Subjects: Physics >> General Physics: Statistical and Quantum Mechanics, Quantum Information, etc. Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-05-14
Abstract: In his 1950 paper cite{Turing1950}, Turing proposed the notion of a thinking machine, namely a machine that can think. But a thinking machine has to follow a certain law of physics, provided it is realized physically. In this paper, we show that Turing’s thinking machine necessrily obeys ’t Hooft’s principle of superposition of states, which was presented by ’t Hooft cite{Hooft2016} in 2016 beyond the usual one as described by Dirac cite{Dirac1958} in the conventional quantum mechanics. Precisely, Turing’s thinking machine must be a quantum machine, while ’t Hooft’s principle characterizes its thinking behavior in a probabilistic way.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-05-09
Abstract: In response to the demand from academia and industry for a scientific classification of malware,this study builds upon existing work and draws insights from Kaspersky’s rigorously multi-stage naming methodology, emphasizing the principles of mutual exclusivity, comprehensive coverage, and convergence,combines the use of threat risk behavior labels ,forms a classification framework for malware that conforms to the MECE (Mutually Exclusive, Collectively Exhaustive) principle, achieves convergence, is compatible with real-world industry classifications,provides effectively support on security defense and governance.
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-04-24
Abstract: In capital market, earlier detection of the influential entities can be beneficial to both market investors’ and regulators’ decision making, those whose change can significantly affect the whole trend of the related ones. Meanwhile, market manipulation in capital markets is a serious concern, encompassing tactics like pump and dump, market cornering, spoofing, and wash trading, which disrupt market fairness and erode investor trust. Market manipulation encompasses a range of activities designed to artificially influence the price or trading volume. By leveraging both information behavior data(stock news opinion/volume) and business behavior(stock trading price/volume), together with trade patterns and communication channels, several herding based manipulation scenes and detection models are discussed and proposed.
Peer Review Status:Awaiting Review
Subjects: Information Science and Systems Science >> Basic Disciplines of Information Science and Systems Science Subjects: Computer Science >> Other Disciplines of Computer Science Subjects: Engineering and technical science >> Engineering Mathematics submitted time 2024-03-30
Abstract: What is intelligence? is one of the core key questions of artificial intelligence, but there is no universally accepted definition. Based on the relationship between intelligence and life, this paper proposes that intelligence is the basic ability and characteristic attribute of living organisms, and it is the ability to achieve the maximum amount of information with the minimum energy as much as possible, and adapt to the environment and maintain existence through information processing. On this basis, this paper puts forward a new view that intelligence is the ability to convert material energy and information, further puts forward new concepts such as the measurement calculation method of intelligence, average intelligence, and comprehensive intelligence, and finally discusses the quantitative conversion relationship between matter, energy and information, points out the upper bound of intelligence and the lower bound of energy conversion into information, and further gives a dimensionless calculation formula for intelligence measurement in order to facilitate practical application. A feasible calculation method is given for the quantitative analysis of the intelligence of the intelligent system..
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-02-20
Abstract: This paper analyzes the main problems of the current LLM and proposes specific solutions, pointing out the fact that: the expression and computation of the conceptualized structural model combined with probability is the key, and provides a brief explanation of the related technology-Deep Semantic Model (DSM), and finally enumerates the direction of the subsequent key work.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-10
Abstract: With the development of science and technology, EEG signals are utilized in more and more fields because of their rich physiological, psychological and pathological contents. After relevant experiments on the dataset of EEG signals, it is shown that the discrimination of Alzheimer's disease based on the LSTM model realized in this project achieves better results, with an average accuracy of about 93%.
Peer Review Status:Awaiting Review
Subjects: Statistics >> Mathematical Statistics Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-08
Abstract: The No Free Lunch (NFL) theorem is an important result of statistical learning theory, according to Bayesian modeling, the expectation of the loss/utility can be deduced with its form related to the selection of the hypothesis space of the prediction functions. If the real prediction function space is considered unknowable, then the arbitrarily selected hypothesis function space may not necessarily yield the expectation of the optimal loss function.
In this paper, the limit behavvior of the NFL theorem is analyzed based on a local form of the uniform convergence of the empirical distribution, i.e. the Glivenko-Cantelli theorem, is obtained: under certain condition of the deterministic and non-deterministic prediction problem, the expectation of the loss/utility is independent of the specific choice of the hypothetical function space as the sample size tends to infinity. A by-product of this work is that the total variation of the distribution can be deduced from the local form of the uniform convergence of the distribution derived in this paper. Previously, this property was generally considered non-existent.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-08
Abstract: With the rapid development of sensor and network technology, a large amount of historical time series data appears, so it is more and more important to predict time series efficiently and accurately. In recent years, the methods of applying deep learning ideas and techniques to time series prediction tasks have developed rapidly and achieved many results. This paper analyzes the research status of time series forecasting methods at home and abroad, discusses the relevant theories involved in time series forecasting, summarizes the traditional methods used in this task, the methods based on machine learning and the methods based on deep learning, and focuses on the comparison and analysis of the advantages and disadvantages of each method based on deep learning. Then, the prediction methods of time series based on deep learning are forecasted.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-07
Abstract: Image classification and recognition are of great significance in modern society. There have been many excellent convolutional neural network works to optimize the accuracy of image classification, one of the outstanding representatives is ResNet 1 , which greatly increases the depth of the neural network, thereby greatly improving the performance of the neural network. At the same time, there are some pluggable performance optimization sub-modules that can help optimize all networks, one of the outstanding representatives is SeNet 3 . However, they do not always perform well when faced with complex scenarios in the real world. The main work of this article is to study how to effectively improve the recognition performance of convolutional neural networks (ResNet) in some special scenes (small pictures, high-noise pictures), and try to analyze the underlying mechanisms of some neural networks.
Peer Review Status:Awaiting Review
Subjects: Environmental Sciences, Resource Sciences >> Environmentology Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-07
Abstract: Objective The aim of this study is to develop a portable gas chromatograph, combined with machine learning, to achieve on-site VOC collection and rapid odor evaluation.
Methods We used a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) to establish an odor intensity prediction model. Due to the small amount of data collected, we used a Generative Adversarial Network (GAN) to generate VOC data for each odor intensity category to enhance model training.
Results After generating the data, we used CNN-LSTM to establish the model again and compared it with Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Gradient Boosting Decision Trees (XG-Boost). The results showed that the test accuracy after using GAN to generate data was better than the original data.
Limitations Future work will focus on further optimizing the model and expanding the dataset to improve the accuracy and stability of the prediction.
Conclusion This study shows that by using deep learning and generative adversarial networks, we can effectively predict the odor intensity inside the car, thereby improving the air quality inside the car. In addition, we will explore the application of this method to air quality prediction under other environmental conditions. This provides new possibilities for future air quality monitoring and improvement. As our equipment is portable and the model structure is small enough to be directly embedded into the device, it can achieve on-site VOC collection and rapid odor evaluation. This provides new possibilities for future air quality monitoring and improvement.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-06
Abstract: With the rapid development of deep learning technology, its wide application in the field of image classification has brought revolutionary changes to the field of computer vision. This paper focuses on the application of deep learning in image classification. Based on CIFAR-10 data set, different models are constructed by convolutional neural network (CNN) to evaluate the classification accuracy of images. I also took well-known models such as ResNet and evaluated their classification performance on the CIFAR-10 dataset. In order to further study the characteristics of the model, we introduce different activation functions in the experiment, and reveal the effect of activation functions on the model performance by comparing their performance in the classification task. In addition, we systematically study the classification accuracy of the model under different training rounds by adjusting hyperparameters such as training rounds. In the experiment, I found that the number of training rounds is not always better, and there may be overfitting phenomenon. However, I solved the overfitting problem to some extent by adjusting the model and adding the Dropout layer. At the same time, through the optimization and adjustment of the model, my classification accuracy has also been greatly improved, up to 0.7509. In the process of continuous optimization, I am more and more confident about the prospect of deep learning in image classification, and I will continue to further study in this direction.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-04
Abstract: Neural Transducer and connectionist temporal classification (CTC) are popular end-to-end automatic speech recognition systems. Due to their frame-synchronous design, blank symbols are introduced to address the length mismatch between acoustic frames and output tokens, which might bring redundant computation. Previous studies managed to accelerate the training and inference of neural Transducers by discarding frames based on the blank symbols predicted by a co-trained CTC. However, there is no guarantee that the co-trained CTC can maximize the ratio of blank symbols. This paper proposes two novel regularization methods to explicitly encourage more blanks by constraining the self-loop of non-blank symbols in the CTC. Experiments on LibriSpeech corpus show that our proposed method accelerates the inference of neural Transducer by 4 times without sacrificing performance. It is interesting to find that the frame reduction ratio of the neural Transducer can approach the theoretical boundary. Additionally, a further gain can be observed when decoding with external language models. To the best of our knowledge, this is the first work to explore the feasibility of neural Transducers with almost no blank symbols.
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-12-18
Abstract: In this paper, we propose a new approach to building a artificial general intelligence with self awareness, which includes: (1) a new method to implement attention mechanisms; (2) a way to give machines self-demands; (3) how to form a value evaluation system compatible with the network; (4) a way to create the world models; (5) how to realize a top-down, hierarchical thinking decision-making chain; (6) a way to achieve general decision-making and response capabilities; (7) a way for a machine to directly obtain human experience through language. In the paper, we first analyze some of the shortcomings of current LLMs (Large Language Model) and propose ideas for improvement. Then we analyze why our scheme can solve the above problems and provide detailed steps for implementing our scheme. In chapter 6, we analyze the advantages and disadvantages of our scheme and propose further research directions. Finally, we propose our thoughts on the next step of AI development.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-12-05
Abstract: Human parsing is a fundamental task aimed at segmenting human images into distinct body parts and holds vast potential applications. Nowadays, the advancement of image-capturing devices has led to a growing number of high-resolution human images. Receptive field, details loss and memory usage are a triplet of contradictions in high-resolution scenarios. Existing human parsing methods designed for low-resolution inputs struggle to process high-resolution images efficiently due to their massive demands for computation and memory. Some methods save resources by overwhelmingly downsampling or encoding high-resolution inputs at the cost of poor performance on details. To resolve the issues above, we propose the Bilateral Edge-Perceiving Network (BiEPNet), consisting of a resources-friendly semantic-perceiving branch to acquire sufficient global information and a simple yet effective edge-perceiving branch used to refine details. The attention mechanism is utilized to simultaneously enhance the perception of context and details, leading to better performance on the boundary regions. To verify the effectiveness of BiEPNet, we contribute a high-resolution human parsing dataset, Human4K, containing 4,000 images with more than five million pixels. Extensive experiments on Human4K demonstrate that our method outperforms state-of-the-art methods while maintaining memory efficiency.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-08-31
Abstract: LLAMA is the most popular open-source Large Language Model(LLM) model in the last few months.
This paper presents its mathematic formulas in detail.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-08-15
Abstract: At present, the mainstream artificial intelligence generally adopts the technical path of "attention mechanism + deep learning" + "reinforcement learning". It has made great progress in the field of AIGC (Artificial Intelligence Generated Content), setting off the technical wave of big models[2][13]. But in areas that need to interact with the actual environment, such as elderly care, home nanny, agricultural production, and vehicle driving, trial and error are expensive and a reinforcement learning process that requires much trial and error is difficult to achieve. Therefore, in order to achieve Artificial General Intelligence(AGI) that can be applied to any field, we need to use both existing technologies and solve the defects of existing technologies, so as to further develop the technological wave of artificial intelligence. In this paper, we analyze the limitations of the technical route of large models, and by addressing these limitations, we propose solutions, thus solving the inherent defects of large models. In this paper, we will reveal how to achieve true AGI step by step.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-07-06
Abstract: In data storage and transmission, file compression is a common technique for reducing the volume of data, reducing data storage space and transmission time and bandwidth. However, there are significant differences in the compression performance of different types of file formats, and the benefits vary. In this paper, 22 file formats with approximately 178GB of data were collected and the Zlib algorithm was used for compression experiments to compare performance in order to investigate the compression gains of different file types. The experimental results show that some file types are poorly compressed, with almost constant file size and long compression time, resulting in lower gains; some other file types are significantly reduced in file size and compression time after compression, which can effectively reduce the data volume. Based on the above experimental results, this paper will then selectively reduce the data volume by compression in data storage and transmission for the file types in order to obtain the maximum compression yield.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science Subjects: Mechanics >> Other Disciplines of Mechanics submitted time 2023-06-15
Abstract: Topology optimization is widely used in the engineering design phase to maximize product performance by mathematically modeling and optimizing the distribution of materials in the design space. However, deep learning to solve the topology optimization problem suffers from insufficient data and weak adaptability of the training model boundary conditions. Therefore, a Topy library-based data sample generation method is used to generate 400,000 2D samples of four types of boundary conditions for random structures, cantilever beams, continuous beams and simply supported beams, each containing two types of resolution data, and to expose this dataset. An improved DoubleU-Net network is proposed for topology optimization with high accuracy prediction in real time. In the generated dataset, the average IoU accuracies of the models for four structures, namely, random beam, cantilever beam, continuous beam and simply supported beam, are 93.26%, 96.71%, 96.35% and 97.38, respectively, and the experimental results show that DoubleU-Net can better adapt to different resolution data. The model trained with the random structure dataset has strong generalization ability and has great potential for real-time structural optimization in large-scale projects.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science Subjects: Mechanics >> Oscillation and Wave submitted time 2023-06-15
Abstract: This paper proposes a modal analysis strategy based on dilated residual convolutional broad network. In modal analysis, vibration analysis of large-scale structures or complex systems usually requires processing large amounts of data and complex calculations.The dilated residual convolution width architecture can reduce the number of parameters and computational complexity of the network, reduce the computational burden, and improve the efficiency of analysis.The dilated residual convolutional broad network applied to modal analysis tasks can improve the extraction ability of vibration features, improve the accuracy of modal identification, and enhance the sensitivity of structural damage detection, and has high computational efficiency and parameter efficiency. The experimental results show that our model achieves excellent performance in the regression task of modal analysis prediction.
Peer Review Status:Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2023-05-22
Abstract: With its excellent expressive power and explicit presentation, Knowledge Graph(KG) has become one of the main forms of knowledge representation and management. However, due to biases in the cognitive and construction processes, their are varying degrees of missing and incorrect problems in KGs. Hence, Knowledge Graph Reasoning(KGR) is needed for completion and denoising of KGs. Rule-based KGR methods reason by mining the underlying logic rules, which empowers them strong generalization ability and interpretability. Rules can also be flexibly combined with Neural Networks. Currently, rule-based knowledge graph reasoning methods are constantly emerging, but there is still a lack of systematic categorization and analysis. This study collects the research works related to rule-based knowledge graph reasoning, summarizes four categories of methods, including Inductive Logic Programming based methods, methods combining rules with probability graph,methods combining rules with embedding representation, methods combining rules with Neural Networks. Core ideas, key technologies, advantages and disadvantages of relevant classical methods are then elaborated. Besides, performance of the methods are compared. Finally, current problems and challenges are summarized, and the future development direction is prospected.
Peer Review Status:Awaiting Review