This paper proposes a reinforcement learning-based control education platform utilizing python and light-weight rapid control prototyping (LW-RCP). The platform employs the Sim-to-Real technique, in which neural netwo...
详细信息
This paper presents opstool, a python package designed to enhance the pre- and post-processing capabilities of OpenSees and OpenSeesPy. It simplifies structural analysis workflows by automating tasks such as mesh gene...
详细信息
This paper presents opstool, a python package designed to enhance the pre- and post-processing capabilities of OpenSees and OpenSeesPy. It simplifies structural analysis workflows by automating tasks such as mesh generation, data management, and data visualization. The package efficiently manages large-scale simulation results, enabling the structured extraction of system, nodal, and element responses. In addition, it integrates adaptive iteration algorithms to improve convergence issues in nonlinear static and dynamic response analyses. By reducing manual modeling effort and enhancing model accuracy, opstool improves workflow efficiency and enables researchers and practitioners to conduct more effective computational simulations using OpenSees and OpenSeesPy, which further supports various task forces in earthquake engineering, such as performance-based design of new structures and regional seismic risk assessment of existing infrastructure systems.
In recent years, python has become widely used language for data processing in machine learning and deep learning. However, the dynamic typing of python can lead to errors caused by array shape mismatches that are onl...
详细信息
In recent years, python has become widely used language for data processing in machine learning and deep learning. However, the dynamic typing of python can lead to errors caused by array shape mismatches that are only detected at runtime. To increase development efficiency, we propose a static method to check python code that can detect shape errors before execution. Existing research activities such as Pytropos provide the capability to manually annotate the shape types of arrays and external datasets with python's type hint feature. However, manual annotation decreases code flexibility and can cause problems, such as incorrect marking due to human error, and wasted labor and time costs. To address these issues, we propose a method that leverages abstract interpretation and abstract syntax tree analysis to statically check code for array and dataset shape types, thus reducing the need for manual annotation and improving code flexibility. We give an implementation of the proposed method named ShapeChecker for the widely-used library NumPy as an example. ShapeChecker extracts the shape type of NumPy arrays and automatically reads external datasets to obtain shape type information, accelerating the checking process and outputting the cause of shape errors when detected. We compared ShapeChecker with existing solutions in various scenarios and obtained promising results, demonstrating the tool's usefulness in improving efficiency and reducing runtime errors. To further enhance its functionality, we plan to extend ShapeChecker's support to more packages and address known issues. Overall, our proposed method and ShapeChecker tool provide a static approach to detecting array shape errors that can improve code quality and improve development efficiency.
With CUDA computational model in mind, we introduce MCTS-NC (Monte Carlo Tree ***). It contains four, fast-operating and thoroughly parallel, variants of the MCTS algorithm. The design of MCTS-NC combines three parall...
详细信息
With CUDA computational model in mind, we introduce MCTS-NC (Monte Carlo Tree ***). It contains four, fast-operating and thoroughly parallel, variants of the MCTS algorithm. The design of MCTS-NC combines three parallelization levels (leaf/root/tree parallelizations). Additionally, all algorithmic stages-selections, expansions, playouts, backups-employ multiple GPU threads. We apply suitable reduction patterns to carry out summations or max/ argmax operations. The implementation uses very few device-host memory transfers, no atomic operations (is lock-free), and takes advantage of threads cooperation. In the mathematical part of this article, we demonstrate how the confidence bounds on estimated action values become tightened by both the number of independent concurrent playouts and the number of independent concurrent trees. The experimental part reports the performance of MCTS-NC on two game examples: Connect4 and Gomoku. All computational results can be exactly reproduced.
Federated learning (FL) is a machine learning setting where clients keep the training data decentralized and collaboratively train a model either under the coordination of a central server (centralized FL) or in a pee...
详细信息
Federated learning (FL) is a machine learning setting where clients keep the training data decentralized and collaboratively train a model either under the coordination of a central server (centralized FL) or in a peer-to-peer network (decentralized FL). Correct orchestration is one of the main challenges. In this paper, we formally verify the correctness of two generic FL algorithms, a centralized and a decentralized one, using the Communicating Sequential Processes (CSP) calculus and the Process Analysis Toolkit (PAT) model checker. The CSP models consist of CSP processes corresponding to generic FL algorithm instances. PAT automatically proves the correctness of the two generic FL algorithms by proving their deadlock freedom (safety property) and successful termination (reachability and liveness property). The CSP models are constructed as a faithful representation of the real python code and are expressed directly in CSP# language that PAT uses. Then they are automatically checked top-down by PAT. The python code follows a restricted actor-based programming model, and the construction of CSP# code from such python code is performed systematically. The process is described in detail, ensuring that the models correspond to the actual code. It represents a basis for developing tools for automatic translation of certain classes of python code to CSP models, expressed in CSP#.
Drought is a hazard that causes great economic, ecological, and human loss. With an ever-growing risk of climate change, their frequency and magnitude are expected to increase. While there are many indices and metrics...
详细信息
Drought is a hazard that causes great economic, ecological, and human loss. With an ever-growing risk of climate change, their frequency and magnitude are expected to increase. While there are many indices and metrics available for the analysis of droughts, assessing their impacts represents one of the best ways to understand their magnitude and extent. However, there are no systematic records outlining these impacts. To help in their ongoing creation, we present a software framework that leverages raw newspaper articles, identifies any drought-related ones, and automatically classifies them according to a set of socioeconomic impacts. The information is provided to the user in a structured format, including geographical coordinates and their date of reporting. Our approach employs state-of-the-art Transformer-based Natural Language Processing (NLP) techniques, which achieve great accuracy. We currently support newspaper articles in the Spanish language within Spain, but our framework can be expanded to other countries and languages.
This study evaluates leading generative AI models for python code generation. Evaluation criteria include syntax accuracy, response time, completeness, reliability, and cost. The models tested comprise OpenAI's GP...
详细信息
This study evaluates leading generative AI models for python code generation. Evaluation criteria include syntax accuracy, response time, completeness, reliability, and cost. The models tested comprise OpenAI's GPT series (GPT-4 Turbo, GPT-4o, GPT-4o Mini, GPT-3.5 Turbo), Google's Gemini (1.0 Pro, 1.5 Flash, 1.5 Pro), Meta's LLaMA (3.0 8B, 3.1 8B), and Anthropic's Claude models (3.5 Sonnet, 3 Opus, 3 Sonnet, 3 Haiku). Ten coding tasks of varying complexity were tested across three iterations per model to measure performance and consistency. Claude models, especially Claude 3.5 Sonnet, achieved the highest accuracy and reliability. They outperformed all other models in both simple and complex tasks. Gemini models showed limitations in handling complex code. Cost-effective options like Claude 3 Haiku and Gemini 1.5 Flash were budget-friendly and maintained good accuracy on simpler problems. Unlike earlier single-metric studies, this work introduces a multi-dimensional evaluation framework that considers accuracy, reliability, cost, and exception handling. Future work will explore other programming languages and include metrics such as code optimization and security robustness.
In this research, we introduce a new python bioinformatics tool. QuaDS (Quantitative/Qualitative Description Statistics) is a pipeline tailored to describe a factor (a qualitative variable of interest) in heterogeneou...
详细信息
In this research, we introduce a new python bioinformatics tool. QuaDS (Quantitative/Qualitative Description Statistics) is a pipeline tailored to describe a factor (a qualitative variable of interest) in heterogeneous datasets consisting of qualitative and quantitative variables. This pipeline separately analyze s the variables related to the factor using appropriate statistical tests. The QuaDS pipeline offers an interactive visualization that describes the factor. Several parameters can be defined by the user to ensure the most personalized results based on their data.
Brain-computer interfaces (BCIs) establish a connection between the human brain and external devices, facilitating novel forms of communication and control. Although BCIs possess significant potential, there is a need...
详细信息
Brain-computer interfaces (BCIs) establish a connection between the human brain and external devices, facilitating novel forms of communication and control. Although BCIs possess significant potential, there is a need for performance improvements. Integrating neural networks into BCIs is crucial for enhancing functionality and promoting broader adoption. This study examines 1,867 articles from the Web of Science core database, covering the period from 1996 to 2024, to identify contemporary hotspots and trends in the application of neural networks within BCIs. This study employs bibliometric methods and python for analysis, examining collaborative relationships, citation networks, keyword bursts, and clustering with visual representations of the findings. The results indicated that current study hotspots predominantly center on "P300," "Long Short-Term Memory," "Motor Imagery," "Epilepsy," "Emotion Recognition," "Feature Extraction," and "Transfer Learning." Future development directions encompass: (1) the establishment of public BCI datasets;(2) the exploration of diverse feature extraction and fusion methods;(3) the enhancement of machine learning and deep learning integration for improved performance and real-time processing;(4) the expansion of application scenarios and the development of portable devices;(5) the optimization of transfer learning algorithms to mitigate performance challenges arising from individual differences. This study provides an overview of the current research landscape and identifies potential future research directions. Furthermore, it assists practitioners in recognizing additional business opportunities and acts as a resource for the formulation of government policy.
We present a converter software program that automatically translates python-based machine learning algorithms into Structured Text codes. This tool empowers engineers to efficiently generate machine learning models i...
详细信息
We present a converter software program that automatically translates python-based machine learning algorithms into Structured Text codes. This tool empowers engineers to efficiently generate machine learning models in a programming language widely used in industrial controllers. It supports the conversion of decision tree and multilayer perceptron models built using scikit-learn library. Moreover, the generated Structure Text code is compatible with ABB's Industrial IT 800xA DCS syntax. A practical example demonstrates the effectiveness of this converter software program and its potential to enhance the integration of machine learning models into industrial automation systems.
暂无评论