We derive information theoretic generalization bounds for supervised learning algorithms based on a new measure of leave-one-out conditional mutual information (loo-CMI). Contrary to other CMI bounds, which are black-...
详细信息
Deep learning algorithms have become widely used in industrial applications to optimize several tasks in many complex systems, particularly for diagnosing and prognosing machinery health, which have leveraged predicti...
详细信息
Protein structures and functions are determined by a contiguous arrangement of amino acid sequences. Designing novel protein sequences and structures with desired geometry and functions is a complex task with large st...
详细信息
Applying limited labeled samples to improve classification results is a challenge in hyperspectral images. Active learning (AL) and Semisupervised learning (SSL) are two promising techniques to achieve this challenge....
详细信息
Applying limited labeled samples to improve classification results is a challenge in hyperspectral images. Active learning (AL) and Semisupervised learning (SSL) are two promising techniques to achieve this challenge. Combining AL with SSL is an excellent idea for hyperspectral image classification. The traditional method, such as the Collaborative Active and Semisupervised learning algorithm (CASSL), may introduce many incorrect pseudolabels and shows premature convergence. To overcome these drawbacks, a novel framework named Double-Strategy-Check Collaborative Active and Semisupervised learning (DSC-CASSL) is proposed in this paper. This framework combines two different AL algorithms and SSL in a collaborative mode. The double-strategy verification can gradually improve the pseudolabeling accuracy and facilitate SSL. We evaluate the performance of DSC-CASSL on four hyperspectral data sets and compare it with that of four hyperspectral image classification methods. Our results suggest that DSC-CASSL leads to consistent improvement for hyperspectral image classification.
Chinese word segmentation (CWS) is a fundamental task of natural language processing. Currently, CWS model using fully supervised learning technology has achieved good results in the common domain. However, it has the...
详细信息
Objective: With the development of remote sensing technology, high-resolution remote sensing images become available to scene recognition of opencast coal mine areas, which is conducive to the supervision of opencast ...
详细信息
Objective: With the development of remote sensing technology, high-resolution remote sensing images become available to scene recognition of opencast coal mine areas, which is conducive to the supervision of opencast coal mine areas for environmental governance. The scene is divided into multiple sub-regions for feature learning and recognition. Aiming at the poor performance of sub-region recognition based on single-label learning, this paper combines a multi-label learning strategy with the first law of geography to propose a scene recognition algorithm based on scene sub-region multi-label learning. Method: In order to distinguish the scene of opencast coal mine areas from its surrounding scene, 6 types of mining tags and 7 non-mining tags are set. The sub-regions, cropped from the scene of opencast coal mine areas and its surrounding scene, are labeled with 13 types of tags to form a multi-label dataset. Train the dataset with the Inception_v3 based on multi-label learning. The input remote sensing images are divided into sub-regions of the same size, and multi-label classification is performed on the sub-regions with the trained model. In order to recognize the sub-regions belonging to the scene of the opencast coal mine areas according to the multi-label classification results, a scene sub-region determination algorithm is introduced. Using the label correlation and the label integrity of the mining tags to determine whether the sub-region, containing the mining tags, belongs to the scene of the opencast coal mine areas. And the recognized sub-regions constitute the scene of the opencast coal mine areas. Result: The results show that, in scene sub-region recognition of opencast coal mine areas, compared with single-label learning algorithms, the F1-score of the proposed method, 0.857, is increased by up to 8 percentage points. On the remote sensing image of the study area, the recognition results of proposed method in scene recognition of opencast coal mine ar
There is a growing interest in novelty search : that is, in sampling a parameter space to search for radical or unexpected behaviour(s), occurring as a consequence of parameter choice, being input to...
详细信息
There is a growing interest in novelty search : that is, in sampling a parameter space to search for radical or unexpected behaviour(s), occurring as a consequence of parameter choice, being input to some downstream complex system, process, or service that will not yield to analysis, without imposing any specific pre-ordained objective function, or fitness function to be optimised. We mean "parameter" in the widest sense, including system learnables, non-autonomous forcing, sequencing and all *** upon the nature of the underlying parameter space of interest one may adopt a rather wide range of search algorithms. We do consider that this search activity has meta-objectives , though: one is of achieving diversity (efficiently reaching out across the space in some way);and one is of achieving some minimum density (not leaving out large unexplored holes). These are in tension. In general, the computational costs of both of these qualities become restrictive as the di- mension of the parameter spaces increase;and consequently their balance is harder to maintain. We may also wish for a substantial random element of search to provide some luck in discovery and to avoid any naive preset sampling *** consider archive-based methods within a range of spaces: finite discrete spaces, where the problem is straightforward (provided we are patient with the random element);Euclidean spaces, of increasing dimension, that become very lonely places;and infinite dimensional spaces. Our aim is to discuss a raft of distinctive search concepts, that respond to identified challenges, and rely on a rather diverse range of mathematical ideas. This arms practitioners with a range of highly practical *** applications requiring novelty search arise, one should avoid rushing to code-up a standard evolving search algorithm and instead give some thought to the nature and requirements of the search: there is a range of effective options available. We give some
Objective: Evidence-based medicine (EBM) is a decision-making process based on the conscious and judicious use of the best available scientific evidence. However, the exponential increase in the amount of information ...
详细信息
Objective: Evidence-based medicine (EBM) is a decision-making process based on the conscious and judicious use of the best available scientific evidence. However, the exponential increase in the amount of information currently available likely exceeds the capacity of human-only analysis. In this context, artificial intelligence (AI) and its branches such as machine learning (ML) can be used to facilitate human efforts in analyzing the literature to foster EBM. The present scoping review aimed to examine the use of AI in the automation of biomedical literature survey and analysis with a view to establishing the state-of-the-art and identifying knowledge *** and methods: Comprehensive searches of the main databases were performed for articles published up to June 2022 and studies were selected according to inclusion and exclusion criteria. Data were extracted from the included articles and the findings ***: The total number of records retrieved from the databases was 12,145, of which 273 were included in the review. Classification of the studies according to the use of AI in evaluating the biomedical literature revealed three main application groups, namely assembly of scientific evidence (n=127;47%), mining the biomedical literature (n=112;41%) and quality analysis (n=34;12%). Most studies addressed the preparation of systematic reviews, while articles focusing on the development of guidelines and evidence synthesis were the least frequent. The biggest knowledge gap was identified within the quality analysis group, particularly regarding methods and tools that assess the strength of recommendation and consistency of ***: Our review shows that, despite significant progress in the automation of biomedical literature surveys and analyses in recent years, intense research is needed to fill knowledge gaps on more difficult aspects of ML, deep learning and natural language processing, and to consolidate the use of automation by en
We study sampling problems associated with non-convex potentials that meanwhile lack smoothness. In particular, we consider target distributions that satisfy either logarithmic-Sobolev inequality or Poincaré ineq...
详细信息
Machine learning models face increased concerns regarding the storage of personal user data and adverse impacts of corrupted data like backdoors or systematic bias. Machine Unlearning can address these by allowing pos...
详细信息
暂无评论