Quantum Bayesian Networks

October 23, 2019

Waiting for the qc quantum chemistry revolution is like waiting for Godot

Filed under: Uncategorized — rrtucci @ 7:31 pm

Check out this very well researched and balanced article:

Waiting for the Quantum Simulation Revolution,
October 21, 2019, by Gabriel Popkin
https://physics.aps.org/articles/v12/112

Google Achieves Quantum Electricity Bill Supremacy

Filed under: Uncategorized — rrtucci @ 6:52 pm

It’s been a torturous journey for Neven in his dreams to achieve quantum supremacy. This blog is ten years old, which is old enough for it to have captured news of when Neven bought his first DWave (in collaboration with NASA, and using Google and NASA money, of course.) Then in Sept 2014, Neven ditched DWave and dangled serious money in front of Martinis if Martinis would build a gate model qc for The Google. And now, 5 years later, Neven and Martinis have achieved what they have long promised, with a Nobel prize gleam in their eyes, Quantum Supremacy. The press and social media (Reddit, Twitter, …) are abuzz with the news. Here is the official Nature paper by Google

https://www.nature.com/articles/s41586-019-1666-5

But the people at IBM are sore losers (not that the people at Google are any more virtuous), so one day before the Nature journal article was released, IBM started a publicity campaign to belittle Google’s achievement.

https://www.ibm.com/blogs/research/2019/10/on-quantum-supremacy/
https://arxiv.org/abs/1910.09534

In a recent paper that was probably leaked intentionally by Google :), they claimed that their 53 qubit Sycamore chip took 200 seconds to perform a task which would take a supercomputer 20,000/2 years to perform. Now IBM, who built what is currently the most powerful supercomputer in the world, the Summit at Oak Ridge National Lab, claims they can do it in on Summit in two days!!?? 3,650,000/2 times faster than what Google claims. IBM Killjoys.

IMHO, it’s a pretty bogus critique from IBM, because running Summit full throttle for 2.5 days would incur an electricity bill that would be many orders of magnitude bigger than running a qc for 200 secs. So, at the very least, Martinis has achieved quantum electric bill supremacy, which is nothing to scoff at.

October 19, 2019

ArXiv paper on squids implementation of single-step multi controlled Not

Filed under: Uncategorized — rrtucci @ 7:42 pm

Check out

“Single-step implementation of high fidelity n-bit Toffoli gate”, by . E. Rasmussen, N. T. Zinner, https://arxiv.org/abs/1910.07548

Milestone work, IMHO, because single-step controlled gates with n controls will be super-useful in quantum AI* and quantum error-correction.

*Why do I say that? Such gates represent a single qubit rotation which is only carried out if n binary questions are made and the answer to those n questions agrees with an answer specified a priori. That is 2^n choices! It would take on the order of 2^n separate steps, each consisting of a single 2 qubit CNOT, to check what this beast of a gate checks in a single step!

Using such multi-control gates, one can program quantum circuits first as dags (directed acyclic graphs, i.e., quantum bayesian networks) and then compile those dags into quantum circuits. Since I first proposed quantum Bayesian networks in 1997 and unveiled my program Quantum Fog (originally in C++ with Mac GUI, now in Python at github), I have spoken about my dream of programming quantum computers this way. This year, I wrote a paper and software showing how to calculate the gradient of one qubit rotations with n-controls and proposing a specific quantum circuit that uses such multi-control gates to do quantum AI and quantum Bayesian networks

https://github.com/artiste-qb-net/qubiter/blob/master/qubiter/adv_applications/threaded_grad.pdf

https://qbnets.wordpress.com/2019/05/01/popular-talk-on-multi-threading-gradients-ai-quantum-computing/

Such multi-controlled quantum gates remind me of a cool Popular Mechanics article I once read

“19 Beautiful and Ludicrous Control Panels” Oh to switch these switches. By Eric Limer Aug 26, 2015

From that article, Apollo Lunar Lander control panel:

gallery-1436391407-lunar

October 12, 2019

Squashed Entanglement can now be calculated

Filed under: Uncategorized — rrtucci @ 7:55 pm

I’m doing a small publicity blitz for my free open source software Entanglish. I’ve posted this story here and on Reddit, Medium, Rigetti Slack, and DataScienceCentral.

A question about the future that I wonder about is: which company and nation of the world will calculate squashed entanglement (a purely classical HPC calculation) for the largest physical system. Will it be the US (Google, IBM-Summit-Oakridge), China(Alibaba) , EU, etc.

If at first you don’t succeed in understanding squashed entanglement, try, try again

Filed under: Uncategorized — rrtucci @ 2:06 am

My first paper proposing an algorithm for calculating squashed quantum entanglement, was in 2001, 18 years ago

“Relaxation Method For Calculating Quantum Entanglement”
by Robert R. Tucci, https://arxiv.org/abs/quant-ph/0101123

The algo was based on the Arimoto-Blahut (AB) algorithm of classical information theory, where it is used to calculate channel capacities. The algo worked, but it was occasionally numerically unstable, and, back then, I didn’t understand too well whence came the instability.

My first public release of Entanglish was on Sep. 27, about 2 weeks ago
https://qbnets.wordpress.com/2019/09/27/welcome-to-first-release-of-entanglish/

That first release of Entanglish came with a paper explaining a new second algo for calculating squashed entanglement. That second algo for calculating squashed entanglement was also based on the AB algo, but it was considerably stabler than the first algo. Two or three days ago, I had a Eureka moment in which I realized that the second algo was giving the wrong answer, and I also realized why the first algo was numerically unstable and how to prevent such instabilities from arising. I immediately went to work and devised a new third algo for calculating squashed entanglement. This third algo, like the first two, is based on AB, but it avoids much better the numerical instabilities that plagued the first algo, and it gives the right answer unlike the second algo. Today I uploaded to github a revision of Entanglish that incorporates this new third algo. Hurray!

For the third algo, I invented a new subroutine called regulate() that reminds me of a “regulator” or “governor” in a steam engine. regulate() prevents matrices from acquiring negative eigenvalues, and keeps the sum of those eigenvalues fixed. Device regulators are used all over science (in electrical and mechanical devices, and in control theory). https://en.wikipedia.org/wiki/Regulator

centrifugal-governor

Blog at WordPress.com.

%d bloggers like this: