Data Processing Archives - Artificial Intelligence

Machine learning reduces microscope data processing time from months to just seconds

aiuniverse — Wed, 09 Jun 2021 06:39:37 +0000

Source – https://www.nanowerk.com/

Ever since the world’s first ever microscope was invented in 1590 by Hans and Zacharias Janssen –a Dutch father and son– our curiosity for what goes on at the tiniest scales has led to development of increasingly powerful devices.

Fast forward to 2021, we not only have optical microscopy methods that allow us to see tiny particles in higher resolution than ever before, we also have non-optical techniques, such as scanning force microscopes, with which researchers can construct detailed maps of a range of physical and chemical properties.

IBEC’s Nanoscale bioelectrical characterization group, led by UB Professor Gabriel Gomila, in collaboration with members of the IBEC’s Nanoscopy for nanomedicine group, have been analysing cells using a special type of microscopy called Scanning Dielectric Force Volume Microscopy, an advanced technique developed in recent years with which they can create maps of an electrical physical property called the dielectric constant.

Each of the biomolecules that make up cells –that is, lipids, proteins and nucleic acids– has a different dielectric constant, so a map of this property is basically a map of cell composition. The technique that they developed has an advantage over the current gold standard optical method, which involves applying a fluorescent dye that can disrupt the cell being studied.

Their approach doesn’t require the addition of any potentially disruptive external agent. However, the application of this technique requires a complex post-processing process to convert the measured observables into the physical magnitudes, which for eukaryotic cells involves huge amounts of computation time. In fact, it can take months to process just one image in a workstation computer, since the dielectric constant is analysed pixel by pixel using local reconstructed geometrical models.

Months to seconds

In this new study, published in the journal Small Methods (“Fast Label-Free Nanoscale Composition Mapping of Eukaryotic Cells Via Scanning Dielectric Force Volume Microscopy and Machine Learning”), the researchers opted for a new technique to speed up the microscope data processing. This time, they used machine learning algorithms instead of conventional computational methods.

The result was drastic: once trained, the machine learning algorithm was able to produce a dielectric biochemical composition map of the cells in just seconds. In the study, no external substances were added to the sample, a long-sought goal in the composition imaging in cell biology. They achieved these rapid results by using a powerful type of algorithm called neural networks, which mimics the way that neurons in the human brain operate.

The study was first-authored by Martí Checa, who carried it out as part of his PhD in Gomila’s group at IBEC. He is now a postdoctoral researcher at the Catalan Institute of Nanoscience and Nanotechnology (ICN2). “It is one of the first studies to provide such a rapid label-free biochemical composition map of dry eukaryotic cells”, Checa explains.

Indeed, in this proof-of-concept study, the researchers used dried out cells, to prevent the huge effects of water in the dielectric measurements due to its high dielectric constant. In a recently published follow-up study (Nanomaterials, “Dielectric Imaging of Fixed HeLa Cells by In-Liquid Scanning Dielectric Force Volume Microscopy “), they also analysed fixed cells in their natural in-liquid state. Here they were able to compare the values obtained in the dry and liquid models in order to render an accurate map of the biomolecules that make up eukaryotic cells. These are the multi-structured cells that animals, plants, fungi and other organisms are composed of.

“The next step in this research is to apply the method to electrically excitable living cells, such as neurons, where intense electrical activity occurs. We are excited to see what can be obtained with our technique in these systems” Prof. Gomila adds.

Biomedical applications

The researchers validated their methodology by comparing their findings to well-known facts about the composition of cells, such as the lipid-rich nature of the cell membrane or the high quantity of nucleic acids present in the nucleus. With this work, they have opened up the possibility of analysing large quantities of cells in record time.

This study is expected to provide an invaluable tool to biologists to conducting basic research, as well as, to open up potential medical applications. For example, changes in the dielectric properties of cells are currently being studied as possible biomarkers of some illnesses, such as cancer or neurodegenerative diseases.

“It is the first study to provide such a rapid nanoscale biochemical composition map from dielectric measurements of dry eukaryotic cells, which are classically seen as being extremely difficult to map due to their complex three-dimensional topography”, declares Martí Checa, first author of the paper.

The post Machine learning reduces microscope data processing time from months to just seconds appeared first on Artificial Intelligence.

QEBR Streamlines Holdings To Concentrate On Filecoin Development And Mining

aiuniverse — Sat, 25 Jul 2020 07:19:27 +0000

Source: aithority.com

QEBR announced that it has divested its ownership in two subsidiaries in order to focus all corporate resources to the buildout of its blockchain-based Filecoin mining operations. Filecoin raised $257 Million in a 2017 Initial Coin Offering, the largest ICO in history at the time, from well-regarded investors such as Sequoia Capital, Andreesen Horowitz, Y Combinator, Naval Ravikant, and Winklevoss Capital. Filecoin expects its mainnet to launch in the second half of 2020, opening the cryptocurrency to global access.

The QEBR technology team previously announced that it has proven its system as a valid Filecoin node with CPU, GPU, bandwidth, and storage compatibility that meets all IPFS guidelines. The QEBR test system has connected with the Filecoin main blockchain and already successfully test-mined filecoin.

Jun Liang, Chief Technology Officer of QEBR, stated, “The divesture of Sheen Boom and Jihye will allow our team to focus solely on the upcoming worldwide launch of Filecoin. QEBR’s subsidiary, Shenzhen DZD Digital Technology Ltd (“DZD”), has a strong background in blockchain development, data mining, encrypted data acquisition, data processing, and researching of data technology.We strongly believe that Filecoin has the ability to be a leading blockchain-based cryptocurrency and will put all efforts into making QEBR a significant player when the Filecoin mainnet launches soon.”

About Filecoin: The Filecoin project is a decentralized storage system based in the cloud. Its InterPlanetary File System, or IPFS, requires FIL coins as payment to miners in exchange for storage space.

Filecoin, developed by Protocol Labs, is a decentralized storage network. The network is expected to give owners of unused storage a means to monetize their storage capacity. It is also expected to bring down the costs of storing data reliably. Given the large amounts of unused storage in data centers and hard drives around the world, a natural market exists for this service.

The post QEBR Streamlines Holdings To Concentrate On Filecoin Development And Mining appeared first on Artificial Intelligence.

Using Transfer Learning to Overcome the Barriers Facing Machine Learning in Materials Science

aiuniverse — Mon, 17 Feb 2020 06:58:00 +0000

Source: allaboutcircuits.com

Machine learning’s ability to perform intellectually demanding tasks across various fields, materials science included, has caused it to receive considerable attention. Many believe that it could be used to unlock major time and cost savings in the development of new materials.

The growing demand for the use of machine learning to derive fast-to-evaluate surrogate models of material properties has prompted scientists at the National Institute for Materials Science in Tsukuba, Japan, to demonstrate that it could be the key driver of the “next frontier” of materials science in recently published research.

Insufficient Materials Data

To learn, machines rely on processing data using both supervised and unsupervised learning.

With no data, however, there is nothing to learn from.

Unfortunately, potential technological advances in machine learning and its potential applications in materials science are not being fully exploited due to a considerable lack of volume and diversity of materials data. This, the Japanese researchers believe, is greatly stifling progress.

However, a machine learning framework known as “transfer learning” is said to have great potential in being able to overcome the problem of a relatively small data supply. This framework relies on the concept that various property types – for example, physical, electrical, chemical, and mechanical – are physically interrelated.

To successfully predict a target property from a limited supply of training data, the researchers used models of related proxy properties that have been pretrained using sufficient data. These models are then able to capture common features that are relevant to the target task.

This repurposing of machine-acquired features on the target task has demonstrated high prediction performance even when the data sets are very small

Facilitating the Widespread Use of Transfer Learning

To facilitate the widespread use and boost the power of transfer learning, the Japanese researchers created their own pre-trained model library, XenonPy.MDL.

In its first release, the library is comprised of more than 140,000 pre-trained models for various properties of small molecules, polymers, and inorganic crystalline materials. Along with these models, the researchers provide literature that describes some of their most outstanding successes of applying transfer learning in varying scenarios. Examples such as being able to build models using the framework in conjunction with only dozens of pieces of materials data nicely hammer home the effectiveness of the transfer learning framework.

The researchers also highlight how they discovered that transfer learning transcends across different disciplines of materials science, such as underlying bridges between organic and inorganic chemistry.

Indispensable to Machine Learning-Centric Workflows

Although transfer learning is being used often across various fields of machine learning, its use in materials science is still lacking.

Furthermore, the limited availability of openly accessible big data will likely continue in the near future, the researchers say, due to a distinct lack of incentives for data sharing, a problem partly caused by the conflicting goals of stakeholders across academia, industry, and public and government organizations. This means that the relevance of transfer learning will only grow and contribute further to the success of machine learning-centric workflows in materials science.

The post Using Transfer Learning to Overcome the Barriers Facing Machine Learning in Materials Science appeared first on Artificial Intelligence.

Digital Vidya, NASSCOM FutureSkills Collaborate For Big Data, Data Science Courses

aiuniverse — Sat, 25 Jan 2020 09:22:57 +0000

Source: bweducation.businessworld.in

Digital Vidya collaborates with FutureSkills, the flagship reskilling initiative by NASSCOM, to upskill/reskill the employees of the Indian IT/ITES sector on identified skill gaps in the current talent pool such as Digital Marketing and Big Data Analysis.

Organisations are constantly looking for employees to solve problems in technology-rich environments and link their work to business value. This has resulted in a significant increase in the number of data science jobs and roles, and consequently an apparent need for a well-defined competency map for these jobs. The competency map defines the principal competence for each job including the knowledge and skills required to do the job well.

In a recently signed MoU, both the organizations have come together to partner and launch foundation courses on Big Data & Data Science Courses. The two courses will be available on Digital Vidya & NASSCOM’s FutureSkills platforms.

Anuj Batra, CEO of Digital Vidya, said, “Digital Vidya is clear in its vision in being an industry integrated ed-tech company that is poised to impart skills which matter the most in the job markets. Our strength lies in creating courses that make the participants job-ready. Our Data Science & Big Data courses are structured in a manner that lay focus on imparting the skills though hands-on case studies, industry-relevant examples and practical training. The trainers and the curriculum creators are not just academics, but practicing industry leaders. After having our courses validated and accepted by IT-ITeS Sector Skills Council NASSCOM through the benchmark of National Occupational Standards, we certainly feel inspired and jubilated.”

The courses are designed to establish an understanding of Big Data Analytics, visualization, Data Processing & Management along with knowledge of various Big Data platforms and their fundamentals. The courses are highly recommended by the Industry and validated by IT-ITeS Sector Skills Council NASSCOM. All FutureSkills subscribers can take up the course at no additional cost on the NASSCOM platform.

Expressing his delight on the partnership, Amit Aggarwal, CEO IT & ITeS Sector Skills Council said, “There is a great need for upskilling in India and NASSCOM already recognizes that. This initiative is a conscious effort from our end to give a concrete resource to students & professionals to hop onto definitive pathways in the initial course of their career journey. Anyone following the course of action structured in these courses will surely reap high returns in their career.”

Digital Vidya’s Big Data course is a foundation level program that will act as a catalyst for advanced learning of this technology. On the other hand, Data Science Course is one wherein Python is used as the primary mechanical coding language, and hence will be an advanced level course propelling the student to expertise in the domain.

The post Digital Vidya, NASSCOM FutureSkills Collaborate For Big Data, Data Science Courses appeared first on Artificial Intelligence.