Connect with us

Biotechnology

Looking into the black box

Avatar

Published

on

Deep learning systems are revolutionizing technology around us, from voice recognition that pairs you with your phone to autonomous vehicles that are increasingly able to see and recognize obstacles ahead. But much of this success involves trial and error when it comes to the deep learning networks themselves. A group of MIT researchers recently reviewed their contributions to a better theoretical understanding of deep learning networks, providing direction for the field moving forward.

“Deep learning was in some ways an accidental discovery,” explains Tommy Poggio, investigator at the McGovern Institute for Brain Research, director of the Center for Brains, Minds, and Machines (CBMM), and the Eugene McDermott Professor in Brain and Cognitive Sciences. “We still do not understand why it works. A theoretical framework is taking form, and I believe that we are now close to a satisfactory theory. It is time to stand back and review recent insights.”

Climbing data mountains

Our current era is marked by a superabundance of data — data from inexpensive sensors of all types, text, the internet, and large amounts of genomic data being generated in the life sciences. Computers nowadays ingest these multidimensional datasets, creating a set of problems dubbed the “curse of dimensionality” by the late mathematician Richard Bellman.

One of these problems is that representing a smooth, high-dimensional function requires an astronomically large number of parameters. We know that deep neural networks are particularly good at learning how to represent, or approximate, such complex data, but why? Understanding why could potentially help advance deep learning applications.

“Deep learning is like electricity after Volta discovered the battery, but before Maxwell,” explains Poggio, who is the founding scientific advisor of The Core, MIT Quest for Intelligence, and an investigator in the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. “Useful applications were certainly possible after Volta, but it was Maxwell’s theory of electromagnetism, this deeper understanding that then opened the way to the radio, the TV, the radar, the transistor, the computers, and the internet.”

The theoretical treatment by Poggio, Andrzej Banburski, and Qianli Liao points to why deep learning might overcome data problems such as “the curse of dimensionality.” Their approach starts with the observation that many natural structures are hierarchical. To model the growth and development of a tree doesn’t require that we specify the location of every twig. Instead, a model can use local rules to drive branching hierarchically. The primate visual system appears to do something similar when processing complex data. When we look at natural images — including trees, cats, and faces — the brain successively integrates local image patches, then small collections of patches, and then collections of collections of patches. 

“The physical world is compositional — in other words, composed of many local physical interactions,” explains Qianli Liao, an author of the study, and a graduate student in the Department of Electrical Engineering and Computer Science and a member of the CBMM. “This goes beyond images. Language and our thoughts are compositional, and even our nervous system is compositional in terms of how neurons connect with each other. Our review explains theoretically why deep networks are so good at representing this complexity.”

The intuition is that a hierarchical neural network should be better at approximating a compositional function than a single “layer” of neurons, even if the total number of neurons is the same. The technical part of their work identifies what “better at approximating” means and proves that the intuition is correct.

Generalization puzzle

There is a second puzzle about what is sometimes called the unreasonable effectiveness of deep networks. Deep network models often have far more parameters than data to fit them, despite the mountains of data we produce these days. This situation ought to lead to what is called “overfitting,” where your current data fit the model well, but any new data fit the model terribly. This is dubbed poor generalization in conventional models. The conventional solution is to constrain some aspect of the fitting procedure. However, deep networks do not seem to require this constraint. Poggio and his colleagues prove that, in many cases, the process of training a deep network implicitly “regularizes” the solution, providing constraints.

The work has a number of implications going forward. Though deep learning is actively being applied in the world, this has so far occurred without a comprehensive underlying theory. A theory of deep learning that explains why and how deep networks work, and what their limitations are, will likely allow development of even much more powerful learning approaches.

“In the long term, the ability to develop and build better intelligent machines will be essential to any technology-based economy,” explains Poggio. “After all, even in its current — still highly imperfect — state, deep learning is impacting, or about to impact, just about every aspect of our society and life.”


Topics: McGovern Institute, Center for Brains Minds and Machines, Brain and cognitive sciences, Quest for Intelligence, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical engineering and computer science (EECS), Artificial intelligence, MIT Schwarzman College of Computing, School of Science, School of Engineering, Research

Source: http://news.mit.edu/2020/looking-black-box-deep-learning-neural-networks-0727

Bioengineer

Electronic skin has a strong future stretching ahead

Avatar

Published

on

A material that mimics human skin in ?strength, stretchability and sensitivity could be used to collect biological data in real time. Electronic skin, or e-skin, may play an important role in next-generation prosthetics, personalized medicine, soft robotics and artificial intelligence.

“The ideal e-skin will mimic the many natural functions of human skin, such as sensing temperature and touch, accurately and in real time,” says KAUST postdoc Yichen Cai. However, making suitably flexible electronics that can perform such delicate tasks while also enduring the bumps and scrapes of everyday life is challenging, and each material involved must be carefully engineered.

Most e-skins are made by layering an active nanomaterial (the sensor) on a stretchy surface that attaches to human skin. However, the connection between these layers is often too weak, which reduces the durability and sensitivity of the material; alternatively, if it is too strong, flexibility becomes limited, making it more likely to crack and break the circuit.

“The landscape of skin electronics keeps shifting at a spectacular pace,” says Cai. “The emergence of 2D sensors has accelerated efforts to integrate these atomically thin, mechanically strong materials into functional, durable artificial skins.”

A team led by Cai and colleague Jie Shen has now created a durable e-skin using a hydrogel reinforced with silica nanoparticles as a strong and stretchy substrate and a 2D titanium carbide MXene as the sensing layer, bound together with highly conductive nanowires.

“Hydrogels are more than 70 percent water, making them very compatible with human skin tissues,” explains Shen. By prestretching the hydrogel in all directions, applying a layer of nanowires, and then carefully controlling its release, the researchers created conductive pathways to the sensor layer that remained intact even when the material was stretched to 28 times its original size.

Their prototype e-skin could sense objects from 20 centimeters away, respond to stimuli in less than one tenth of a second, and when used as a pressure sensor, could distinguish handwriting written upon it. It continued to work well after 5,000 deformations, recovering in about a quarter of a second each time. “It is a striking achievement for an e-skin to maintain toughness after repeated use,” says Shen, “which mimics the elasticity and rapid recovery of human skin.”

Such e-skins could monitor a range of biological information, such as changes in blood pressure, which can be detected from vibrations in the arteries to movements of large limbs and joints. This data can then be shared and stored on the cloud via Wi-Fi.

“One remaining obstacle to the widespread use of e-skins lies in scaling up of high-resolution sensors,” adds group leader Vincent Tung; “however, laser-assisted additive manufacturing offers new promise.”

“We envisage a future for this technology beyond biology,” adds Cai. “Stretchable sensor tape could one day monitor the structural health of inanimate objects, such as furniture and aircraft.”

###

Source: https://bioengineer.org/electronic-skin-has-a-strong-future-stretching-ahead/

Continue Reading

Bioengineer

German researchers compile world’s largest inventory of known plant species

Avatar

Published

on

Leipzig could mean for the future of plant taxonomy what Greenwich meant for world time until 1972: it could become the reference city for correct scientific plant names. In an outstanding feat of research, the curator of the Botanical Garden of Leipzig University, Dr Martin Freiberg, and colleagues from iDiv and UL have compiled what is now the largest and most complete list of scientific names of all known plant species in the world. The Leipzig Catalogue of Vascular Plants (LCVP) enormously updates and expands existing knowledge on the naming of plant species, and could replace The Plant List (TPL) – a catalogue created by the Royal Botanic Gardens, Kew in London which until now has been the most important reference source for plant researchers.

“In my daily work at the Botanical Garden, I regularly come across species names that are not clear, where existing reference lists have gaps,” said Freiberg. “This always means additional research, which keeps you from doing your actual work and above all limits the reliability of research findings. I wanted to eliminate this obstacle as well as possible.”

World’s most comprehensive and reliable catalogue of plant names

With 1,315,562 scientific names, the LCVP is the largest of its kind in the world describing vascular plants. Freiberg compiled information from accessible relevant databases, harmonized it and standardised the names listed according to the best possible criteria. On the basis of 4500 other studies, he investigated further discrepancies such as different spellings and synonyms. He also added thousands of new species to the existing lists – species identified in recent years, mainly thanks to rapid advances in molecular genetic analysis techniques.

The LCVP now comprises 351,180 vascular plant species and 6160 natural hybrids across 13,460 genera, 564 families and 84 orders. It also lists all synonyms and provides further taxonomic details. This means that it contains over 70,000 more species and subspecies than the most important reference work to date, TPL. The latter has not been updated since 2013, making it an increasingly outdated tool for use in research, according to Freiberg.

“The catalogue will help considerably in ensuring that researchers all over the world refer to the same species when they use a name,” says Freiberg. Originally, he had intended his data set for internal use in Leipzig. “But then many colleagues from other botanical gardens in Germany urged me to make the work available to everyone.”

LCVP vastly expands global knowledge of plant diversity

“Almost every field in plant research depends on reliably naming species,” says Dr Marten Winter of iDiv, adding: “Modern science often means combining data sets from different sources. We need to know exactly which species people refer to, so as not to compare apples and oranges or to erroneously lump different species.” Using the LCVP as a reference will now offer researchers a much higher degree of certainty and reduce confusion. And this will also increase the reliability of research results, adds Winter.

“Working alone, Martin Freiberg has achieved something truly incredible here,” says the director of the Botanical Garden and co-author Prof Christian Wirth (UL, iDiv). “This work has been a mammoth task, and with the LCVP he has rendered an invaluable service to plant research worldwide. I am also pleased that our colleagues from iDiv, with their expertise in biodiversity informatics, were able to make a significant contribution to this work.”

###

This research was in part supported by the DFG – Deutsche Forschungsgemeinschaft (FZT-118).

Source: https://bioengineer.org/german-researchers-compile-worlds-largest-inventory-of-known-plant-species/

Continue Reading

Biotechnology

Benefits of Purchasing Used Lab Equipment

Avatar

Published

on

In the laboratory industry, there can be a lot of pressure to always have the newest and latest technology. Due to concerns about having outdated or inefficient equipment, many professionals don’t even consider purchasing used equipment. However, there are numerous advantages to doing so. Here are some of the key benefits of purchasing used lab equipment to keep in mind next time you need to purchase equipment.


Financial Benefits

One of the most significant and obvious benefits of purchasing used lab equipment is that doing so can save you a lot of money. Used laboratory equipment is significantly less expensive than brand new models. Often laboratories can save up to 50 percent or more by purchasing quality second-hand equipment.

Because laboratory equipment is often one of the top expenses that laboratories incur, purchasing used can substantially reduce laboratory costs and free up some much-needed room in tight budgets.

 

Environmental Benefits

Large laboratory equipment can take up a lot of room in landfills and often contain toxic components such as lead or mercury. Such toxins can seep into the earth and contaminate groundwater over time. By purchasing used laboratory equipment rather than new, you can reduce the amount of equipment that ends up in landfills. Plus, you will also reduce the number of raw materials used to manufacture new equipment which will decrease your lab’s negative impact on the environment.

Increased Insight

Purchasing a new type of equipment right after it is released comes with some risks. When a piece of equipment is fresh on the market, there aren’t many reviews from past customers attesting to how well or poorly it operates.\

Used equipment, however, is far less risky. Because the equipment has been on the market for a while, there are likely plenty of reviews that you can reference and any potential issues have likely been well-documented. Just make sure to purchase from a reliable and trustworthy reseller that took the proper measures to ensure the equipment is in optimal operating condition.

Reduced Wait Time

If you need a piece of equipment in a short period of time, buying used is often the best option. Many manufacturers have long wait times that can require you to wait for several weeks or even months before the equipment will arrive. If you have deadlines you need to meet, that simply won’t do. In cases when you need equipment in a timelier manner, used equipment is already built and ready to go so you often only have to wait for shipping.


 

Source: Christina Duron is a freelance writer for multiple online publications where she can showcase her affinity for all things digital. She has focused her career around digital marketing and writes to explore topics that spark her interest.

 

Continue Reading

Bioengineer

An ionic forcefield for nanoparticles

Avatar

Published

on

Tunable coating allows hitch-hiking nanoparticles to slip past the immune system to their target

Nanoparticles are promising drug delivery tools, offering the ability to administer drugs directly to a specific part of the body and avoid the awful side effects so often seen with chemotherapeutics.

But there’s a problem. Nanoparticles struggle to get past the immune system’s first line of defense: proteins in the blood serum that tag potential invaders. Because of this, only about 1 percent of nanoparticles reach their intended target.

“No one escapes the wrath of the serum proteins,” said Eden Tanner, a former postdoctoral fellow in bioengineering at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS).

Now, Tanner and a team of researchers led by Samir Mitragotri, the Hiller Professor of Bioengineering and Hansjorg Wyss Professor of Biologically Inspired Engineering at SEAS, have developed an ionic forcefield that prevents proteins from binding to and tagging nanoparticles.

In mouse experiments, nanoparticles coated with the ionic liquid survived significantly longer in the body than uncoated particles and, surprisingly, 50 percent of the nanoparticles made it to the lungs. It’s the first time that ionic liquids have been used to protect nanoparticles in the blood stream.

“The fact that this coating allows the nanoparticles to slip past serum proteins and hitch a ride on red blood cells is really quite amazing because once you are able to fight the immune system effectively, lots of opportunities open up,” said Mitragotri, who is also a Core Faculty Member of Harvard’s Wyss Institute for Biologically Inspired Engineering

The research is published in Science Advances.

Ionic liquids, essentially liquid salts, are highly tunable materials that can hold a charge.

“We knew that serum proteins clear out nanoparticles in the bloodstream by attaching to the surface of the particle and we knew that certain ionic liquids can either stabilize or destabilize proteins,” said Tanner, who is now an Assistant Professor of Chemistry & Biochemistry at the University of Mississippi. “The question was, could we leverage the properties of ionic liquids to allow nanoparticles to slip past proteins unseen.”

“The great thing about ionic liquids is that every small change you make to their chemistry results in a big change in their properties,” said Christine Hamadani, a former graduate student at SEAS and first author of the paper. “By changing one carbon bond, you can change whether or not it attracts or repels proteins.”

Hamadani is currently a graduate student at Tanner’s lab at the University of Mississippi.

The researchers coated their nanoparticles with the ionic liquid choline hexenoate, which has an aversion to serum proteins. Once in the body, these ionic-liquid coated nanoparticles appeared to spontaneously attach to the surface of red-blood cells and circulate until they reached the dense capillary system of the lungs, where the particles sheared off into the lung tissue.

“This hitchhiking phenomenon was a really unexpected discovery,” said Mitragotri. “Previous methods of hitchhiking required special treatment for the nanoparticles to attach to red blood cells and even then, they only stayed at a target location for about six hours. Here, we showed 50 percent of the injected dose still in the lungs after 24 hours.”

The research team still needs to understand the exact mechanism that explains why these particles travel so well to lung tissue, but the research demonstrates just how precise the system can be.

“This is such a modular technology,” said Tanner, who plans to continue the research in her lab at University of Mississippi. “Any nanoparticle with a surface change can be coated with ionic liquids and there are millions of ionic liquids that can be tuned to have different properties. You could tune the nanoparticle and the liquid to target specific locations in the body.”

“We as a field need as many tools as we can to fight the immune system and get drugs where they need to go,” said Mitragotri. “Ionic liquids are the latest tool on that front.”

###

The research was co-authored by Morgan J. Goetz.

Source: https://bioengineer.org/an-ionic-forcefield-for-nanoparticles/

Continue Reading
Cleantech2 hours ago

Four Corners EV Charging: Utah & Colorado Are Leaving NM & Arizona Behind

Cleantech2 hours ago

Aptera Announces First “Never Charge” Electric Vehicle

Cleantech3 hours ago

Gayam Motor Works & Sokowatch Launch East Africa’s First Commercial Electric Tuk-Tuks

Cleantech3 hours ago

The German Constitution May Protect A Right To Human Driving

Cleantech4 hours ago

2021 Toyota RAV4 Prime Fails Moose Avoidance Test

SaaS5 hours ago

Top 10 SaaStr Videos of the Week: MongoDB, Splash, Slack + Yammer, Gainsight and More!

Cleantech5 hours ago

Supercell Technology From Cadenza Is Centerpiece Of New York Energy Storage Project — CleanTechnica Exclusive (Video)

SaaS5 hours ago

How to Create PPC Campaigns for Real Estate Marketing

Cleantech5 hours ago

Cleantech ETFs Vastly Outperform Dow Jones, Oil & Gas In 2020

Cleantech6 hours ago

ICE Racing Can Still Teach Us Things

Cleantech6 hours ago

California’s Low Carbon Fuel Standard Accelerating Transportation Electrification

SaaS7 hours ago

7 Warning Signs You Have Product Flop on Your Hands (and How to Fix It!)

Ecommerce7 hours ago

Amazon Marketing Consulting

Globe NewsWire7 hours ago

McPhy Energy : Déclaration du nombre total des droits de vote et du nombre d’actions au 30 novembre 2020

Cleantech8 hours ago

Breaking News! Oakland & Seattle Ban Natural Gas as Cities Continue to Lead on Electrification

SaaS8 hours ago

Did You Ship At Least 3 Game-Changing Features This Year?

SaaS9 hours ago

When Should You Use Microsites

Globe NewsWire10 hours ago

Dassault Aviation : Roll-out du Falcon 6X

Globe NewsWire10 hours ago

Dassault Aviation: Falcon 6X rollout

Energy12 hours ago

Crescent Point Announces 2021 Budget

Energy12 hours ago

PHNIX dévoile une nouvelle pompe à chaleur pour le chauffage, le refroidissement et l’eau chaude des habitations, destinée au marché européen

Energy12 hours ago

PHNIX stellt neue Wärmepumpe für Hausheizung, Kühlung und Warmwasser für den europäischen Zielmarkt vor

Esports12 hours ago

NiP announce “Path of a Ninja” talent program

Energy12 hours ago

Wilbur-Ellis Agribusiness Acquires Probe Schedule

Energy12 hours ago

High Purity Alumina Market Estimated to Expand at a CAGR of 12.5% over the Forecast Period of 2020 to 2030 – Persistence Market Research

Energy15 hours ago

Gunvor inaugura una nueva línea de crédito de 540 millones de dólares dedicada al biodiésel

Energy15 hours ago

Gunvor Launches New US $540 Million Biodiesel Borrowing Base

Techcrunch16 hours ago

China’s internet regulator takes aim at forced data collection

Blockchain16 hours ago

Amazing Blocks joins startup and innovation hub TechQuartier

Blockchain17 hours ago

Amazing Blocks attended the European Blockchain Convention

Trending