In this paper, we present two online structure learning algorithms for NeuralLog, NeuralLog+OSLR and NeuralLog+OMIL. NeuralLog is a system that compiles first-order logic programs into neural networks. Both learning a...
详细信息
ISBN:
(数字)9783030974541
ISBN:
(纸本)9783030974541;9783030974534
In this paper, we present two online structure learning algorithms for NeuralLog, NeuralLog+OSLR and NeuralLog+OMIL. NeuralLog is a system that compiles first-order logic programs into neural networks. Both learning algorithms are based on Online Structure Learner by Revision (OSLR). NeuralLog+OSLR is a port of OSLR to use NeuralLog as inference engine;while NeuralLog+OMIL uses the underlying mechanism from OSLR, but with a revision operator based on Meta-Interpretive Learning. We compared both systems with OSLR and RDN-Boost on link prediction in three different datasets: Cora, UMLS and UWCSE. Our experiments showed that NeuralLog+OMIL outperforms boththe compared systems on three of the four target relations from the Cora dataset and in the UMLS dataset, while both NeuralLog+OSLR and NeuralLog+OMIL outperform OSLR and RDNBoost on the UWCSE, assuming a good initial theory is provided.
Elastograms suffer from noise and undesirable artifacts, making it necessary to enhance their signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for accurate detection of tissue deformations, determination ...
详细信息
ISBN:
(纸本)9798350359732;9798350359749
Elastograms suffer from noise and undesirable artifacts, making it necessary to enhance their signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for accurate detection of tissue deformations, determination of its displacement fields, and representation of tissue characteristics to medical professionals. this research aims to improve the quality of elastograms. Among various displacement estimation methods, the dynamic programming (DP) approach is chosen due to its superior speed, accuracy, and lower computational complexity and cost. the research objectives are defined as enhancing the SNR and CNR metrics and evaluating the resulting improvement in the elastogram quality. To achieve DP refinement, a perturbation is employed. the variables of the DP cost function are evaluated, and finally, the adjustment weight (w) is selected as a suitable criterion for optimization. After assessing different multi-objective optimization algorithms, the Electric Fish Optimization algorithm with Non-dominated Sorting (NS-EFO) is chosen. In the designed optimization process, the DP cost function is considered as the objective function, SNR and CNR as objectives, and w as the design variable. To evaluate the improved performance of DP, two RF samples are used, recorded from a CIRS phantom and a Polyvinyl alcohol (PVA) single-layer mammary gland phantom. Following determining the optimal w, the SNR and CNR values are computed for both one-dimensional (1D) and two-dimensional (2D) models under different conditions. the results indicate that the SNR and CNR values for the CIRS phantom, on average, increased respectively by 556.74% and 853.72% in the 1D model and by 83.56% and 127.91% in the 2D model, compared to the primitive values. For the single-layer glandular phantom, the SNR and CNR values, on average, increased respectively by 66.98% and 33.03% in the 1D model and by 102.67% and 2178.99% in the 2D model, compared to the primary values.
In this paper, we address the issue of using logic rules to explain the results from legal case retrieval. the task is critical to legal case retrieval because the users (e.g., lawyers or judges) are highly specialize...
详细信息
Transformer models, such as BERT, are often taken off-the-shelf and then fine-tuned on a downstream task. Although this is sufficient for many tasks, low-resource settings require special attention. We demonstrate an ...
详细信息
the proceedings contain 64 papers. the topics discussed include: a deep-learning framework for predicting congestion during FPGA placement;lightweight side-channel protection using dynamic clock randomization;executin...
ISBN:
(纸本)9781728199023
the proceedings contain 64 papers. the topics discussed include: a deep-learning framework for predicting congestion during FPGA placement;lightweight side-channel protection using dynamic clock randomization;executing ARMv8 loop traces on reconfigurable accelerator via binary translation framework;precise pointer analysis in high-level synthesis;LFTSM: lightweight and fully testable SEU mitigation system for Xilinx processor-based SoCs;a high throughput MobileNetV2 FPGA implementation based on a flexible architecture for depthwise separable convolution;hardware acceleration of Monte-Carlo sampling for energy efficient robust robot manipulation;and automated design of FPGAs facilitated by cycle-free routing.
programming courses provide students withthe skills to develop complex business applications. Teaching and learning programming is challenging, and collaborative learning is proposed to help withthis challenge. Onli...
详细信息
ISBN:
(纸本)9789869721493
programming courses provide students withthe skills to develop complex business applications. Teaching and learning programming is challenging, and collaborative learning is proposed to help withthis challenge. Online discussion forums promote networking with other learners such that they can build knowledge collaboratively. It aids students open their horizons of thought processes to acquire cognitive skills. Cognitive analysis of discussion is critical to understand students' learning process. In this paper, we propose Bloom's taxonomy based cognitive model for programming discussion forums. We present machine learning (ML) based solution to extract students' cognitive skills. Our evaluations on compupting courses show that ensemble model performs better with an average F1-score of 76%.
In the Industry 4.0 scene, Artificial Intelligence (AI) is sought after as a new way of getting a competitive advantage from other market competitors. this technology can support not only in-line production status ass...
详细信息
ISBN:
(纸本)9798350362442;9798350362435
In the Industry 4.0 scene, Artificial Intelligence (AI) is sought after as a new way of getting a competitive advantage from other market competitors. this technology can support not only in-line production status assessment processes, which enable a better control over the quality of the final product, but also to identify potential bottlenecks and other inefficiencies that can exist or occur in production processes. However, this technology has some obstacles that make its access difficult for businesses that do not have the necessary resources for implementing AI solutions, whether due to the intrinsic difficulty to handle such technologies, which require specialists (engineers, data scientists) that are not normally part of industrial human resources, or due to the integration and management of these technologies with already established processes and environments. To approach these technological accessibility challenges, some concepts are being applied, such as in the case of no code/low code solutions, i.e., the reduction or complete removal of programming requirements while using these technologies, and Machine Learning Operations (MLOps), where the integration and lifecycle management of these solutions use the same approach as DevOps but applied and adapted to AI technologies. this paper presents an innovative, open-source and scalable approach towards AI pipeline creation, integration, and lifecycle management in Industry 4.0 scenarios, in which these no code/low code and MLOps concepts are used, as well as a real-life application in the manufacturing industry.
We present a novel scheduling model that leverages Constraint programming (CP) to enhance problem solving performance in Temporal Planning. Building on the established strategy of decomposing causal and temporal reaso...
详细信息
the proceedings contain 74 papers. the topics discussed include: automatic music score difficulty classification;a feature-based approach for the recognition of image quality degradation in automotive applications;fuz...
ISBN:
(纸本)9798350337297
the proceedings contain 74 papers. the topics discussed include: automatic music score difficulty classification;a feature-based approach for the recognition of image quality degradation in automotive applications;fuzzy logic based algorithm for electrical load management in touring buses;electricity theft detection based on temporal convolutional networks with self-attention;piracy detection in online soccer streaming with video content inspection: an application to the Portuguese market;digitalization in power energy sector: principles of cybersecurity;music recommendation system for shared environments;an approach to EEG based BCI for motor imagery using explainable transfer learning;occlusion-aware pedestrian detection and tracking;proportional fairness in wireless powered mobile edge computing networks;cross-lingual text-to-speech with prosody embedding;and comparing image processing techniques for bubble separation and counting in underwater leaks.
One of the primary factors that encourage developers to contribute to open source software (OSS) projects is the collaborative nature of OSS development. However, the collaborative structure of these communities large...
详细信息
ISBN:
(纸本)9781665452786
One of the primary factors that encourage developers to contribute to open source software (OSS) projects is the collaborative nature of OSS development. However, the collaborative structure of these communities largely remains unclear, partly due to the enormous scale of data to be gathered, processed, and analyzed. In this work, we utilize the World Of Code dataset, which contains commit activity data for millions of OSS projects, to build collaboration networks for ten popular programming language ecosystems, containing in total over 290M commits across over 18M projects. We build a collaboration graph representation for each language ecosystem, having authors and projects as nodes, which enables various forms of social network analysis on the scale of language ecosystems. Moreover, we capture the information on the ecosystems' evolution by slicing each network into 30 historical snapshots. Additionally, we calculate multiple collaboration metrics that characterise the ecosystems' states. We make the resulting dataset publicly available% including the constructed graphs and the pipeline enabling the analysis of more ecosystems.
暂无评论