Search

Introduction: Recently, Google introduced Pathways as its next-generation AI architecture. Pathways must address three critical challenges: learning one general model for several continuous tasks, ensuring tasks can leverage each other without forgetting old tasks, and learning from multi-modal data such as images and audio. Additionally, Pathways must maintain sparsity in both learning and deployment. Current lifelong multi-task learning approaches are inadequate in addressing these challenges., Methods: To address these challenges, we propose SEN, a Sparse and Expandable Network. SEN is designed to handle multiple tasks concurrently by maintaining sparsity and enabling expansion when new tasks are introduced. The network leverages multi-modal data, integrating information from different sources while preventing interference between tasks., Results: The proposed SEN model demonstrates significant improvements in multi-task learning, successfully managing task interference and forgetting. It effectively integrates data from various modalities and maintains efficiency through sparsity during both the learning and deployment phases., Discussion: SEN offers a straightforward yet effective solution to the limitations of current lifelong multi-task learning methods. By addressing the challenges identified in the Pathways architecture, SEN provides a promising approach for developing AI systems capable of learning and adapting over time without sacrificing performance or efficiency., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2024 Ling, Wang and Wang.)

Tuning the values of kernel parameters plays a vital role in the performance of kernel methods. Kernel path algorithms have been proposed for several important learning algorithms, including support vector machine and kernelized Lasso, which can fit the piecewise nonlinear solutions of kernel methods with respect to the kernel parameter in a continuous space. Although the error path algorithms have been proposed to ensure that the model with the minimum cross validation (CV) error can be found, which is usually the ultimate goal of model selection, they are limited to piecewise linear solution paths. To address this problem, in this article, we extend the classic error path algorithm to the nonlinear kernel solution paths and propose a new kernel error path algorithm (KEP) that can find the global optimal kernel parameter with the minimum CV error. Specifically, we first prove that error functions of binary classification and regression problems are piecewise constant or smooth w.r.t. the kernel parameter. Then, we propose KEP for support vector machine and kernelized Lasso and prove that it guarantees to find the model with the minimum CV error within the whole range of kernel parameter values. Experimental results on various datasets show that our KEP can find the model with minimum CV error with less time consumption. Finally, it would have better generalization error on the test set, compared with grid search and random search.

Net primary productivity (NPP) is an indicator to reflect the production capacity of terrestrial ecosystems, as well as a key indicator for ecological quality. NPP at large scale is difficult to be measured. At present, most of the assessment of ecosystem quality uses NPP products with low resolution, which cannot capture the detailed characteristics of the ecosystem and is not conducive to the assessment of ecosystem quality at small-scale. The establishment of a rapid and efficient assessment method for small-scale ecosystem quality will greatly promote the protection and restoration of ecosystems in China. We focused on the calculation method of ecosystem quality assessment and NPP, and optimized the calculation process of the NPP, and obtained NPP by remote sensing without ground observation data. We established a regression model for remote sensing inversion of leaf area index, and estimated the vegetation coverage by using dimidiate pixel model, developed a set of methods for rapid assessment of ecosystem quality by remote sensing. Taking Nanwenghe National Nature Reserve as an example, we evaluated the change of ecosystem quality from 2001 to 2022. The results showed that from 2001 to 2022, the ecosystem qua-lity of the reserve was good as a whole, and that the area with good and excellent quality accounted for more than 85% in 2022. High vegetation coverage was the backbone of the sustainable good ecosystem quality of the reserve. From 2001 to 2022, ecosystem quality of the reserve showed a trend of first decreasing and then increasing, with the lowest point of ecosystem quality in 2013. This method had achieved good results in the evaluation of ecosystem quality in Nanwenghe National Nature Reserve. The index optimization method proposed in this study could facilitate the rapid and accurate assessment of ecosystem quality in small-scale regions, and thus provides technical reference for the precise improvement of ecosystem quality.

Learning knowledge from different tasks to improve the general learning performance is crucial for designing an efficient algorithm. In this work, we tackle the Multi-task Learning (MTL) problem, where the learner extracts the knowledge from different tasks simultaneously with limited data. Previous works have been designing the MTL models by taking advantage of the transfer learning techniques, requiring the knowledge of the task index, which is not realistic in many practical scenarios. In contrast, we consider the scenario that the task index is not explicitly known, under which the features extracted by the neural networks are task agnostic. To learn the task agnostic invariant features, we implement model agnostic meta-learning by leveraging the episodic training scheme to capture the common features across tasks. Apart from the episodic training scheme, we further implemented a contrastive learning objective to improve the feature compactness for a better prediction boundary in the embedding space. We conduct extensive experiments on several benchmarks compared with several recent strong baselines to demonstrate the effectiveness of the proposed method. The results showed that our method provides a practical solution for real-world scenarios, where the task index is agnostic to the learner and can outperform several strong baselines, achieving state-of-the-art performances., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Background: Many smokers have undiagnosed chronic obstructive pulmonary disease (COPD), and yet screening for COPD is not recommended. Smokers who know that they have airflow limitation are more likely to quit smoking. This study aims to identify the prevalence and predictors of airflow limitation among smokers in primary care., Methods: Current smokers ≥ 40 years old who were asymptomatic clinic attendees in a primary care setting were recruited consecutively for two months. We used a two-step strategy. Step 1: participants filled in a questionnaire. Step 2: Assessment of airflow limitation using a pocket spirometer. Multiple logistic regression was utilised to determine the best risk predictors for airflow limitation., Results: Three hundred participants were recruited. Mean age was 58.35 (SD 10.30) years old and mean smoking history was 34.56 pack-years (SD 25.23). One in two smokers were found to have airflow limitation; the predictors were Indian ethnicity, prolonged smoking pack-year history and Lung Function Questionnaire score ≤ 18. Readiness to quit smoking and the awareness of COPD were low., Conclusions: The high prevalence of airflow limitation and low readiness to quit smoking imply urgency with helping smokers to quit smoking. Identifying airflow limitation as an additional motivator for smoking cessation intervention may be considered. A two-step case-finding method is potentially feasible.

Guava seeds are produced as a waste product by the guava processing industry. Their high carbohydrate contents may suit the carbohydrate needs of the feed sector but their high dietary fiber content limits their feed value. The feed values of fruit seeds can be improved through germination, which involves the mobilization of nutrients through seed enzymes and alters the seed carbohydrate composition. The changes of selected carbohydrates in guava ( Psidium guajava L.) seeds brought by germination to those in red bean ( Vigna angularis ) and winter wheat ( Triticum aestivum L.) were compared. The contents of soluble carbohydrates, digestible starch, resistant starch and cellulose in the seeds were determined. The radial diffusion method was used to detect carbohydrate-degrading enzymes in the seed extracts. Guava seeds were rich in cellulose (402.2 mg/g), which decreased progressively during germination, probably through the action of cellulase. Winter wheat contained the highest starch content (412.2 mg/g) and also distinct quantities of α-amylase and cellulase. The starch contents of all the seeds decreased, but the soluble carbohydrate contents in red beans and guava seeds increased significantly by the end of germination, suggesting the transient oversupply of reserve metabolites. The content of hydrolyzed polysaccharides increased in the germinated seeds with detectable amounts of cellulose-degrading enzymes present, indicating improved value as feed. Further research is warranted to explore the potential of guava seeds as a source of low-cost animal feed supplements.

This paper presents an intuitive nose surgery planning and simulation system, using 3D laser scan image and lateral X-ray image, to provide high quality prediction of the postoperative appearance, and design the patient specific prosthesis model automatically. After initial registration, the internal surface of soft tissue at the nose region was generated by the statistical data for soft tissue thickness adapted by the individual thickness information from the X-ray image. Then, the sketch contour of the 3D scan data on the lateral X-ray image was modified manually or adjusted automatically according to some aesthetic statistical data, to drive the simulation in real time by the state-of-the-art Laplacian surface deformation method. When satisfied with the 3D postoperative appearance, the deformation was mapped to the internal surface of soft tissue, and the change before and after simulation was utilized to generate the patient specific prosthesis model automatically. The surgeons who used the system confirmed that this planning system is attractive and has potential for daily clinical practice., (Copyright © 2010 Elsevier Ireland Ltd. All rights reserved.)

Differential evolution (DE) is a simple, yet efficient, evolutionary algorithm for global numerical optimization, which has been widely used in many areas. However, the choice of the best mutation strategy is difficult for a specific problem. To alleviate this drawback and enhance the performance of DE, in this paper, we present a family of improved DE that attempts to adaptively choose a more suitable strategy for a problem at hand. In addition, in our proposed strategy adaptation mechanism (SaM), different parameter adaptation methods of DE can be used for different strategies. In order to test the efficiency of our approach, we combine our proposed SaM with JADE, which is a recently proposed DE variant, for numerical optimization. Twenty widely used scalable benchmark problems are chosen from the literature as the test suit. Experimental results verify our expectation that the SaM is able to adaptively determine a more suitable strategy for a specific problem. Compared with other state-of-the-art DE variants, our approach performs better, or at least comparably, in terms of the quality of the final solutions and the convergence rate. Finally, we validate the powerful capability of our approach by solving two real-world optimization problems.

Protein-protein interactions (PPIs) are frequently mediated by the binding of a modular domain in one protein to a short, linear peptide motif in its partner. The advent of proteomic methods such as peptide and protein arrays has led to the accumulation of a wealth of interaction data for modular interaction domains. Although several computational programs have been developed to predict modular domain-mediated PPI events, they are often restricted to a given domain type. We describe DomPep, a method that can potentially be used to predict PPIs mediated by any modular domains. DomPep combines proteomic data with sequence information to achieve high accuracy and high coverage in PPI prediction. Proteomic binding data were employed to determine a simple yet novel parameter Ligand-Binding Similarity which, in turn, is used to calibrate Domain Sequence Identity and Position-Weighted-Matrix distance, two parameters that are used in constructing prediction models. Moreover, DomPep can be used to predict PPIs for both domains with experimental binding data and those without. Using the PDZ and SH2 domain families as test cases, we show that DomPep can predict PPIs with accuracies superior to existing methods. To evaluate DomPep as a discovery tool, we deployed DomPep to identify interactions mediated by three human PDZ domains. Subsequent in-solution binding assays validated the high accuracy of DomPep in predicting authentic PPIs at the proteome scale. Because DomPep makes use of only interaction data and the primary sequence of a domain, it can be readily expanded to include other types of modular domains.

We present a new preprocessing method, PeakSelect, to improve the accuracy and efficiency of Tandem Mass-Spec peptide (protein) identification. The fundamental difference between noise and fragment ions in spectra is that ions have isotopes but noise does not. We propose a new and important concept of an Isotope Pattern Vector (IPV) which characterizes the isotope cluster of fragment ions. Then the noise and real peaks can be distinguished by the quantitative IPV values. PeakSelect first uses a new method of the Gaussian Mixture Model and Expectation-Maximization (EM) algorithm to find the base intensity level (baseline) in a spectrum. Then PeakSelect selects features based on the IPV and baseline, and constructs a decision tree to automatically classify the peaks into different categories such as noise, single ion peaks, and overlapping peaks. Experiments show that PeakSelect can help to reduce the Mascot searching time and increase the reliability of peptide identifications. In particular, PeakSelect performs well on complex spectra with a large number of peaks from large peptides, and supports more sequence identification than other well-known systems.

We propose a novel classification learning method called customized support pattern learner (CSPL). Given an instance to be classified, CSPL explores and discovers support patterns (SPs), which are essentially attribute value subsets of the instance to be classified. The final prediction of the class label is performed by combining some statistics of the discovered useful SPs. One advantage of the CSPL method is that it can explore a richer hypothesis space and discover useful classification patterns involving attribute values with almost indistinguishable information gain. The customized learning characteristic also allows that the target class can vary for different instances to be classified. It facilitates extremely easy training instance maintenance and updates. We have evaluated our method with real-world problems and benchmark data sets. The results demonstrate that CSPL can achieve good performance and high reliability.

Unlabelled: A software package, IndexToolkit, aimed at overcoming the disadvantage of FASTA-format databases for frequent searching, is developed to utilize an indexing strategy to substantially accelerate sequence queries. IndexToolkit includes user-friendly tools and an Application Programming Interface (API) to facilitate indexing, storage and retrieval of protein sequence databases. As open source, it provides a sequence-retrieval developing framework, which is easily extensible for high-speed-request proteomic applications, such as database searching or modification discovering. We applied IndexToolkit to database searching engine pFind to demonstrate its effect. Experimental studies show that IndexToolkit is able to support significantly faster searches of protein database., Availability: The IndexToolkit is free to use under the open source GNU GPL license. The source code and the compiled binary can be freely accessed through the website http://pfind.jdl.ac.cn/IndexToolkit. In this website, the more detailed information including screenshots and documentations for users and developers is also available.

Summary: Research in proteomics requires powerful database-searching software to automatically identify protein sequences in a complex protein mixture via tandem mass spectrometry. In this paper, we describe a novel database-searching software system called pFind (peptide/protein Finder), which employs an effective peptide-scoring algorithm that we reported earlier. The pFind server is implemented with the C++ STL, .Net and XML technologies. As a result, high speed and good usability of the software are achieved.

Motivation: The correlation among fragment ions in a tandem mass spectrum is crucial in reducing stochastic mismatches for peptide identification by database searching. Until now, an efficient scoring algorithm that considers the correlative information in a tunable and comprehensive manner has been lacking., Results: This paper provides a promising approach to utilizing the correlative information for improving the peptide identification accuracy. The kernel trick, rooted in the statistical learning theory, is exploited to address this issue with low computational effort. The common scoring method, the tandem mass spectral dot product (SDP), is extended to the kernel SDP (KSDP). Experiments on a dataset reported previously demonstrate the effectiveness of the KSDP. The implementation on consecutive fragments shows a decrease of 10% in the error rate compared with the SDP. Our software tool, pFind, using a simple scoring function based on the KSDP, outperforms two SDP-based software tools, SEQUEST and Sonar MS/MS, in terms of identification accuracy., Supplementary Information: http://www.jdl.ac.cn/user/yfu/pfind/index.html

Supporters of eliminative connectionism have argued for a pattern association-based explanation of language learning and language processing. They deny that explicit rules and symbolic representations play any role in language processing and cognition in general. Their argument is based to a large extent on two artificial neural network (ANN) models that are claimed to be able to learn the past tenses of English verbs (Rumelhart & McClelland, 1986, Parallel distributed processing, Vol. 2, Cambridge, MA: MIT Press; MacWhinney & Leinbach, 1991, Cognition, 40, 121-157). In this article we critically review Rumelhart and McClelland's as well as MacWhinney and Leinbach's ANN models and conclude that they do not succeed in the assigned task of learning the past tenses of English verbs. In order to answer their challenge to the symbolic processing approach, we present our symbolic pattern associator (SPA)-a general-purpose pattern associator that can learn to associate arbitrary discrete patterns. We carried out several experiments with the SPA using the same set of verbs that was used in MacWhinney and Leinbach's simulation with more realistic training and testing procedures. The SPA outperformed the connectionist models by a wide margin in the accuracy of learning, and successful inductive generalizations to unseen verbs. Our SPA has very natural and psychologically realistic explanations to many psychological effects such as U-shaped learning curve, and is much closer to human subjects in predicting past tense of the pseudo-verbs. In contrast to ANNs, whose internal representations are entirely opaque, the SPA can represent the acquired knowledge in the form of production rules that allow for further higher-level processing and integration, resulting in linguistically realistic associative templates for irregular verbs and production rules for regular verbs. In the light of these findings, we conclude that eliminative connectionists' vision of cognition as simple pattern association and pattern recognition without symbolic representation is inadequate. Pattern association as such does not imply rule-less or cue-based models of language acquisition or of human learning in general.