SANTM: Efficient Self-attention-driven Network for Text Matching

Self-attention mechanisms have recently been embraced for a broad range of text-matching applications. Self-attention model takes only one sentence as an input with no extra information, i.e., one can utilize the final hidden state or pooling. However, text-matching problems can be interpreted either in symmetrical or asymmetrical scopes. For instance, paraphrase detection is an asymmetrical task, while textual entailment classification and question-answer matching are considered asymmetrical tasks. In this article, we leverage attractive properties of self-attention mechanism and proposes an attention-based network that incorporates three key components for inter-sequence attention: global pointwise features, preceding attentive features, and contextual features while updating the rest of the components. Our model follows evaluation on two benchmark datasets cover tasks of textual entailment and question-answer matching. The proposed efficient Self-attention-driven Network for Text Matching outperforms the state of the art on the Stanford Natural Language Inference and WikiQA datasets with much fewer parameters.

Ensemble Data Analytics Approaches for Fast Parametrization Screening and Validation

Ensemble approaches are increasingly used for history matching also with large scale models. However, the iterative nature and the high computational resources required, demands careful and consistent parameterization of the initial ensemble of models, to avoid repeated and time-consuming attempts before an acceptable match is achieved. The objective of this work is to introduce ensemble-based data analytic techniques to validate the starting ensemble and early identify potential parameterization problems, with significant time saving. These techniques are based on the same definition of the mismatch between the initial ensemble simulation results and the historical data used by ensemble algorithms. In fact, a notion of distance among ensemble realizations can be introduced using the mismatch, opening the possibility to use statistical analytic techniques like Multi-Dimensional Scaling and Generalized Sensitivity. In this way a clear and immediate view of ensemble behavior can be quickly explored. Combining these views with advanced correlation analysis, a fast assessment of ensemble consistency with observed data and physical understanding of the reservoir is then possible. The application of the proposed methodology to real cases of ensemble history matching studies, shows that the approach is very effective in identifying if a specific initial ensemble has an adequate parameterization to start a successful computational loop of data assimilation. Insufficient variability, due to a poor capturing of the reservoir performance, can be investigated both at field and well scales by data analytics computations. The information contained in ensemble mismatches of relevant quantities like water-breakthrough and Gas-Oil-ratio is then evaluated in a systematic way. The analysis often reveals where and which uncertainties have not enough variability to explain historical data. It also allows to detect what is the role of apparently inconsistent parameters. In principle it is possible to activate the heavy iterative computation also with an initial ensemble where the analytics tools show potential difficulties and problems. However, experiences with large scale models point out that the possibility to obtain a good match in these situations is very low, leading to a time-consuming revision of the entire process. On the contrary, if the ensemble is validated, the iterative large-scale computations achieve a good calibration with a consistency that enables predictive ability. As a new interesting feature of the proposed methodology, ensemble advanced data analytics techniques are able to give clues and suggestions regarding which parameters could be source of potential history matching problems in advance. In this way it is possible anticipate directly on initial ensemble the uncertainties revision for example modifying ranges, introducing new parameters and better tuning other ensemble factors, like localization and observations tolerances that controls the ultimate match quality.

Phase transition in the bipartite z-matching

We study numerically the maximum z-matching problems on ensembles of bipartite random graphs. The z-matching problems describes the matching between two types of nodes, users and servers, where each server may serve up to z users at the same time. Using a mapping to standard maximum-cardinality matching, and because for the latter there exists a polynomial-time exact algorithm, we can study large system sizes of up to $$10^6$$ 10 6 nodes. We measure the capacity and the energy of the resulting optimum matchings. First, we confirm previous analytical results for bipartite regular graphs. Next, we study the finite-size behaviour of the matching capacity and find the same scaling behaviour as before for standard matching, which indicates the universality of the problem. Finally, we investigate for bipartite Erdős–Rényi random graphs the saturability as a function of the average degree, i.e. whether the network allows as many customers as possible to be served, i.e. exploiting the servers in an optimal way. We find phase transitions between unsaturable and saturable phases. These coincide with a strong change of the running time of the exact matching algorithm, as well with the point where a minimum-degree heuristic algorithm starts to fail. Graphical Abstract

Graphene/AlGaN/GaN RF Switch

RF switches, which use a combination of graphene and two-dimensional high-density electron gas (2DEG) in the AlGaN/GaN system, were proposed and studied in the frequency band from 10 MHz to 114.5 GHz. The switches were integrated into the coplanar waveguide, which allows them to be used in any system without the use of, e.g., bonding, flip-chip and other technologies and avoiding the matching problems. The on-state insertion losses for the designed switches were measured to range from 7.4 to 19.4 dB, depending on the frequency and switch design. Although, at frequencies above 70 GHz, the switches were less effective, the switching effect was still evident with an approximately 4 dB on–off ratio. The best switches exhibited rise and fall switching times of ~25 ns and ~17 ns, respectively. The use of such a switch can provide up to 20 MHz of bandwidth in time-modulated systems, which is an outstanding result for such systems. The proposed equivalent circuit describes well the switching characteristics and can be used to design switches with required parameters.

A Two-sided Matching Decision-making Approach based on PROMETHEE Under the Probabilistic Linguistic Environment

Matching problems in daily life can be effectively solved by two-sided matching decision-making (TSMDM) approaches. The involved matching intermediary is to match two sides of subjects. This paper proposes a TSMDM approach based on preference ranking organization method (PROMETHEE) under the probabilistic linguistic environment. The probabilistic linguistic evaluations are firstly normalized and transformed to the benefit types. Then, the preference degrees of a subject over other subjects from the same side are obtained by using six types of preference function. Afterwards, groups of preference degrees of a subject are aggregated to the preference indexes by considering the weights of criteria. Hereafter, the preference degrees of a subject over other subjects from the same side are aggregated to the outgoing flow, while the preference degrees of other subjects from the same side over this subject are aggregated to the incoming flow. Furthermore, the net-flows, which is recognized as the satisfaction degrees are calculated by using outgoing flows to minus incoming flows. On the basis of this, the multi-objectives TSMDM model is built by considering the matching aspirations. A model with respect to the matching aspirations is built and solved by using the Lagrange function. The multi-objectives TSMDM model is further transformed to the single-objective model, the solution of which is the matching scheme. A matching problem related to the intelligent technology intermediary is solved to verify the effectiveness and the feasibility of the proposed approach.

Implementasi Deep Learning untuk Entity Matching pada Dataset Obat (Studi Kasus K24 dan Farmaku)

Data processing speed in companies is important to speed up their analysis. Entity matching is a computational process that companies can perform in data processing. In conducting data processing, entity matching plays a role in determining two different data but referring to the same entity. Entity matching problems arise when the dataset used in the comparison is large. The deep learning concept is one of the solutions in dealing with entity matching problems. DeepMatcher is a python package based on a deep learning model architecture that can solve entity matching problems. The purpose of this study was to determine the matching between the two datasets with the application of DeepMatcher in entity matching using drug data from farmaku.com and k24klik.com. The comparison model used is the Hybrid model. Based on the test results, the Hybrid model produces accurate numbers, so that the entity matching used in this study runs well. The best accuracy value of the 10th training with an F1 value of 30.30, a precision value of 17.86, and a recall value of 100.

Neuropsychological Functioning in Users of Serotonergic Psychedelics – A Systematic Review and Meta-Analysis

Background: Serotonergic psychedelics (SPs) like LSD, psilocybin, DMT, and mescaline are a heterogeneous group of substances that share agonism at 5-HT2a receptors. Besides the ability of these substances to facilitate profoundly altered states of consciousness, persisting psychological effects have been reported after single administrations, which outlast the acute psychedelic effects. In this review and meta-analysis, we investigated if repeated SP use associates with a characteristic neuropsychological profile indicating persisting effects on neuropsychological function.Methods: We conducted a systematic review of studies investigating the neuropsychological performance in SP users, searching studies in Medline, Web of Science, embase, ClinicalTrials.gov, and EudraCT. Studies were included if they reported at least one neuropsychological measurement in users of SPs. Studies comparing SP users and non-users that reported mean scores and standard deviations were included in an exploratory meta-analysis.Results: 13 studies (N = 539) published between 1969 and 2020 were included in this systematic review. Overall, we found that only three SPs were specifically investigated: ayahuasca (6 studies, n = 343), LSD (5 studies, n = 135), and peyote (1 study, n = 61). However, heterogeneity of the methodological quality was high across studies, with matching problems representing the most important limitation. Across all SPs, no uniform pattern of neuropsychological impairment was identified. Rather, the individual SPs seemed to be associated with distinct neuropsychological profiles. For instance, one study (n = 42) found LSD users to perform worse in trials A and B of the Trail-Making task, whereas meta-analytic assessment (5 studies, n = 352) of eleven individual neuropsychological measures indicated a better performance of ayahuasca users in the Stroop incongruent task (p = 0.03) and no differences in the others (all p > 0.05).Conclusion: The majority of the included studies were not completely successful in controlling for confounders such as differences in non-psychedelic substance use between SP-users and non-users. Our analysis suggests that LSD, ayahuasca and peyote may have different neuropsychological consequences associated with their use. While LSD users showed reduced executive functioning and peyote users showed no differences across domains, there is some evidence that ayahuasca use is associated with increased executive functioning.