Events Add an event Speakers Talks Collections
 
Duration 17:36
16+
Video

Q2B 2020 | QC Ware Forge - Delivering Performance | Fabio Sanches, Sean J. Weinberg, Natalie Parham

Fabio Sanches
Senior Research Scientist at QC Ware Corp
+ 2 speakers
  • Video
  • Table of contents
  • Video
Q2B 2020
December 8, 2020, Online, USA
Q2B 2020
Request Q&A
Q2B 2020
From the conference
Q2B 2020
Request Q&A
Video
Q2B 2020 | QC Ware Forge - Delivering Performance | Fabio Sanches, Sean J. Weinberg, Natalie Parham
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Add to favorites
423
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About the talk

Fabio Sanches, Quantum Computing Services Lead, Sean J. Weinberg, Quantum Engineer, and Natalie Parham, Quantum Engineer of QC Ware, present to attendees on December 8, 2020 at Q2B20 - Practical Quantum Computing, an annual conference hosted by QC Ware.

Recorded sessions and conference details can be found at: https://q2b.qcware.com/

More information on QC Ware can be found at: https://qcware.com/

Get in touch with the QC Ware team at: https://qcware.com/contact

To learn about how to get started with Forge, QC Ware's data science platform running on quantum computers, visit https://forge.qcware.com/

00:15 QC Ware Forge

00:34 Two types of users

01:39 Multiple backends available

03:10 Distance Estimation

03:41 Data Loaders

06:48 Optimization on Quantum Annealers

07:21 Accelerating the performance of the D-Wave quantum Annealer

08:06 Demonstration

10:17 Optimization on Gate-Based Quantum Computers

14:02 Advanced Circuit Editing & Simulation

14:20 Import quantum circuits into Forge

About speakers

Fabio Sanches
Senior Research Scientist at QC Ware Corp
Sean J. Weinberg
Quantum Computing Researcher at QC Ware Corp.
Natalie Parham
MMath student at University of Waterloo

Fabio did his PhD in theoretical Physics at UC Berkeley where he focused on quantum gravity, black holes, and holography. He is developing applications for quantum algorithms in near term quantum computers.

View the profile
Share

I'm Fabio and I'm here with Sean and Natalie from DC where and I'm going to be telling you about how cute she wears. Forge platform delivers performance to our customers. The forge is a Quantum Computing software platform in the main focus of Borg is on delivering the best quantum, algorithms targeting specific problems. So even though Quantum algorithms are the main focus. We also bundle Hardware access to deliver a better user experience. Show me pills Forge with two types of users in mind. The first star, the expert

users. So those are the users who are already familiar with Quantum Computing and may want to write and stimulate their own Quantum circuits for those users. First provides basic circuit editing functionality as well as gpu-accelerated simulation, and the ability to import circuits directly from kids get insert. Rebuilt Ford 450 recall new users. So those are users who aren't really familiar with Quantum Computing but still have use cases that may benefit from using Quantum algorithms. Do for those users. Forge provides easy-to-use.

TurnKey, algorithm implementation. And all those algorithms are back by unique, algorithms, redeveloped Syracuse you are and everything is delivered through a posted. Jupyter Notebook. On the web. This is actually a snapshot of what force looks like. Are you see this more closely? When Sean and Natalie give the demos later. Pictures of Atkins that we currently have available that you can access through Forge for Quantum on Euler's. We have both be ways Advantage machine that the newer machine with over 5,000 Cuban as well as a 2000 Cubit machine for circuit model Hardware. We

provide access to ion cues iron trap as well as we get a superconducting device to the Amazon bracket platform. Will soon a table at the loss of IBM falcons as well. And like I mentioned we also have CPU and GPU simulators available. This is the high-level architecture of Forge. Like I mentioned earlier, the main focus is on the quantum, algorithms portion, as well as the essential performance. Booster add on that. We that we developed internally for each of those algorithms. So, we'll see. More closely the binary optimization functionality as well as the machine

learning functionality and will soon release Monte Carlo, functionality into next major release. These algorithms are often delivered either directly or through some custom solvers that we also built targeting specific problems. A lot of this functionality is built using some common Middle where that we've developed internally. That just makes essentially connecting to the different platforms as well as Hardware back in more straight for us. I want to dive in a little bit into the individual libraries, and we're going to start with a Quantum machine learning library.

So our Quantum machine learning algorithms are built using our Quantum distance estimation algorithm. So, this is estimation is really at the core of a lot of machine learning algorithms. So speeding that up using quantum computer, I said, she allows us to speed up a collection of different machine learning. Algorithms developed is a way to do distance, estimation, on a quantum computer. That is actually logarithmic in the dimensionality of the vectors. In addition to do machine learning, we need access to the data. And if you're doing Quantum algorithms, you need Quantum access to

classical data, that's going to require users, to load the classical data onto a quantum computer. In a lot of you who are familiar with, this may have already heard of the new Ram problem. So what we've developed is a set of data loaders, that provides a flexible way to load classical data on a quantum computer. So now I'm going to introduce Jon Weinberg a Quantum engineering. You see where and he's going to be demoing this machine learning functionality specifically. What we're going to see is our cue near Century Eye Care, nearest, neighbors, and r k means out

Rhythm essentially, with a side-by-side comparison between our library and scikit-learn, and all of those out rhythms are backed by our distance estimation, end loader functionality. So go ahead shotton. Hi, I'm Sean Weinberg, a Quantum engineer and a researcher. I can see where I'm going to show you, how we can use Forge to implement Quantum machine, learning applications. As a first step. I'm going to generate some data with this code. I generate 40 data points each one of which has two Dimensions. Now, I'm going to use our Quantum nearest centrala classifier. This is a supervised

learning algorithm. Take a look at the results. On top. We have the quantum, algorithms results. And on the bottom. We have a comprable classical clustering algorithm. Very similar. Next we're going to do a nearest neighbor classification. Now, wildest runs. I just want to emphasize that a lot of important things happened under the hood. There's a Quantum distance estimation, as well as the algorithm to Fabio top told you about for loading classical data into the quantum circuit are loners. So looking again, we see that the classical and Quantum clusters. Look very similar.

Finally, Let's do an unsupervised. Learning example, we can use our Quantum algorithm called humans. This is a Quantum version of the classical algorithm K means that you may be familiar with and because of his unsupervised learning. There's no labels on the data points. So the clustering is I'm done without labels and the results are very similar. The thing that I want you to take home from this demo, is that a lot of the tools that we built with Forge allowed suppose that you have training in the end classical data science, you can use that

training and just basically use for exactly what you would use a tool like scikit learn. And that way, it's very easy to get your foot into the door and such an esoteric subjects. Like, okay, thanks to use the quantum, machine learning, algorithms. We've built, and a really directly do compared with the functionality. That a lot of you already familiar from scikit learn. Move on to the binary optimization Library. So what ways to solve optimization problems using quantum computers is making use of quantum of dealers dealers, eventually provide a heuristic way to

solve specific optimization problems. And as is the case with lots of heuristics State, they can kind of very into quality of dissolution and what we've developed is essentially, it waited to certain Anil annealing parameters on device device that actually make improves the outcome of quantum on you. So, this is what I mean. We're, we're fighting. Here is a sense. He's a total count of good solutions for a specific optimization problem that we chose. Of course, Morgan Solutions is a good thing and you were comparing the same number of run side-by-side. So one on

on one of the bars Reese's kind of Z-Wave with standard parameters, and the other one is but augmented using Where's the new offsets out for the smaller problem, which is actually easier are. I'll grab them, actually does underperform, but you can see that as a problem skills and becomes harder to sell the assets algorithm developed by, to see where the liver is a much higher count of good solution. Introduce Natalie. Parham also a Quantum engineer to see where we going to. Give you a demonstration of this functionality. Specifically Will C Wood?

Standard-setting side-by-side with QC. Where's and you'll go ahead, Natalie? Hi everyone. My name is Natalie and I'm also a Quantum engineer. Excuse you are where I research and development algorithms. Okay. So in this notebook, I'm going to be showing forges Advanced functionality on. Machines so you can see right here. I have to find the Q Matrix, which specifies, the optimization problem that we would like to solve. This problem is defined on 100 cubits. So do you have a lousy? Its users to specify a variety of different

parameters? When when running a Quantum Anil? And then we talked about standard Quantum annealing, what we mean by that, is the default transmitters that do wave stats when the user doesn't specify any and hear a q. See where we've developed here is sick, that often determined better parameters compared to standard, kneeling in terms of getting a better probability of success will be comparing the two. So, in the cell that I just started running, I am running. Standard, kneeling on do is 2000 Q. And you can see that we get a probability of success of

7.3%. Now, I'm running to see where is enhanced version of quantum annealing in this cell and success probability of 13%. So that's almost two times as good as before. And I want to emphasize hear that. In order to do this last call. You see where we only had to specify a single parameter, really, making it simple for users to take advantage of advanced specifications on the TV. That's all for Quantum, annealing back to you Bobby. Oh, thanks Natalie. So as you can see.

Easy-to-use already lost its algorithm, but also it is for certain problems that can really improve the success probability on Z-Wave. So now I want to talk about another way of solving optimization problems on quantum computers, and that's using essentially circuit model devices. So one promising algorithm for the near-term. Quantum computers is the Quantum approximate optimization out for them. And for those of you that are familiar with the QA away, are you also know that you actually need to choose certain parameters are angles? And I can take quite a bit of time.

So what we've also developed a Q-See where is an algorithm that can find the optimal angles for the keyway way. It's specifically for the shirt you a service. For those of you are familiar with a fraction of the time. So now Natalie is going to demo the functionality again. So specifically over going to see is the calculation of the QA way angles using default setting as well as the QC. Where is angle calculation. Thanks Bobby. Oh, so as he said, this notebook is about finding the optimal parameters for the Cure White

Station algorithm and it's become very popular in recent years. So, here we go to find the optimization problem again, and this time, it's only specified on four cubits. So the performance of the QA heavily relies on whether or not you pick that are good. So I just started running is the typical approach for evaluating the performance of different parameters. And as you can see, it's taking kind of considering that this problem is only specified on four cubits.

So when this completes, we're going to see a nice heat map. And so here's the heat map and we can interpret this heat map as the blue spots correspond to promoters at Legion good Solutions and the red spots, corresponds to bad Solutions. And what I want to emphasize here is the complexity of this and that there's some good spots, near bad spot. So if you if you're a little bit off from good perimeter, as you might end up getting a bag solution. So I really want to emphasize hear the complexity and importance the right parameters

into Iowa. And lastly, this algorithm took 2/22 seconds, which for only four cubits as I said, and is problematic. So heroic, you see where we've developed an algorithm that calculates the exact same thing as above, but in a fraction of the time. So you can see this took less than 2 and 1/2 seconds. Is a significant improvement over 20 seconds, is to be able to do this calculation in a fraction of the time. Reduces the overall overhead of this algorithm significantly. That's all. Thanks again, Natalie. So as you can see,

using our algorithm is extremely straightforward. I also want to emphasize that to need is qay. Parameters, is really important to get those solutions that are better than the alternative algorithms using the Cure away and doing that much faster. Really improves the time to solution. So the last thing I want to talk about are the expert tools. So specifically, those are the circuit editing and simulation tools that I mentioned earlier. So here's a snapshot of what that looks like. For those of you who are familiar. It's probably really similar to other open-source libraries.

You used to do circuit editing and simulation. We provide essentially the standard functionality, including simulation getting measurements as well as Computing. The expectation value of poly operators and the gradient of the expectation value of those operators with respect to parameters in your circuit. Another functionality. We provided the ability to directly import circus from casket insert. And what I want to show you now is essentially the GPU simulation functionality that we make available through Forge. So, of course, lots

of other open source. Libraries, also provide essential egpu, simulation tools, and Provisions. The GPU for you, which makes it really easy to Century used us sped up from the sped up. Simulation using a cheap you say specifically what Sean's going to demo to you is a circuit circuit circuit running locally on his advice and then side by side with the with a circuit imported into running on our Provisions. If you go ahead, Sean, okay. So unlike the other Dubose, you seen this notebook that you see here. Actually just runs on my personal laptop. So I have some

codes are used it to make a circuit with 20 cubits and 984 Gates. That's pretty non-trivial. And I then simulated On my computer and as you see it took 30 seconds just for your reference is running on a MacBook Pro. So I want to show you how easy it is to take towed the road and kiss get like this copy and paste it into Forge imported into Forge and take advantage of forges provision Jeep. You use the GPS simulator. Then as I advertise, I just copied and pasted decode from that other notebook into Forge. And then right here, this line on this one line here is for importing the code

importing. A casket circuit into Forge 2984 Gates just like before. So let's simulator on the GPU. Okay. Took about 3 and 1/2 seconds. It's about a 10 x speed up the point here. The real thing to notice is that you don't have to go to the work of provisioning, a GPU learning how to write to decode, Etc. You can just take advantage of of the ability to take that to the road and kiss, get copy and paste into Forge imported and easily brought it on the cheap. You just like you would run it. Local simulation your computer.

You used to simulate circus using Forge, everything's provision for you, easy to use. So, that's it for our demos today. I also want to point out that anyone can sign up for a trial. Of course. Do you see where the trial account? Gives you enough time to test out all these features and we encourage you to do so, as well as if anyone has feedback. Please do reach out to us to request until We're always happy to hear what the community wants to see and what they think about these

features already. That's it for today. Thanks for listening. Also. Thanks again to Sean and Natalie for the day.

Cackle comments for the website

Buy this talk

Access to the talk “Q2B 2020 | QC Ware Forge - Delivering Performance | Fabio Sanches, Sean J. Weinberg, Natalie Parham”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “Q2B 2020”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “IT & Technology”?

You might be interested in videos from this event

November 9 - 17, 2020
Online
50
252
future of ux, behavioral science, design engineering, design systems, design thinking process, new product, partnership, product design, the global experience summit 2020, ux research

Similar talks

Alejandro Perdomo Ortiz
Associate Director of Quantum AI at Zapata Computing, Inc.
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Yianni Gamvros
Head of Business Development at QC Ware Corp.
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Sonika Johri
Senior Quantum Applications Research Scientist at IonQ
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Buy this video

Video
Access to the talk “Q2B 2020 | QC Ware Forge - Delivering Performance | Fabio Sanches, Sean J. Weinberg, Natalie Parham”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
944 conferences
37487 speakers
14316 hours of content
Fabio Sanches
Sean J. Weinberg
Natalie Parham