Neural Networks (NNs) in Confinement Fusion

I just found the following paper written in the "Proceedings of the National Academy of Sciences of the USA" (PNAS), which describes the application of machine learning surrogate models for prediction in confinement fusion:

Rushil Anirudh, Jayaraman J. Thiagarajan, Peer-Timo Bremer, and Brian K. Spears, "Improved surrogates in inertial confinement fusion with manifold and cycle consistencies"

Proceedings of the National Academy of Sciences of the USA (PNAS), 117 (18), 9741-9746, 2020.

1) Link to the paper www, HERE
2) Link to the .pdf file, HERE

It seems some specific types of neural networks may work well in prediction tasks, in these so hard to simulate plasma physics scenarios.

News at the National Ignition Facility (NIF) website, HERE.

[Almost] - Everything / Anything unified in ML?

I would like to think that the idea proposed by Prof. Eric P. Xing about the possibility to set out a standard model in Machine Learning has good chances to be accepted, or at least carefully analyzed.

Have a look at the video below:

And then - Magic!

Just uau!

I recently discovered Diego Catalán and I am amazed by his painting and drawing work.

I can not reproduce it here. Just go to the links above and enjoy!

Disentangling Luke Skywalker - Darth Vader connection

Two questions:

- Would Machine Learning be able to convince Luke Skywalker that Darth Vader is his father?

- What would have happened in the case of Luke and Princess Leia?

I just found by chance a very interesting paper by Prof. Theo Gevers et al about the possibility to infer not only the probability that two people are kins, but also the type of kinship between them. 

I was not even aware that kinship verification was another research topic in machine learning. More specifically, the aim now is about kinship identification.

Details are as follows:

1) Wei Wang, Shaodi You, Sezer Karaoglu, Theo Gevers, "Kinship Identification through Joint Learning Using Kinship Verification Ensemble" 

arXiv:2004.06382

PDF HERE 

Scientific Déjà Vu

I have just read a blog entry (HERE) by Prof. Yoshua Bengio about the current publishing dynamics in machine learning.

Insightful analysis about how scientific reviewing process might be improved.

Reading this blog has allowed me to discover The Slow Science Manifesto

Do your own homework: Fermi problem

This is not an Enrico Fermi problem, but somehow close to it (or inspired by).

Would you be really interested to know the impact a phenomenon like COVID-19 may have in a country's life?

Play with numbers a little bit:

1) Write down the entire population of a country: A

2) Assess the percentage of people would be expected to be in contact with the virus, and infected: 40%, for instance

3) Assume a 3% mortality rate in relation to the number of people that were infected

4) Do the following product:

Total death number = A * 0.4 * 0.03

Do you think this is not serious enough?

Building selfless and knowledgeable communities

A list of 24 sources of free machine learning knowledge:

https://blog.paralleldots.com/data-science/24-best-and-free-books-to-understand-machine-learning/

Enjoy learning!

Substantiation of rationality - human weaknesses

I just found a paper describing a way to beat a deep fake detector. In particular, the paper analyses how a state-of-the-art-deep face detector may determine that a face is real, when it is not.

The paper is:

Paarth Neekhara, Shehzeen Hussain, Malhar Jere, Farinaz Koushanfar, Julian McAuley,
"Adversarial Deepfakes: Evaluating Vulnerability of Deepfake Detectors to Adversarial Examples"
arXiv:2002.12749
PDF file HERE

No comments...

Hybrid Intelligence - Hybrid Moral Intelligence?

I understand that there might be a useful point in trying to enhance human decisions with some help (hint, whatever you would like to call it) from artificial intelligence (AI).

When should we take advantage of AI in our decisions?

As per Wikipedia:

"Morality (from Latin: moralitas, lit. 'manner, character, proper behavior') is the differentiation of intentions, decisions and actions between those that are distinguished as proper and those that are improper...Morality can be a body of standards or principles derived from a code of conduct from a particular philosophy, religion or culture, or it can derive from a standard that a person believes should be universal... Morality may also be specifically synonymous with "goodness" or "rightness".

Morality is a creation of the human being.

- Do we really need to have the help of an AI system for that?

- Would an AI agent help us decide whether an action is proper or improper?

I just read some information regarding research lines aiming at infer how AI can support human moral judgments.

NO need to help me. Thank you!

NO to no man's land

Nice video...A few months now, but hits the nail on the head.

Old code - New Loops

I just found by chance in the NASA web page (HERE) that they have been running a model named ROCKE-3D that:

"...calculates the details of the orbit of any planet around its star. This has been modified from the original Earth model so that it can handle any kind of planet in any kind of orbit, including planets that are "tidally locked,"..."

The model (code) can be found HERE.

My first approach to FORTRAN was during my BSc in Physics, in 1996. NASA still uses it, and for very modern purposes (so to speak).

"They have mouths, but cannot speak, eyes, but cannot see..."

2019 was the 2nd hottest year on record - as per NASA.

Psalm 115:5-7: "They have mouths, but cannot speak, eyes, but cannot see. They have ears, but cannot hear, noses, but cannot smell. They have hands, but cannot feel, feet, but cannot walk, nor can they utter a sound with their throats".