The curse of dimensionality; new particle discovered; battling crop disease with smartphones.
Extracting the right information in large data sets
When solving complex scientific problems, researchers sometimes encounter what is called the curse of dimensionality, that is, they have so much data that they cannot efficiently analyze it.
Large data sets can also be expensive and time consuming to acquire, so it is critical to gather only what is necessary. To this end, University of Illinois Assistant Professor Yihong Wu, and Professor Tsachy Weissman of Stanford University were awarded $500,000 from the National Science Foundation to identify optimal algorithms that efficiently extract the relevant information from large data sets.
The researchers explained that with advanced technology and data acquisition methods, data sets can be larger than they ever have been, and while it might seem that a surplus of data would be advantageous to researchers, analyzing large data sets, called high-dimensional data, is challenging.
Wu’s algorithms are meant to cut the number of data sets, especially when measuring entropy — which is the amount of relevant information contained in the data — measured in bits.
The researchers assert that a smarter algorithm will not only help scientists extract the most relevant data, but if they have a specific goal, the algorithm can initially collect less data to fulfill that goal, saving both time and resources.
The algorithm developed by the Illinois team comes with a theoretical guarantee of optimality, meaning that it cannot be significantly improved by other schemes. One of the inspirations for solving this problem originates from what is called the “species problem” of “estimating the unseen.”
Wu hopes to use these algorithms in machine learning, which is the science of getting computers to act without being explicitly programmed.
“Many machine learning algorithms rely on the accurate estimation of information. Our improved estimators can lead to better performance and serve as building blocks for machine learning tasks on a high-dimensional model.”
New particle discovered
ETH Zurich researchers studying peculiar properties of a long known metallic material have chanced upon a new particle that is related to the so-called Weyl fermions that the mathematician Hermann Weyl predicted almost ninety years ago. The researchers remarked that Weyl had overlooked the particle, which could have interesting applications in electronics.
Physicists at the ETH, together with researchers at Princeton University and at the Chinese Academy of Sciences, have now found a thus far unknown particle that their illustrious forebear had overlooked in Weyl’s calculations and that had remained undetected for almost ninety years thereafter. They named it “type-2 Weyl fermion.”
The researchers explained that they came upon the new kind of particle when trying to understand strange physical properties of the metal tungsten ditelluride (WTe2). They explained that they were hoping to find especially so-called ‘topological’ properties that make certain quantum states more resistant to perturbations in this metal material.
Scientists are already speculating that this could open up entirely new possibilities for electronic devices.
Battling crop disease with smartphones
To address the problem of lack of access to advanced diagnostics and treatment advice for crop diseases — which in turn can lead to famine — scientists from EPFL and Penn State University are releasing 50,000 open access images of infected and healthy crops. They said the images will allow machine-learning experts to develop algorithms that automatically diagnose the disease of a crop. The tool will then be put into the hands of farmers – in the form of a smartphone app.