates stress detection using electroencephalographic (EEG) signals, which have proven valuable for studying neural correlates of stress. Stress was induced in participants, and physiological data was recorded as part of the experimental setup. Different feature sets were extracted and four machine learning models, including LightGBM, Convolutional Neural Network (CNN), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM), were utilized for classification tasks. The findings indicate that the mean and standard deviation of 19 channels consistently outperform other feature sets. LightGBM demonstrates superior performance across all scenarios compared to CNN, KNN, and SVM. Overall, this study presents an effective stress detection approach using EEG signals and demonstrates the potential of integrating simple statistical features for enhanced classification accuracy. The findings contribute to the advancement of stress monitoring technologies, with potential applications in wearables and BCIs for real-time stress management.

To solve real-world expensive constrained multi-objective optimization problems (ECMOPs), surrogate/approximation models are commonly incorporated in evolutionary algorithms to pre-select promising candidate solutions for evaluation. However, the performance of existing approaches are highly dependent on the relative position of unconstrained and constrained Pareto fronts (UPF and CPF, respectively). In addition, the uncertainty information of surrogate models is often ignored, which can misguide the search. To mitigate these key issues (among others), an efficient probabilistic selection based constrained multi-objective EA is proposed, referred to as PSCMOEA. It comprises novel elements such as (a) an adaptive search bound identification scheme based on the feasibility and convergence status of evaluated solutions (b) a probabilistic selection method backed by theoretical formulations of model mean and uncertainties to conduct search in the predicted space to identify promising solutions (c) an efficient single infill sampling approach to balance feasibility, convergence and diversity across different stages of the search and (d) an adaptive switch to unconstrained search based on certain search conditions. Numerical experiments are conducted on an extensive range of challenging constrained problems using low evaluation budgets to simulate ECMOPs. The performance of PSCMOEA is benchmarked against five competitive state-of-the-art algorithms, to demonstrate its competitive and consistent performance.

Global healthcare systems face growing pressure as populations rise. This can lead to longer wait times and an increased risk of treatment delays or misdiagnosis. Artificial intelligence (AI) diagnostic systems are being developed to address these challenges, but concerns exist about their accuracy and data security. This study introduces a robust AI telehealth system that offers a two-pronged approach. It utilizes a cutting-edge image analysis method, vision transformer, to enhance diagnostic accuracy, while also incorporating post-quantum cryptography algorithm, Kyber, to ensure patient privacy. Furthermore, an interactive visualization tool aids in interpreting the diagnostic results, providing valuable insights into the model's decisionmaking process. This translates to faster diagnoses and potentially shorter wait times for patients. Extensive testing with various datasets has demonstrated the system's effectiveness. The optimized model achieves a remarkable 95.79% accuracy rate in diagnosing COVID-19 from chest X-rays, with the entire process completed in under five seconds.

The advent of smart grid in the power infrastructure of the nations has provisioned a proactive and highly efficient power management at the generation end and user end as well. The smart meters being a fundamental part of the grid system are responsible for the intermittent sharing of usage data that further assists the utility for decision making. However, the inclusion of digital technology accompanies some major challenges like data security, data loss and communication issues etc. Smart meters are lightweight devices hence susceptible to errors frequently and therefore can cause loss of data. If the unavailability of usage data is significantly high, this can be problematic for the utility in context with further power generation and distribution decisions. This article presents a comprehensive discussion of the fault-tolerant aggregation models for the smart grid environment. A thorough investigation of the most recent research on handling of unavailable data is provided. Finally, the paper identifies the major challenges and future research possibilities for an efficiently fault-tolerant aggregation in the smart grid environment.

This study, conducted with the cooperation of the long-established Nishikawa Construction Company in Mie Prefecture, Japan, explores the possibility of transmitting centuries-old architectural traditions from human to artificial intelligence. Utilizing sensors and data collection methods to gather human behavior, speech, vital data, and decision-making processes, we propose a framework to distill the architectural philosophy and creativity of the daimyo architects of the Sengoku period into a digital format, thereby applying it to contemporary architectural practice. Specifically, we juxtapose the enduring influence of historical constructions such as Himeji Castle and Nijo Castle with the modern construction methodologies employed by Nishikawa Construction Company, focusing on the utilization of data to enhance the precision and robustness of AI. This research demonstrates that AI can be a potent tool in preserving and utilizing the knowledge and experience of past masters, resulting in a convergence of enduring architectural principles with innovative AI applications. It reveals a symbiotic existence of historical wisdom and technological advancement, suggesting the potential for modern architects and builders to make a more significant societal impact through enhanced design and construction integrity.

Airport operational efficiency is significantly enhanced by the accurate prediction of meteorological conditions. It enables safe and effective air traffic control both en-route and on the ground. Aerodrome systems demand continuous operation, which means that they regularly generate a lot of data thus verifying a forecast's value, consistency and quality is the aim of forecast verification. If a forecast accurately predicts the observed conditions based on objective or subjective criteria, it is considered high quality. A forecast's usefulness in aviation is crucial since it directly affects operational regularity, economy and safety. When the predicted value is within predetermined bounds, the forecast is validated as "correct" or otherwise "incorrect". At present, real time verification is conducted for every component of TAF, TREND forecast, takeoff forecast, area/local forecast and aerodrome warnings and light aircraft warnings. This study assesses the accuracy of the performance-based analysis of the forecasting system of the India Meteorological Department for aviation meteorology as verified with the Meteorological Aerodrome Report (METAR) in conformity with the ICAO's prescribed standards. The evaluation is done over the period from January 2018 to December 2023. The purpose of the analysis is to help duty officers of aerodrome weather forecasts by providing them with insights from previous operations.

Hyperspectral Image Change Detection (HSI-CD) focuses on identifying differences in multi-temporal HSIs. Graph Convolutional Networks (GCN) have demonstrated greater promise than convolutional neural networks (CNNs) in remote sensing, particularly for processing HSIs. This is due to GCN's ability to handle non-Euclidean graph-structured information, as opposed to the fixed kernel operations of CNN based on Euclidean structures. Specifically, GCN operates predominantly on superpixel-based nodes. This paper proposes a method, named Hyperspectral Multilevel GCN and CNN Feature Fusion for Change Detection (HMGCF-CD), that integrates superpixel-level GCN with pixel-level CNN for feature extraction and efficient change detection in HSIs. The proposed method utilizes the strengths of both CNN and GCN; the CNN branch focuses on feature learning in small-scale, regular regions, while the GCN branch handles large-scale, irregular regions. This approach generates complementary spectral-spatial features at both pixel and superpixel levels. To bridge the structural incompatibility between the Euclidean-data-oriented CNN and the non-Euclideandata-oriented GCN, HMGCF introduces a graph encoder and decoder. These elements help in propagating features between image pixels and graph nodes, allowing CNN and GCN to function within an integrated end-to-end framework. HMGCF integrates graph encoding into the network, edge weights, and node representations from training data. This innovation improves node feature learning and adapts the graph more effectively to HSI content. Fusion of features from different branches within the HMGCF enables precise changes in HSI. Experiments carried out on four datasets demonstrate that the proposed HMGCF method outperforms existing methods, marking a significant advancement in HSI-CD techniques.

A document by Swaroop Gopalam . Click on the document to view its contents.

This paper presents the design and development of a low-cost ventilator intended for emergency situations and resource-constrained environments. An overview of various research is done on how human lungs are used for respiration to understand the working principle of the ventilator. The ventilator is engineered to deliver breaths within a range of 5 to 30 per minute through a pulley and lever mechanism driven by an electromagnetic stepper motor, drawing power from a 12V supply. Offering flexibility, the device provides a tidal volume output ranging from 400ml to 1000ml and an adjustable inspiration to expiration ratio of 1:3, which can be tailored to meet individual patient requirements. To enhance usability, the ventilator is equipped with an LCD screen user interface that displays the set tidal volume, breaths per minute, and inspiration to expiration time ratio. Additionally, the system is equipped with two pressure sensors integrated into the patient connector, enabling precise measurements of flow rate and pressure changes in the oxygen delivery tube. Ensuring reliability, the device has been carefully calibrated to maintain a 10% accuracy rate in delivering the set output. This prototype showcases the potential of automated Ambu bag compression, offering a cost-effective and portable ventilator system that delivers essential ventilation features, rivaling existing technology at a significantly reduced cost.