Serving the Quantitative Finance Community

 
User avatar
rrtucci
Topic Author
Posts: 24
Joined: February 7th, 2010, 7:26 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 26th, 2018, 7:11 pm

"do you provide complimentary Quantum computers and Q compilers?"

We don't build any quantum computers ( hardware). But Bayesforge is jupyter notebook centric. It uses Jupyter notebooks as an interface to the cloud. From a Jupyter notebook, you can connect seamlessly with the Dwave,  IBM, Rigetti and Google (Google not online yet but they have promised) quantum computers


Bayesforge (and github) has some open source software made by us, but that is an infinitesimal part of the open source software bundle that BayesForge provides

This is our quantum compiler
https://github.com/artiste-qb-net/qubiter

Our tool for doing classical and quantum bayesian networks
https://github.com/artiste-qb-net/quantum-fog

Our tool (very young) for doing quantum neural nets
https://github.com/artiste-qb-net/Quantum_Edward

plus other stuff not public yet
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 26th, 2018, 7:53 pm

Is there a reason for using Python? is it just a front-end to TensorFlow et al or does your s/w do heavy lifting as well? Are you writing fundamental algorithms or using others' libraries? If the latter you don't have a 'product' as such and you become captive user.
Why no C++?
 
User avatar
rrtucci
Topic Author
Posts: 24
Joined: February 7th, 2010, 7:26 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 26th, 2018, 8:44 pm

Is there a reason for using Python? is it just a front-end to TensorFlow et al or does your s/w do heavy lifting as well?
Why no C++?
Bayesforge supports 3 languages:  Python, R and Octave(a Matlab clone). You can open Jupyter notebooks in all 3. You can even open a BASH shell jupyter notebook.


In the past, I wrote quantum computing programs in C++, Java and Octave/Matlab. But now i am using Python, because
1.  it has an ecosystem of software libraries for science, math and engineering, most of them free and open source, that no other language comes even close to matching (jupyter, numpy, scipy, pandas, etc.)
2.  it is in fashion in qc. All the major quantum computing platforms IBM, Google, Rigetti are using it (All except MS. Microsoft has their own proprietary language for quantum computing called Q#, which I'm sure will fail).
3.  It's in fashion in AI. Tensorflow and Pytorch use it.
4. Whenever a section of a  Python program is too slow for your purposes, you can write a program in a faster language and write a python wrapper for it. For example, the Python library Numpy is mostly a wrapper for C code, and also a wrapper for LAPACK, which is written in Fortran

All computer languages go out of style eventually.  Python is no exception. But right now, Python is king, at least in qc.
Anyway, it seems to me that all computer languages are slowly converging to a common  point in some kind of metric,  because they copy features from each other For example, Java had no lists for a long time, and now it does just like Python

 
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 26th, 2018, 10:00 pm

because they copy features from each other For example, Java had no lists for a long time, and now it does just like Python

Really?? we used to give Java 1.2 courses in the 90s and they had ArrayList. 

The big problem going forward -and an opportunity- is the hordes of Python maintenance programmers!
 
User avatar
rmax
Posts: 374
Joined: December 8th, 2005, 9:31 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 27th, 2018, 11:23 am

I saw something on Linkedin that I liked - this is rare so I thought I would share.

If you see something written in Python is probably Machine Learning.
If you see something written in PowerPoint it is probably AI.
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 3:27 pm

Q
How much of the software approach is innovative in the sense that you attempt to design the core algorithms and load-balancing software yourselves.  Do you avoid trying to ‘piggyback’ on external  libraries to do the hard number-crunching and just writing simple front-ends (for example, in Python) to them. Instead, you wish to write all the code on the ‘system critical path’ yourselves.
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 3:42 pm

I saw something on Linkedin that I liked - this is rare so I thought I would share.

If you see something written in Python is probably Machine Learning.
If you see something written in PowerPoint it is probably AI.
Personally, I dislike if/if not/else logic inside Python for loops. Maybe it doesn't matter these days and we should switch our brains off and stop worrying, what?

I must say that the Pythin numpy-dumpy, scipy libraries are cool way to learn numerics.
 
User avatar
rmax
Posts: 374
Joined: December 8th, 2005, 9:31 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 4:41 pm

Q
How much of the software approach is innovative in the sense that you attempt to design the core algorithms and load-balancing software yourselves.  Do you avoid trying to ‘piggyback’ on external  libraries to do the hard number-crunching and just writing simple front-ends (for example, in Python) to them. Instead, you wish to write all the code on the ‘system critical path’ yourselves.
Various in the shops where I work/ed. I have generally seen Datasynpase for the distribution onto the comp framework, and then used various systems to distribute, mainly as the way that the calculation distribution is very based on what you are trying to measure.
 
User avatar
rmax
Posts: 374
Joined: December 8th, 2005, 9:31 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 4:44 pm

I saw something on Linkedin that I liked - this is rare so I thought I would share.

If you see something written in Python is probably Machine Learning.
If you see something written in PowerPoint it is probably AI.
Personally, I dislike if/if not/else logic inside Python for loops. Maybe it doesn't matter these days and we should switch our brains off and stop worrying, what?

I must say that the Pythin numpy-dumpy, scipy libraries are cool way to learn numerics.
I personally used to hate Python, until I had some problems to solve with data, and it worked very well. Like with all tools, I don't really care whether it is a Philips, flat-head or allan. But if I had a choice I would generally prefer allan keys  - especially in metal.
 
User avatar
rrtucci
Topic Author
Posts: 24
Joined: February 7th, 2010, 7:26 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 6:22 pm

I mentioned previously some things that I like about  Python.   But there are somethings about it that I dislike too. My Python coding style would  be considered sacrilegous by many Python zealots. I intentionally imitate a Java style in my code: I put only one class per file, and I try to use no global functions. If I crave to make a function global,  I insert it into a class as a static method, A few functions that seem too general to insert as static functions in a class, I put into a last resort file called ''utilities". I also like to write a "main()" method at the end of each class  file to test the file's class. Python zealots use files they call modules that might contain several classes and global functions. Sorry, but I don't like their approach.
Last edited by rrtucci on December 23rd, 2018, 11:16 pm, edited 4 times in total.
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 7:00 pm

You are an OO programmer. The move these days is to stateless functional programming. Much easier for parallel programming.

IMO Python will probably teach you to become a 'hobby programmer' (nothing wrong with that of course, but it is a long way from the airs and graces of those Fortran types).

What is the lifetime of undocumented Python code. i.e. when you decide to put it out to pasture and rewrite it in C++?
 
User avatar
rrtucci
Topic Author
Posts: 24
Joined: February 7th, 2010, 7:26 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 7:30 pm

You are an OO programmer. The move these days is to stateless functional programming. Much easier for parallel programming.

IMO Python will probably teach you to become a 'hobby programmer' (nothing wrong with that of course, but it is a long way from the airs and graces of those Fortran types).

What is the lifetime of undocumented Python code. i.e. when you decide to put it out to pasture and rewrite it in C++?
Yes. I am an incurable OO programmer. If in the future, I have to translate my code to C++ or java or some newer OO language, my coding style will make translation easier. I write open source that will last forever 💍💎
Last edited by rrtucci on November 30th, 2018, 7:39 pm, edited 1 time in total.
 
User avatar
FaridMoussaoui
Posts: 327
Joined: June 20th, 2008, 10:05 am
Location: Genève, Genf, Ginevra, Geneva

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 7:35 pm

Unless you are working on a collaborative project where you have guidelines you "must" fellow, do what works for you. In computing, we have the chance "to learn by doing".

I got a lot of discussion with people telling me why I use R for my statistical analysis instead of python. I find it a useless discussion as R fits my requirements and I combine it with C++ for performance (if needed).
Last edited by FaridMoussaoui on November 30th, 2018, 7:57 pm, edited 1 time in total.
 
User avatar
rrtucci
Topic Author
Posts: 24
Joined: February 7th, 2010, 7:26 am

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 7:48 pm

Lot's of great stuff in R. Bayesforge favors AI done with Bayesian Networks so it contains the program  bnlearn by Marco Scutari, which is written mostly in C, but it is called from an R wrapper. Bayesforge allows you to open notebooks in Python 3, R and Octave. It contains Anaconda for Python , but also most of the R distribution.
 
User avatar
katastrofa
Posts: 7440
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Quantum Neural Networks, Tensorflow, Quantum Computer Programming on Amazon Cloud

November 30th, 2018, 8:52 pm

You are an OO programmer. The move these days is to stateless functional programming. Much easier for parallel programming.

IMO Python will probably teach you to become a 'hobby programmer' (nothing wrong with that of course, but it is a long way from the airs and graces of those Fortran types).

What is the lifetime of undocumented Python code. i.e. when you decide to put it out to pasture and rewrite it in C++?
Yes. I am an incurable OO programmer. If in the future, I have to translate my code to C++ or java or some newer OO language, my coding style will make translation easier. I write open source that will last forever 💍💎
In theory, any code can be translated as long as the languages are Turin complete. The quantum computer is not a Turing machine, though.