Stefano Markidis
Stefano Markidis
Institution:
Email:
<jats:p>Physics-Informed Neural Networks (PINN) emerged as a powerful tool for solving scientific computing problems, ranging from the solution of Partial Differential Equations to data assimilation tasks. One of the advantages of using PINN is to leverage the usage of Machine Learning computational...
More
<jats:p>Physics-Informed Neural Networks (PINN) emerged as a powerful tool for solving scientific computing problems, ranging from the solution of Partial Differential Equations to data assimilation tasks. One of the advantages of using PINN is to leverage the usage of Machine Learning computational frameworks relying on the combined usage of CPUs and co-processors, such as accelerators, to achieve maximum performance. This work investigates the design, implementation, and performance of PINNs, using the Quantum Processing Unit (QPU) co-processor. We design a simple Quantum PINN to solve the one-dimensional Poisson problem using a Continuous Variable (CV) quantum computing framework. We discuss the impact of different optimizers, PINN residual formulation, and quantum neural network depth on the quantum PINN accuracy. We show that the optimizer exploration of the training landscape in the case of quantum PINN is not as effective as in classical PINN, and basic Stochastic Gradient Descent (SGD) optimizers outperform adaptive and high-order optimizers. Finally, we highlight the difference in methods and algorithms between quantum and classical PINNs and outline future research challenges for quantum PINN development.</jats:p>
Less
2 days ago
Tiffany Jiang
Tiffany Jiang
Institution:
Email:
An unprecedented amount of access to data, “big data (or high dimensional data),” cloud computing, and innovative technology have increased applications of artificial intelligence in finance and numerous other industries. Machine learning is used in process automation, security, underwriting and...
More
An unprecedented amount of access to data, “big data (or high dimensional data),” cloud computing, and innovative technology have increased applications of artificial intelligence in finance and numerous other industries. Machine learning is used in process automation, security, underwriting and credit scoring, algorithmic trading and robo-advisory. In fact, machine learning AI applications are purported to save banks an estimated $447 billion by 2023. Given the advantages that AI brings to finance, we focused on applying supervised machine learning to an investment problem. 10-K SEC filings are routinely used by investors to determine the worth and status of a company–Warren Buffett is frequently cited to read a 10-K a day. We sought to answer–“Can machine learning analyze more than thousands of companies and spot patterns? Can machine learning automate the process of human analysis in predicting whether a company is fit to merge? Can machine learning spot something that humans cannot?” In the advent of rising antitrust discussion of growing market concentrations and the concern for decrease in competition, we analyzed merger activity using text as a data set. Merger activity has been traditionally hard to predict in the past. We took advantage of the large amount of publicly available filings through the Securities Exchange Commission that give a comprehensive summary of a company, and used text, and an innovative way to analyze a company. In order to verify existing theory and measure harder to observe variables, we look to use a text document and examined a firm’s 10-K SEC filing. To minimize over-fitting, the L2 LASSO regularization technique is used. We came up with a model that has 85% accuracy compared to a 35% accuracy using the “bag-of-words” method to predict a company’s likelihood of merging from words alone on the same period’s test data set. These steps are the beginnings of tackling more complicated questions, such as “Which section or topic of words is the most predictive?” and “What is the difference between being acquired and acquiring?” Using product descriptions to characterize mergers further into horizontal and vertical mergers could eventually assist with the causal estimates that are of interest to economists. More importantly, using language and words to categorize companies could be useful in predicting counterfactual scenarios and answering policy questions, and could have different applications ranging from detecting fraud to better trading.
Less
2 days ago
Thomas C Erren
Thomas C Erren
Institution:
Email:
Medical advice is key to the relationship between doctor and patient. The question I will address is “how may chatbots affect the interaction between patients and doctors in regards to medical advice?” I describe what lies ahead when using chatbots and identify questions galore for the daily wo...
More
Medical advice is key to the relationship between doctor and patient. The question I will address is “how may chatbots affect the interaction between patients and doctors in regards to medical advice?” I describe what lies ahead when using chatbots and identify questions galore for the daily work of doctors. I conclude with a gloomy outlook, expectations for the urgently needed ethical discourse, and a hope in relation to humans and machines.
Less
1 week ago
Youngsam Chun,
Youngsam Chun
Institution:
Email:
Jisoo Hur,
Jisoo Hur
Institution:
Email:
Junseok Hwang
Junseok Hwang
Institution:
Email:
This study investigates the factors influencing specialization in artificial intelligence (AI) technology, a critical element of national competitiveness. We utilized a revealed comparative advantage matrix to evaluate technological specialization across countries and employed a three-way fixed-effe...
More
This study investigates the factors influencing specialization in artificial intelligence (AI) technology, a critical element of national competitiveness. We utilized a revealed comparative advantage matrix to evaluate technological specialization across countries and employed a three-way fixed-effect panel logit model to examine the relationship between AI specialization and its determinants. The results indicate that the development of AI technology is strongly contingent on a nation’s pre-existing technological capabilities, which significantly affect AI specialization in emerging domains. Additionally, this study reveals that scientific knowledge has a positive impact on technological specialization, highlighting the necessity of integrating scientific advancements with technological sectors. Although complex technologies positively influence AI specialization, their effect is less pronounced than that of scientific knowledge. This suggests that in rapidly advancing fields, such as AI, incorporating new scientific knowledge into related industries may be more advantageous than simply advancing existing technologies to outpace competitors. This insight points nations toward enhancing AI competitiveness in new areas, emphasizing the vital importance of both scientific and technological capabilities, and the integration of novel AI knowledge with established sectors. This research offers critical guidance for policymakers in less technologically and economically developed countries, as these nations may not have the technological infrastructure required to foster AI specialization through increased technical complexity.
Less
1 week ago
Chen Liu
Chen Liu
Institution:
Email:
More
In this study, we used unidirectional and bidirectional long short-term memory (LSTM) deep learning networks for Chinese news classification and characterized the effects of contextual information on text classification, achieving a high level of accuracy. A Chinese glossary was created using jieba—a word segmentation tool—stop-word removal, and word frequency analysis. Next, word2vec was used to map the processed words into word vectors, creating a convenient lookup table for word vectors that could be used as feature inputs for the LSTM model. A bidirectional LSTM (BiLSTM) network was used for feature extraction from word vectors to facilitate the transfer of information in both the backward and forward directions to the hidden layer. Subsequently, an LSTM network was used to perform feature integration on all the outputs of the BiLSTM network, with the output from the last layer of the LSTM being treated as the mapping of the text into a feature vector. The output feature vectors were then connected to a fully connected layer to construct a feature classifier using the integrated features, finally classifying the news articles. The hyperparameters of the model were optimized based on the loss between the true and predicted values using the adaptive moment estimation (Adam) optimizer. Additionally, multiple dropout layers were added to the model to reduce overfitting. As text classification models for Chinese news articles, the Bi-LSTM and unidirectional LSTM models obtained f1-scores of 94.15% and 93.16%, respectively, with the former outperforming the latter in terms of feature extraction.
Less
1 week ago
Houda Chakib,
Houda Chakib
Institution: Data4Earth Laboratory, Faculty of Sciences and Technics
Email: houda.chakib@yahoo.fr
Najlae Idrissi,
Najlae Idrissi
Institution: 1Data4Earth Laboratory, Faculty of Sciences and Technics
Email: n.idrissi@usms.ma
Oussama Jannani
Oussama Jannani
Institution: Data4Earth Laboratory, Faculty of Sciences and Technics
Email: o.jannani@gmail.com
In recent years, image compression techniques have received a lot of attention from researchers as the number of images at hand keep growing. Digital Wavelet Transform is one of them that has been utilized in a wide range of applications and has shown its efficiency in image compression field. Moreo...
More
In recent years, image compression techniques have received a lot of attention from researchers as the number of images at hand keep growing. Digital Wavelet Transform is one of them that has been utilized in a wide range of applications and has shown its efficiency in image compression field. Moreover, used with other various approaches, this compression technique has proven its ability to compress images at high compression ratios while maintaining good visual image quality. Indeed, works presented in this paper deal with mixture between Deep Learning algorithms and Wavelets Transformation approach that we implement in different color spaces. In fact, we investigate RGB and Luminance/Chrominance YCbCr color spaces to develop three image compression models based on Convolutional Auto-Encoder (CAE). In order to evaluate the models’ performances, we used 24 raw images taken from Kodak database and applied the approaches on every one of them and compared achieved experimental results with those obtained using standard compression method. We draw this comparison in terms of performance parameters: Structural Similarity Index Metrix SSIM, Peak Signal to Noise Ratio PSNR and Mean Square Error MSE. Reached results indicates that with proposed schemes we gain significate improvement in distortion metrics over traditional image compression method especially SSIM parameter and we managed to reduce MSE values over than 50%. In addition, proposed schemes output images with high visual quality where details and textures are clear and distinguishable.
Less
1 year ago
Huan -Yu Chen,
Huan -Yu Chen
Institution: Department of Computer Science and Information Engineering, National Taichung University of Science and Technology
Email: info@rnfinity.com
Chuen-Horng Lin,
Chuen-Horng Lin
Institution: Department of Computer Science and Information Engineering, National Taichung University of Science and Technology
Email: info@rnfinity.com
Jyun-Wei Lai,
Jyun-Wei Lai
Institution: Department of Computer Science and Information Engineering, National Taichung University of Science and Technology
Email: info@rnfinity.com
Yung-Kuan Chan
Yung-Kuan Chan
Institution: Department of Management Information Systems, National Chung Hsing University
Email: info@rnfinity.com
This paper proposes a multi–convolutional neural network (CNN)-based system for the detection, tracking, and recognition of the emotions of dogs in surveillance videos. This system detects dogs in each frame of a video, tracks the dogs in the video, and recognizes the dogs’ emotions. The system ...
More
This paper proposes a multi–convolutional neural network (CNN)-based system for the detection, tracking, and recognition of the emotions of dogs in surveillance videos. This system detects dogs in each frame of a video, tracks the dogs in the video, and recognizes the dogs’ emotions. The system uses a YOLOv3 model for dog detection. The dogs are tracked in real time with a deep association metric model (DeepDogTrack), which uses a Kalman filter combined with a CNN for processing. Thereafter, the dogs’ emotional behaviors are categorized into three types—angry (or aggressive), happy (or excited), and neutral (or general) behaviors—on the basis of manual judgments made by veterinary experts and custom dog breeders. The system extracts sub-images from videos of dogs, determines whether the images are sufficient to recognize the dogs’ emotions, and uses the long short-term deep features of dog memory networks model (LDFDMN) to identify the dog’s emotions. The dog detection experiments were conducted using two image datasets to verify the model’s effectiveness, and the detection accuracy rates were 97.59% and 94.62%, respectively. Detection errors occurred when the dog’s facial features were obscured, when the dog was of a special breed, when the dog’s body was covered, or when the dog region was incomplete. The dog-tracking experiments were conducted using three video datasets, each containing one or more dogs. The highest tracking accuracy rate (93.02%) was achieved when only one dog was in the video, and the highest tracking rate achieved for a video containing multiple dogs was 86.45%. Tracking errors occurred when the region covered by a dog’s body increased as the dog entered or left the screen, resulting in tracking loss. The dog emotion recognition experiments were conducted using two video datasets. The emotion recognition accuracy rates were 81.73% and 76.02%, respectively. Recognition errors occurred when the background of the image was removed, resulting in the dog region being unclear and the incorrect emotion being recognized. Of the three emotions, anger was the most prominently represented; therefore, the recognition rates for angry emotions were higher than those for happy or neutral emotions. Emotion recognition errors occurred when the dog’s movements were too subtle or too fast, the image was blurred, the shooting angle was suboptimal, or the video resolution was too low. Nevertheless, the current experiments revealed that the proposed system can correctly recognize the emotions of dogs in videos. The accuracy of the proposed system can be dramatically increased by using more images and videos for training the detection, tracking, and emotional recognition models. The system can then be applied in real-world situations to assist in the early identification of dogs that may exhibit aggressive behavior.
Less
1 year ago
Joan Danielle K. Ongchoco,
Joan Danielle K. Ongchoco
Institution: Department of Psychology
Email: info@rnfinity.com
Madeline Gedvila,
Madeline Gedvila
Institution: Department of Psychology
Email: info@rnfinity.com
Wilma A. Bainbridge
Wilma A. Bainbridge
Institution: Department of Psychology
Email: info@rnfinity.com
Time is the fabric of experience — yet it is incredibly malleable in the mind of the observer: seeming to drag on, or fly right by at different moments. One of the most influential drivers of temporal distortions is attention, where heightened attention dilates subjective time. But an equally impo...
More
Time is the fabric of experience — yet it is incredibly malleable in the mind of the observer: seeming to drag on, or fly right by at different moments. One of the most influential drivers of temporal distortions is attention, where heightened attention dilates subjective time. But an equally important feature of subjective experience involves not just the objects of attention, but also what information gets tagged to be remembered or forgotten in the first place, independent of attention (i.e. intrinsic image memorability). Here we test how memorability influences time perception. Observers viewed scenes in an oddball paradigm, where the last scene could be a forgettable “oddball” amidst memorable ones, or vice versa. Subjective time dilation occurred only for forgettable oddballs, but not memorable ones — demonstrating an oddball effect where the oddball did not differ in low-level visual features, image category, or even subjective memorability. But more importantly, these results emphasize how memory can interact with temporal experience: forgettable endings amidst memorable sequences dilate our experience of time.
Less
1 year ago
Abstract This article presents a fast parallel lossless technique and a lossy image compression technique for 16-bit single-channel images. Nowadays, such techniques are “a must” in robotics and other areas where several depth cameras are used. Since many of these algorithms need to be run in lo...
More
Abstract This article presents a fast parallel lossless technique and a lossy image compression technique for 16-bit single-channel images. Nowadays, such techniques are “a must” in robotics and other areas where several depth cameras are used. Since many of these algorithms need to be run in low-profile hardware, as embedded systems, they should be very fast and customizable. The proposal is based on the consideration of depth images as surfaces, so the idea is to split the image into a set of polynomial functions that each describes a part of the surface. The developed algorithm herein proposed can achieve a similar—or better—compression rate and especially higher speed rates than the existing techniques. It also has the potential of being fully parallelizable and to run on several cores. This feature, compared to other approaches, makes it useful for handling and streaming multiple cameras simultaneously. The algorithm is assessed in different situations and hardware. Its implementation is rather simple and is carried out with LIDAR captured images. Therefore, this work is accompanied by an open implementation in C++.
Less
1 year ago
Baekcheon Seong,
Baekcheon Seong
Institution: Yonsei University
Email: info@rnfinity.com
Abstract Several image-based biomedical diagnoses require high-resolution imaging capabilities at large spatial scales. However, conventional microscopes exhibit an inherent trade-off between depth-of-field (DoF) and spatial resolution, and thus require objects to be refocused at each lateral locati...
More
Abstract Several image-based biomedical diagnoses require high-resolution imaging capabilities at large spatial scales. However, conventional microscopes exhibit an inherent trade-off between depth-of-field (DoF) and spatial resolution, and thus require objects to be refocused at each lateral location, which is time-consuming. Here, we present a computational imaging platform, termed E2E-BPF microscope, which enables large-area, high-resolution imaging of large-scale objects without serial refocusing. This method involves a physics-incorporated, deep-learned design of binary phase filter (BPF) and jointly optimized deconvolution neural network, which altogether produces high-resolution, high-contrast images over extended depth ranges. We demonstrate the method through numerical simulations and experiments with fluorescently labeled beads, cells and tissue section, and present high-resolution imaging capability over a 15.5-fold larger DoF than the conventional microscope. Our method provides highly effective and scalable strategy for DoF-extended optical imaging system, and is expected to find numerous applications in rapid image-based diagnosis, optical vision, and metrology.
Less
1 year ago