Toggle light / dark theme

Concept whitening: A strategy to improve the interpretability of image recognition models

Over the past decade or so, deep neural networks have achieved very promising results on a variety of tasks, including image recognition tasks. Despite their advantages, these networks are very complex and sophisticated, which makes interpreting what they learned and determining the processes behind their predictions difficult or sometimes impossible. This lack of interpretability makes deep neural networks somewhat untrustworthy and unreliable.

Researchers from the Prediction Analysis Lab at Duke University, led by Professor Cynthia Rudin, have recently devised a technique that could improve the interpretability of deep neural networks. This approach, called whitening (CW), was first introduced in a paper published in Nature Machine Intelligence.

“Rather than conducting a post hoc analysis to see inside the hidden layers of NNs, we directly alter the NN to disentangle the latent space so that the axes are aligned with known concepts,” Zhi Chen, one of the researchers who carried out the study, told Tech Xplore. “Such disentanglement can provide us with a much clearer understanding of how the network gradually learns concepts over layers. It also focuses all the information about one concept (e.g., “lamp,” “bed,” or “person”) to go through only one neuron; this is what is meant by disentanglement.”

‘Mars on Earth’ experience opens in the Wadi Rum desert

Circa 2018


“Vast, echoing and God-like.” That is how T.E. Lawrence, the British archaeologist and army officer who inspired the 1962 film “Lawrence of Arabia,” described Wadi Rum.

As you approach the wind-swept mountains that fiercely jut out of the burnt orange sand in Jordan’s largest desert, it’s easy to see what he meant. The landscape here is like something from another world.

So it’s perhaps no surprise that a hotel in Wadi Rum has just opened a dramatic Martian Experience in the heart of this wilderness, which lets visitors feel as though they have landed on the Red Planet.

The Martian city of the future is landing in Dubai, and it’s going to be an out-of-this-world experience

Circa 2020


Unreal and beyond most of our trippiest dreams, the city of Dubai is a living, breathing sci-fi movie—firefighters in jetpacks, anyone? Now try adding an entire Martian city concept to that.

The United Arab Emirates is on the same wavelength as Elon Musk when it comes to colonizing Mars. They want an entire human population on the Red Planet within the next century. Architects from Bjarke Ingels Group were asked to design Mars Science City, a prototype for what is going to turn into a hyper-futuristic lab for the Mohammed Bin Rashid Space Center (MBRSC), which will keep developing space tech that will allow humans to stay alive on a frozen planet almost 80 million miles (40 on a good day) from Earth.

Here’s How Humans Have Evolved In the Last 100 Years

You can buy Universe Sandbox 2 here: http://amzn.to/2yJqwU6
Or get a shirt: https://teespring.com/stores/whatdamath.

Hello and welcome! My name is Anton and in this video, we will talk about some of the recent discoveries in regards to our own evolution in the last 250 years.
Paper: https://onlinelibrary.wiley.com/doi/10.1111/joa.

Support this channel on Patreon to help me make this a full time job:
https://www.patreon.com/whatdamath.

Space Engine is available for free here: http://spaceengine.org.
Enjoy and please subscribe.

Twitter: https://twitter.com/WhatDaMath.
Facebook: https://www.facebook.com/whatdamath.
Twitch: http://www.twitch.tv/whatdamath.

Bitcoins to spare? Donate them here to help this channel grow!

Neuroscientists identify brain circuit that encodes timing of events

When we experience a new event, our brain records a memory of not only what happened, but also the context, including the time and location of the event. A new study from MIT neuroscientists sheds light on how the timing of a memory is encoded in the hippocampus, and suggests that time and space are encoded separately.

In a study of mice, the researchers identified a hippocampal circuit that the animals used to store information about the timing of when they should turn left or right in a maze. When this circuit was blocked, the mice were unable to remember which way they were supposed to turn next. However, disrupting the circuit did not appear to impair their of where they were in space.

The findings add to a growing body of evidence suggesting that when we form new memories, different populations of neurons in the brain encode time and place information, the researchers say.

‘Galaxy-sized’ observatory sees potential hints of gravitational waves

Scientists have used a “galaxy-sized” space observatory to find possible hints of a unique signal from gravitational waves, or the powerful ripples that course through the universe and warp the fabric of space and time itself.

The new findings, which appeared recently in The Astrophysical Journal Letters, hail from a U.S. and Canadian project called the North American Nanohertz Observatory for Gravitational Waves (NANOGrav).

For over 13 years, NANOGrav researchers have pored over the light streaming from dozens of pulsars spread throughout the Milky Way Galaxy to try to detect a “gravitational wave background.” That’s what scientists call the steady flux of gravitational radiation that, according to theory, washes over Earth on a constant basis. The team hasn’t yet pinpointed that target, but it’s getting closer than ever before, said Joseph Simon, an astrophysicist at the University of Colorado Boulder and lead author of the new paper.

/* */