Tag Datasets Research Articles

Jet Tagging with More-Interaction Particle Transformer

Abstract In this study, we introduce the More-Interaction Particle Transformer (MIParT), a novel deep learning neural network designed for jet tagging. This framework incorporates our own design, the More-Interaction Attention (MIA) mechanism, which increases the dimensionality of particle interaction embeddings. We tested MIParT using the top tagging and quark-gluon datasets. Our results show that MIParT not only matches the accuracy and AUC of LorentzNet and a series of Lorentz-equivariant methods, but also significantly outperforms the ParT model in background rejection. Specifically, it improves background rejection by approximately 25\% at a 30\% signal efficiency on the top tagging dataset and by 3\% on the quark-gluon dataset. Additionally, MIParT requires only 30\% of the parameters and 53\% of the computational complexity needed by ParT, proving that high performance can be achieved with reduced model complexity. For very large datasets, we double the dimension of particle embeddings, referring to this variant as MIParT-Large (MIParT-L). We find that MIParT-L can further capitalize on the knowledge from large datasets. From a model pre-trained on the 100M JetClass dataset, the background rejection performance of the fine-tuned MIParT-L improved by 39\% on the top tagging dataset and by 6\% on the quark-gluon dataset, surpassing that of the fine-tuned ParT. Specifically, the background rejection of fine-tuned MIParT-L improved by an additional 2\% compared to the fine-tuned ParT. The results suggest that MIParT has the potential to advance efficiency benchmarks for jet tagging and event identification in particle physics.Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. Article funded by SCOAP3 and published under licence by Chinese Physical Society and the Institute of High Energy Physics of the Chinese Academy of Science and the Institute of Modern Physics of the Chinese Academy of Sciences and IOP Publishing Ltd

Towards an improved of teaching practice using Sentiment Analysis in Student Evaluation

Student evaluation of teaching (SET) is an ad-hoc way of assessing teaching effectiveness in higher education institutions. In this paper, we present an approach to analyzing sentiments expressed in SET comments using a large language model (LLM). By employing natural language processing techniques, we extract and analyze sentiments expressed by students when the course has ended, aiming to provide educators and administrators with valuable insights into teaching quality and elements to improve teaching practice. Our study demonstrates the effectiveness of LLMs in sentiment analysis of comments, highlighting their potential to enhance the evaluation process. Our experiments with a crowdsourced tagged dataset show a 93% of accuracy in the classification of feedback messages. We discuss the implications of our findings for educational institutions and propose future directions for research in this domain.

Studying animal locomotion with multiple data loggers: quantifying time drift between tags

Temporal accuracy is a fundamental characteristic of logging technology and is needed to correlate data streams. Single biologgers sensing animal movement (accelerometers, gyroscope, magnetometers, collectively inertial measurement unit; IMU) have been extensively used to study the ecology of animals. To better capture whole body movement and increase the accuracy of behavior classification, there is a need to deploy multiple loggers on a single individual to capture the movement of multiple body parts. Yet due to temporal drift, accurately aligning multiple IMU datasets can be problematic, especially as deployment duration increases. In this paper we quantify temporal drift and errors in commercially available IMU data loggers using a combination of robotic and animal borne experiments. The variance in drift rate within a tag is over an order of magnitude lower (σ = 0.001 s h−1) than the variance between tags (σ = 0.015 s·h−1), showing that recording frequency is a characteristic of each tag and not a random variable. Furthermore, we observed a large offset (0.54 ± 0.016 s·h−1) between two groups of tags that had differing recording frequencies, and we observed three instances of instantaneous temporal jumps within datasets introducing errors into the data streams. Finally, we show that relative drift rates can be estimated even when deployed on animals displaying various behaviors without the tags needing to be simultaneously moved. For the tags used in this study, drift rates can vary significantly between tags, are repeatable, and can be accurately measured in the field. The temporal alignment of multiple tag datasets allows researchers to deploy multiple tags on an individual animal which will greatly increase our knowledge of movement kinematics and expand the range of movement characteristics that can be used for behavioral classification.

CEPTNER: Contrastive learning Enhanced Prototypical network for Two-stage few-shot Named Entity Recognition

Few-shot Named Entity Recognition (NER) systems aim to classify unseen named entity types with limited labeled examples. Significant progress has been made in the use of large-scale pre-trained language models. However, boundary information and hidden relationships between entities beyond the sequence, which provide additional information and play crucial roles in few-shot NER, have received little attention in recent methods that achieve the state of the art. In this paper, we propose CEPTNER, a Contrastive learning Enhanced Prototypical network for Two-stage few-shot Named Entity Recognition, which leverages meta-learning and prototypical network to identify unseen entity types with limited labeled data. Concretely, we first detect candidate boundaries of entities in the stage of boundary detector, then we employ a prototypical network to filter false boundaries and type for the remaining spans in the stage of entity classifier. Additionally, we conduct entity-level contrastive learning to explore the internal relationship between entities that provides extra information while optimizing the prototypical network. CEPTNER is evaluated on two widely-used few-shot NER datasets and a few-shot slot tagging dataset: Few-NERD, CrossNER and SNIPS. Extensive experiments on various benchmarks show the superiority of CEPTNER over previous methods for few-shot NER.

Ontologies for increasing the FAIRness of plant research data

The importance of improving the FAIRness (findability, accessibility, interoperability, reusability) of research data is undeniable, especially in the face of large, complex datasets currently being produced by omics technologies. Facilitating the integration of a dataset with other types of data increases the likelihood of reuse, and the potential of answering novel research questions. Ontologies are a useful tool for semantically tagging datasets as adding relevant metadata increases the understanding of how data was produced and increases its interoperability. Ontologies provide concepts for a particular domain as well as the relationships between concepts. By tagging data with ontology terms, data becomes both human- and machine- interpretable, allowing for increased reuse and interoperability. However, the task of identifying ontologies relevant to a particular research domain or technology is challenging, especially within the diverse realm of fundamental plant research. In this review, we outline the ontologies most relevant to the fundamental plant sciences and how they can be used to annotate data related to plant-specific experiments within metadata frameworks, such as Investigation-Study-Assay (ISA). We also outline repositories and platforms most useful for identifying applicable ontologies or finding ontology terms.

Hierarchical high-point Energy Flow Network for jet tagging

Jet substructure observable basis is a systematic and powerful tool for analyzing the internal energy distribution of constituent particles within a jet. In this work, we propose a novel method to insert neural networks into jet substructure basis as a simple yet efficient interpretable IRC-safe deep learning framework to discover discriminative jet observables. The Energy Flow Polynomial (EFP) could be computed with a certain summation order, resulting in a reorganized form which exhibits hierarchical IRC-safety. Thus inserting non-linear functions after the separate summation could significantly extend the scope of IRC-safe jet substructure observables, where neural networks can come into play as an important role. Based on the structure of the simplest class of EFPs which corresponds to path graphs, we propose the Hierarchical Energy Flow Networks and the Local Hierarchical Energy Flow Networks. These two architectures exhibit remarkable discrimination performance on the top tagging dataset and quark-gluon dataset compared to other benchmark algorithms even only utilizing the kinematic information of constituent particles.

A Multi-Scale Hybrid Attention Network for Sentence Segmentation Line Detection in Dongba Scripture

Dongba scripture sentence segmentation is an important and basic work in the digitization and machine translation of Dongba scripture. Dongba scripture sentence segmentation line detection (DS-SSLD) as a core technology of Dongba scripture sentence segmentation is a challenging task due to its own distinctiveness, such as high inherent noise interference and nonstandard sentence segmentation lines. Recently, projection-based methods have been adopted. However, these methods are difficult when dealing with the following two problems. The first is the noisy problem, where a large number of noise in the Dongba scripture image interference detection results. The second is the Dongba scripture inherent characteristics, where many vertical lines in Dongba hieroglyphs are easily confused with the vertical sentence segmentation lines. Therefore, this paper aims to propose a module based on the convolutional neural network (CNN) to improve the accuracy of DS-SSLD. To achieve this, we first construct a tagged dataset for training and testing DS-SSLD, including 2504 real images collected from Dongba scripture books and sentence segmentation targets. Then, we propose a multi-scale hybrid attention network (Multi-HAN) based on YOLOv5s, where a multiple hybrid attention unit (MHAU) is used to enhance the distinction between important features and redundant features, and the multi-scale cross-stage partial unit (Multi-CSPU) is used to realize multi-scale and richer feature representation. The experiment is carried out on the Dongba scripture sentence segmentation dataset we built. The experimental results show that the proposed method exhibits excellent detection performance and outperforms several state-of-the-art methods.

Tissue-based absolute quantification using large-scale TMT and LFQ experiments.

Relative and absolute intensity-based protein quantification across cell lines, tissue atlases and tumour datasets is increasingly available in public datasets. These atlases enable researchers to explore fundamental biological questions, such as protein existence, expression location, quantity and correlation with RNA expression. Most studies provide MS1 feature-based label-free quantitative (LFQ) datasets; however, growing numbers of isobaric tandem mass tags (TMT) datasets remain unexplored. Here, we compare traditional intensity-based absolute quantification (iBAQ) proteome abundance ranking to an analogous method using reporter ion proteome abundance ranking with data from an experiment where LFQ and TMT were measured on the same samples. This new TMT method substitutes reporter ion intensities for MS1 feature intensities in the iBAQ framework. Additionally, we compared LFQ-iBAQ values to TMT-iBAQ values from two independent large-scale tissue atlas datasets (one LFQ and one TMT) using robust bottom-up proteomic identification, normalisation and quantitation workflows.

Informing Management of Atlantic Bluefin Tuna Using Telemetry Data

Sustainable management of exploited marine fish and wildlife populations requires knowledge about their productivity. Survival from natural causes of mortality is a key component of population productivity, but is notoriously difficult to estimate. We evaluate the potential for acoustic telemetry as a long-term monitoring tool to estimate rates of natural mortality. We present a Bayesian multistate mark-recapture model for telemetry data collected over a decade from 188 Atlantic bluefin tuna (Thunnus thynnus) and apply it to estimate the rate of natural mortality using only acoustic tag detections for all animals, or using acoustic tag detections for 96 single-tagged tuna plus acoustic tag detections combined with estimated positions from pop-up satellite archival tags for 92 double-tagged animals. We support the model for bluefin tuna with a simulation study to quantify bias in estimates of population dynamics parameters and investigate the effect of auxiliary information from satellite tagging on mortality rate estimates for different acoustic tag detection probability scenarios.We obtained posterior estimates of the instantaneous annual natural mortality (survival) rate across a decade of tagging for Atlantic bluefin tuna of 0.17 yr-1 (0.84 yr-1) both using only acoustic tagging data, and using a combination of acoustic and satellite tagging data. Use of a prior implying a higher rate of fishing mortality yielded an instantaneous annual natural mortality (survival) estimate of 0.10 yr-1 (0.90 yr-1), with combined acoustic and satellite tag data. Results from the simulation study indicate that the use of satellite tags can improve the precision and accuracy of estimates of detection probabilities, area-specific movement probabilities and mortality rates, where the extent of the improvement depends on true underlying acoustic tag detection probabilities. Our work demonstrates that long-term acoustic tagging data sets have strong potential for monitoring of highly migratory marine fish and wildlife populations, providing information on a number of key parameters, including survival and movement rates. However, improved information on tag reporting rates or fishing mortality is needed to better separate natural and fisheries mortality for Atlantic bluefin tuna.

Times are changing, but has natural mortality? Estimation of mortality rates for tropical tunas in the western and central Pacific Ocean

A variety of density-dependent and -independent processes have been proposed to influence natural mortality rates, potentially leading to variation through time. Processes of natural mortalities are rarely directly observed, making estimation of natural mortality rates difficult. Mark-recapture data allow estimation of total mortality rates, which can be separated into natural and fishing mortality with information on rates of tag reporting, tag shedding and tag-induced mortality. We fitted attrition models and length-based Brownie models to four decades of mark-recapture data from skipjack and yellowfin tuna in the Western and Central Pacific Ocean, a period representing a sustained expansion of associated fisheries in the region as well as rapid changes to the marine environment. The modelled dataset included c. 250,000 skipjack and 100,000 yellowfin tag releases, with 45,000 and 17,000 recoveries of skipjack and yellowfin respectively, released from 1977 to 2017. Increases in fishing mortality were detected over this time for both skipjack and yellowfin, with evidence of temporal changes in selectivity for yellowfin. Estimates of natural mortality were highest for the smallest size class and generally lower for larger sizes, though there was large uncertainty in the largest size groups due to lower sample sizes of tagged fish. There was no clear evidence of temporal changes in natural mortality rates for either species, though there was some evidence of changes in natural mortality for the smaller yellowfin size classes (< 61 cm). However, there was likely insufficient statistical power to test for plausible changes in natural mortality rates for yellowfin due to low precision of estimates during the earlier years of the tag dataset.

Hope speech detection in YouTube comments

Recent work on language technology has tried to recognize abusive language such as those containing hate speech and cyberbullying and enhance offensive language identification to moderate social media platforms. Most of these systems depend on machine learning models using a tagged dataset. Such models have been successful in detecting and eradicating negativity. However, an additional study has lately been conducted on the enhancement of free expression through social media. Instead of eliminating ostensibly unpleasant words, we created a multilingual dataset to recognize and encourage positivity in the comments, and we propose a novel custom deep network architecture, which uses a concatenation of embedding from T5-Sentence. We have experimented with multiple machine learning models, including SVM, logistic regression, K-nearest neighbour, decision tree, logistic neighbours, and we propose new CNN based model. Our proposed model outperformed all others with a macro F1-score of 0.75 for English, 0.62 for Tamil, and 0.67 for Malayalam.

Perspectives and Challenges of AI Techniques in the Field of Social Sciences and Communication

&lt;div&gt;&lt;p class="4"&gt;In the past decade, the methods and technologies of artificial intelligence (AI) have made great progress. In many cases, they have become part of the usual landscape of solving new or old problems in different fields of human knowledge. In this progress, there are several aspects, especially three aspects: the availability and universality of data in many fields of human activities; a deeper understanding of the mathematics of the basic control algorithm; and the availability and capability of hardware and computing which allows a wide range and a large number of data experiments. Considering these aspects, the key challenge for each problem and application area is to understand how to use these technologies, to what extent they may reach, and what constraints need to be overcome in order to obtain beneficial results (in terms of production cost, value, etc.). This challenge includes identifying data sources and their integration and recovery requirements, the necessity and cost of acquiring or constructing tag data sets, volume data required for measurement, verifying its feasibility, technical method of data analysis task and its consistency with the final application goal, and social and communication sciences are no exception. The knowledge in these fields is related to artificial intelligence, but they do have particularities that define the most appropriate type of artificial intelligence technology and method (i.e. natural language processing). The successful use of AI technology in these disciplines involves not only technical knowledge, but also the establishment of a viable application environment, including the availability of data, the appropriate complexity of tasks to be performed, and verification procedures with experts in the field. This paper introduces the methodology of generating artificial intelligence model, summarizes the artificial intelligence methods and services most likely to be used in social and communication sciences, and finally gives some application examples to illustrate the practical and technical considerations in this regard.&lt;/p&gt;&lt;/div&gt;

Characterizing cellular heterogeneity in chromatin state with scCUT&Tag-pro.

Technologies that profile chromatin modifications at single-cell resolution offer enormous promise for functional genomic characterization, but the sparsity of the measurements and integrating multiple binding maps represent substantial challenges. Here we introduce scCUT&Tag-pro, a multimodal assay for profiling protein-DNA interactions coupled with the abundance of surface proteins in single cells. In addition, we introduce scChromHMM, which integrates data from multiple experiments to infer and annotate chromatin states based on combinatorial histone modification patterns. We apply these tools to perform an integrated analysis across nine different molecular modalities in circulating human immune cells. We demonstrate how these two approaches can characterize dynamic changes in the function of individual genomic elements across both discrete cell states and continuous developmental trajectories, nominate associated motifs and regulators that establish chromatin states, and identify extensive and cell type-specific regulatory priming. Finally, we demonstrate how our integrated reference can serve as a scaffold to map and improve the interpretation of additional scCUT&Tag datasets.

Depression Discovery in Covid-19 Communities Using Deep Learning

: Seemingly, in the contemporary era, data is the new currency. Facing the exponential explosion of data through its various online sources, there arises a need for the industry to tap into this source. The art of ascribing sentiment to a piece of text is Sentiment Analysis making it relevant to a wide array of fields. The world has witnessed assortments of a wide family of coronaviruses, 229E and NL63 being the alpha coronaviruses and OC43 &amp; HKU1 being the beta coronaviruses. A version of coronavirus MERS-CoV seemed to be identified in Saudi Arabia. This work focuses on The Deep Neural Network and Natural Language processing to detect depression in COVID-19 patients, and the model is trained with the depression labeled tagged dataset with an accuracy of 98%. The foundation of this work shall be instrumental in aiding the doctors for the timely diagnosis and treatment of their patients.

NGMA-4. Creation of a MADR brain tumor single-cell atlas for examination of inter-/intratumor heterogeneity and the results of genetic perturbations in a diverse array of brain tumor subtypes

We have recently established mosaic analysis by dual recombinase-mediated cassette exchange (MADR), which permits stable labeling of mutant cells expressing transgenic elements from precisely defined chromosomal loci. MADR provides a toolkit of elements for combinatorial labeling, inducible/reversible transgene manipulation, VCre recombinase expression, and genetic manipulation of human cells. Further, we have demonstrated the versatility of MADR by creating glioma models with mixed, reporter-identified zygosity or with “personalized” driver mutations from pediatric glioma. For example, introducing H3f3a (aka H3.3) mutation variants with MADR regulates the spatiotemporal profile of glioma, and single-cell RNA and ATAC sequencing analysis demonstrates a recapitulation of developmental hierarchy seen in K27M mutant human glioma. Moreover, we have generated novel models of H3.3 WT glioma, H3.3 G34R glioma, and supratentorial ependymoma using patient-derived oncofusion transgenes. These models display a high degree of phenotypic fidelity and we now compare these models on a single-cell level with our previous models, mouse single-cell RNA glioma datasets from other studies, and human tumor cell transcriptomes. Our multi-omics approach includes integration of ChIP-seq/Cut&Tag datasets, single-cell ATAC, and single-cell Cut&Tag datasets.Moreover, we have engineered a novel methodology for inducible gain- and loss-of function perturbation studies in vivo. Using ETS transcription factors as a proof-of-principle, we overlay these genetic perturbations on the glioma atlas to examine the gene networks altered by precise molecular manipulations.We hope that these combined approaches will enable researchers to discover disease mechanisms with increased resolution and test therapeutics in credentialed pre-clinical disease models.

Projected shifts in loggerhead sea turtle thermal habitat in the Northwest Atlantic Ocean due to climate change

It is well established that sea turtles are vulnerable to atmospheric and oceanographic shifts associated with climate change. However, few studies have formally projected how their seasonal marine habitat may shift in response to warming ocean temperatures. Here we used a high-resolution global climate model and a large satellite tagging dataset to project changes in the future distribution of suitable thermal habitat for loggerheads along the northeastern continental shelf of the United States. Between 2009 and 2018, we deployed 196 satellite tags on loggerheads within the Middle Atlantic Bight (MAB) of the Northwest Atlantic continental shelf region, a seasonal foraging area. Tag location data combined with depth and remotely sensed sea surface temperature (SST) were used to characterize the species’ current thermal range in the MAB. The best-fitting model indicated that the habitat envelope for tagged loggerheads consisted of SST ranging from 11.0° to 29.7 °C and depths between 0 and 105.0 m. The calculated core bathythermal range consisted of SSTs between 15.0° and 28.0 °C and depths between 8.0 and 92.0 m, with the highest probability of presence occurred in regions with SST between 17.7° and 25.3 °C and at depths between 26.1 and 74.2 m. This model was then forced by a high-resolution global climate model under a doubling of atmospheric CO2 to project loggerhead probability of presence over the next 80 years. Our results suggest that loggerhead thermal habitat and seasonal duration will likely increase in northern regions of the NW Atlantic shelf. This change in spatiotemporal range for sea turtles in a region of high anthropogenic use may prompt adjustments to the localized protected species conservation measures.

Movement patterns and growth rate of the whitespotted wedgefish Rhynchobatus djiddensis in southern Africa based on tag-recapture data

Information on the movement ecology of endangered species is critical for the implementation of effective conservation measures. This study made use of a long-term dart tagging dataset to reveal the movement patterns and growth rates of two size classes of the Critically Endangered whitespotted wedgefish Rhynchobatus djiddensis within its southern African distribution, which can have important implications for fisheries management. A total of 4 768 individuals were tagged with 340 recaptures recorded, ranging from 1 to 2 639 days (7.2 years) at liberty. Most of the tag releases and recaptures occurred within the KwaZulu-Natal central region in South Africa, with catches increasing significantly during summer (October to March). Most recaptures (43%) were recorded within 5 km of the tagging (release) site. Tagged adults recorded significantly greater distances moved than juveniles (p < 0.002) but there was no significant difference between juveniles or adults in terms of their direction of movement (p > 0.30). A Francis growth model showed that smaller individuals had a substantially faster growth rate (198.69 [SE 21.75] mm year−1) compared with larger individuals (57.41 [SE 27.83] mm year−1) confirming that the species is relatively slow-growing. Ultimately, this study identified important knowledge gaps in the broadscale movement patterns of R. djiddensis and provided new information on the growth rate of this Critically Endangered species. Filling in these knowledge gaps will aid in conservation measures for two important size classes of the R. djiddensis population as the species faces increasing targeted fishing pressure.

Stock-of-origin catch estimation of Atlantic bluefin tuna (Thunnus thynnus) based on observed spatial distributions

Atlantic bluefin tuna (Thunnus thynnus) are a large, highly migratory fish distributed throughout the North Atlantic Ocean and adjacent seas currently managed as two discrete stocks: western and eastern. Both stocks forage in the North Atlantic, and a high degree of intermixing occurs, which combined with limited single-stock survey data makes it difficult to assess the abundance and status of individual populations. In this study, we used movement patterns from a multidecadal tagging dataset to create monthly distribution maps for these two major stocks. We then used these maps to separate the overall catch records into stock-specific catch (catch per unit effort, CPUE) time series. We identified an increase in the past two decades in the proportion of catch estimated to come from the eastern stock, attributable to a decrease in CPUE in regions dominated by the western stock, relative to other regions. The stock-specific catch series can be used to improve the accuracy of stock assessments and inform spatial management.

Migrating curlews on schedule: departure and arrival patterns of a long-distance migrant depend on time and breeding location rather than on wind conditions

BackgroundDeparture decisions in long-distance migratory bird species may depend on favourable weather conditions and beneficial resources at the destination location, overarched by genetic triggers. However, few studies have tried to validate the significance of these three concepts simultaneously, and long-term, high-resolution tagging datasets recording individual movements across consecutive years are scarce. We used such a dataset to explore intraspecific and intra-individual variabilities in departure and arrival decisions from/to wintering grounds in relation to these three different concepts in bird migration.MethodsWe equipped 23 curlews (Numenius arquata) wintering in the Wadden Sea with Global Positioning System data loggers to record their spatio-temporal patterns of departure from and arrival at their wintering site, and the first part of their spring migration. We obtained data for 42 migrations over 6 years, with 12 individuals performing repeat migrations in consecutive years. Day of year of departure and arrival was related to 38 meteorological and bird-related predictors using the least absolute shrinkage and selection operator (LASSO) to identify drivers of departure and arrival decisions.ResultsCurlews migrated almost exclusively to Arctic and sub-Arctic Russia for breeding. They left their wintering site mainly during the evening from mid- to late April and returned between the end of June and mid-July. There was no difference in departure times between the sexes. Weather parameters did not impact departure decisions; if departure days coincided with headwind conditions, the birds accounted for this by flying at higher altitudes of up to several kilometres. Curlews breeding further away in areas with late snowmelt departed later. Departures dates varied by only < 4 days in individual curlews tagged over consecutive years.ConclusionsThese results suggest that the trigger for migration in this long-distance migrant is largely independent of weather conditions but is subject to resource availability in breeding areas. The high intra-individual repeatability of departure days among subsequent years and the lack of relationship to weather parameters suggest the importance of genetic triggers in prompting the start of migration. Further insights into the timing of migration in immatures and closely related birds might help to further unravel the genetic mechanisms triggering migration patterns.

MindLink-Eumpy: An Open-Source Python Toolbox for Multimodal Emotion Recognition.

Emotion recognition plays an important role in intelligent human–computer interaction, but the related research still faces the problems of low accuracy and subject dependence. In this paper, an open-source software toolbox called MindLink-Eumpy is developed to recognize emotions by integrating electroencephalogram (EEG) and facial expression information. MindLink-Eumpy first applies a series of tools to automatically obtain physiological data from subjects and then analyzes the obtained facial expression data and EEG data, respectively, and finally fuses the two different signals at a decision level. In the detection of facial expressions, the algorithm used by MindLink-Eumpy is a multitask convolutional neural network (CNN) based on transfer learning technique. In the detection of EEG, MindLink-Eumpy provides two algorithms, including a subject-dependent model based on support vector machine (SVM) and a subject-independent model based on long short-term memory network (LSTM). In the decision-level fusion, weight enumerator and AdaBoost technique are applied to combine the predictions of SVM and CNN. We conducted two offline experiments on the Database for Emotion Analysis Using Physiological Signals (DEAP) dataset and the Multimodal Database for Affect Recognition and Implicit Tagging (MAHNOB-HCI) dataset, respectively, and conducted an online experiment on 15 healthy subjects. The results show that multimodal methods outperform single-modal methods in both offline and online experiments. In the subject-dependent condition, the multimodal method achieved an accuracy of 71.00% in the valence dimension and an accuracy of 72.14% in the arousal dimension. In the subject-independent condition, the LSTM-based method achieved an accuracy of 78.56% in the valence dimension and an accuracy of 77.22% in the arousal dimension. The feasibility and efficiency of MindLink-Eumpy for emotion recognition is thus demonstrated.