RESUMEN
The prediction of molecular properties is a crucial aspect in drug discovery that can save a lot of money and time during the drug design process. The use of machine learning methods to predict molecular properties has become increasingly popular in recent years. Despite advancements in the field, several challenges remain that need to be addressed, like finding an optimal pre-training procedure to improve performance on small datasets, which are common in drug discovery. In our paper, we tackle these problems by introducing Relative Molecule Self-Attention Transformer for molecular representation learning. It is a novel architecture that uses relative self-attention and 3D molecular representation to capture the interactions between atoms and bonds that enrich the backbone model with domain-specific inductive biases. Furthermore, our two-step pretraining procedure allows us to tune only a few hyperparameter values to achieve good performance comparable with state-of-the-art models on a wide selection of downstream tasks.
RESUMEN
Modern construction materials, including steels, have to combine strength with good formability. In metallic materials, these features are obtained for heterogeneous multiphase microstructures. Design of such microstructures requires advanced numerical models. It has been shown in our earlier works that models based on stochastic internal variables meet this requirement. The focus of the present paper is on deterministic and stochastic approaches to modelling hot deformation of multiphase steels. The main aim was to survey recent advances in describing the evolution of dislocations and grain size accounting for the stochastic character of the recrystallization. To present a path leading to this objective, we reviewed several papers dedicated to the application of internal variables and statistical approaches to modelling recrystallization. Following this, the idea of the model with dislocation density and grain size being the stochastic internal variables is described. Experiments composed of hot compression of cylindrical samples are also included for better presentation of the utility of this approach. Firstly, an empirical data describing the loads as a function of time during compression and data needed to create histograms of the austenite grain size after the tests were collected. Using the measured data, identification and validation of the models were performed. To present possible applications of the model, it was used to produce a simulation imitating industrial hot-forming processes. Finally, calculations of the dislocation density and the grain size distribution were utilized as inputs in simulations of phase transformations during cooling. Distributions of the ferrite volume fraction and the ferrite grain size after cooling recapitulate the paper. This should give readers good overview on the application of collected equations in practice.
RESUMEN
BACKGROUND: In 2006, the Global Alliance against Chronic Respiratory Diseases (GARD) was established. The GARD strategy is based mainly on activities aimed at implementing preventive tasks at a local and global level. In Poland, the National Health Program (NHP) is the strategic document describing the tasks of governmental and local administration in the field of public health. One of the activities under the NHP is a project to carry out mass allergy screening (of children and adolescents at school age) and to organize an information and education campaign. METHODS: In 2017-2018, the screening will cover a group of at least 10,000 children and adolescents at school age (6-18 years of age) in 10 of 16 voivodships in Poland. In the study, the e-health tool for early pre-medical risk assessment of allergic diseases was used. The algorithm determines the risk of bronchial asthma (BA) and allergic rhinitis (AR) based on a series of 38 questions on the symptoms of allergic diseases. In order to assess the effectiveness of the screening test (algorithm) a validation study was carried out before screening. Moreover, the algorithm was calibrated on the basis of the obtained results. The screening is accompanied by a nationwide information and education campaign carried out by means of new media. RESULTS: A total of 1,008 children and adolescents participated in the validation study. In outpatient examination AR was diagnosed in 46.4% and BA in 11.2%. In the case of AR, the sensitivity of the calibrated version of the algorithm was 0.852 and the specificity was 0.840. In the case of BA, it was 0.841 and 0.912, respectively. In 2017, 1,512 people used the screening tool, of which 1,472 respondents went through all stages of the assessment. BA positive AR result (probably sick) was found in 19.5% of respondents and BA in 8.4%. CONCLUSIONS: The use of e-health tools in mass screening and new media to conduct information and education campaigns allows for the reduction of costs and for efficient implementation of activities. It is important to popularize the use of this type of solutions both at national and local level.