Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizens for the clue delivery of the holy Buddha! This article comes from the official account of Wechat: SF Chinese (ID:kexuejiaodian), author: SF

Machine learning is a science that studies how to use computers to simulate or realize human learning activities, and it is the core field of artificial intelligence. Scientists use a lot of human experience to train machine learning models in order to make the models surpass human beings in some way and accomplish things that are difficult or impossible for human beings to accomplish. Recently, scientists have successfully developed a machine learning model that transcends the human eye.

Wen Jing | Wen Jing

Have you ever used sugar as salt or salt as sugar and added the wrong seasoning to make the food taste strange? At the same time, he will be ridiculed by his mother. When you react, take a closer look at the two seasonings, and find that they are really different, and you can even see that because you have been using a spoon to pick up the seasoning, the spoon is covered with both sugar and salt.

Mom didn't taste it before putting the seasoning, why wouldn't she make such a mistake? After a loss, why can you distinguish between the two seasonings with the naked eye? What we are going to talk about below may answer your confusion.

The recognition ability of the human eye is to determine the composition of the solid mixture, we can use nuclear magnetic resonance, high resolution mass spectrometer and spectrometer and so on. These methods can obtain quite accurate quantitative analysis results. However, naked eye observation is still an important method in chemical experiments, which can get preliminary evaluation results, and has the advantages of fast and non-destructive. In history, Louis Pasteur, a famous French microbiologist and chemist, distinguished tartrate by carefully observing the crystal with the naked eye. With this breakthrough as a starting point, scientists have established and developed the theory of molecular chirality, and now molecular chirality has been widely used in the field of drug development.

For the solid mixture, the naked eye observation is mainly based on its color, texture, particle size, transparency to determine what ingredients it contains, as well as the approximate proportion of each component. However, due to the influence of experience and intuition, different researchers may come to different results, which makes the results of naked eye observation contain considerable uncertainty.

The image recognition function of artificial intelligence is to imitate the naked eye observation ability of human beings. Therefore, scientists came up with the idea of using machine learning technology to enable models to learn how to analyze the composition of solid mixtures with the naked eye (especially those with rich knowledge and experience). And get the accuracy beyond the most experienced chemist.

Beyond the recognition of the human eye, researchers have successfully developed a machine learning model that can know the composition of a solid mixture just by taking pictures, according to an article published in the academic journal Industrial & Engineering Chemistry Research.

In the beginning, the researchers used pictures of a mixture of sugar and salt to train machine learning models. Although there are only 300 original photos, the researchers use random cropping, flipping, rotation and other means to process the original photos to create more sub-images for training and testing. The test results showed that the machine learning model was twice as accurate in recognizing these photos as those with the highest visual accuracy in the research team.

The researchers also applied the model to the evaluation of different solid mixtures, especially the model also successfully distinguished different polycrystalline forms and enantiomers. The same substance exists in the form of two or more crystal structures, and they are called polymorphs to each other, while the substances that are enantiomers to each other have the same molecular weight and atomic composition, but the arrangement of atoms is slightly different. it's actually two kinds of substances. These two differences are important in the pharmaceutical industry, and it usually takes a lot of time and effort to distinguish them.

In addition, the researchers used the model to analyze a more complex situation-mixtures with four components.

The machine learning model for real-time analysis of dynamic processes can quickly identify mixture components, so it is more suitable for real-time analysis of some dynamic processes than traditional analysis methods. Using this model, we can find the problem in time and intervene in the process of production or experiment as soon as possible before the hidden trouble breeds the accident, so as to reduce the loss.

For example, aminosalicylic acid (PAS) is an antibiotic used in tuberculosis drug treatment. Solid aminosalicylic acid can undergo thermal decarboxylation to form m-aminophenol (MAP). As the reaction goes on, the physical appearance of the mixture of reactants and products will continue to change. By letting the machine learning model analyze the relevant pictures, the researchers can accurately know the reaction progress and the reaction yield at each stage.

The researchers also carried out supplementary training on the model so that it can analyze photos with average definition, so that we can easily use mobile phones for analysis.

The researchers say the model can be used for continuous and rapid assessment in the future, such as monitoring chemical reactions in chemical plants and laboratories. In addition, the model can also help visually impaired people to observe the outside world.

References:

Machine Learning-Based Analysis of Molar and Enantiomeric Ratios and Reaction Yields Using Images of Solid Mixtures | Industrial & Engineering Chemistry Research (acs.org)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.