We’re considering coding a julia application in 2022. We’re based in streeterville ( by the MCA) and would like to chat with any NW’ers about julia. My wife is a NW alum. Anyone that can get to Streeterville is welcome.
We’re going to need some help in our application and prepared to pay for it.
We’ll probably use the same approach as we did in silicon valley ( real life NOT Pied Piper :-)) which was pretty successful.
Hi Logan
welcome. As to what we are doing, we’re writing a trading augmentation tool for our little partnership. The 1st one ( TONTINE) worked spectacularly in a volatile options and futures environment. That was written in Python. The next ( TONTINE2, see we lack imagination ) will be written in julia probably using dagger.jl and MAYBE using Pluto.jl as the IDE. Don’t panic it’s a complete rewrite so a blank sheet of paper.
The actual goal is to have an adaptive system which produces graphical representation of complex data sets to Tufte design rules.
Rather than doing imperative programming with Dagger.jl have you considered a declarative approach? The difference being, with declarative you tell the computer what you want done versus imperative’s step by step instructions.
The declarative approach looks at the request and generates optimal code (code-gen). Typically the declarative interface is the Structured Query Language (SQL). The authors of these various systems have spent decades building optimizations into the cost based optimizers, etc to make the declarative level interface very fast. You are unlikely to exceed these expert’s efforts unless you have hundreds of millions of dollars in R&D.
With Swift for TensorFlow being deprecated by Google and Python at its evolutionary end…Julia is the way forward for machine learning and data science. You get the best of both worlds by combining a declarative tuple relational calculus (SQL) with Julia’s machine learning packages.
some history I think. In our youth we were mentored by Dr Jim Gray who literally wrote the precusor to SQL (R) and the book Transaction Processing: and developed a considerable amount of the body of work that is the database theory of today.
so we have some experience with high throughput massively distributed systems using SQL and it’s ilk. Because we use a data abstraction layer we don’t care what the data looks like or where it lives. We watch the fine people at ursalabs and many others to make sure we are architecting this properly to adapt to as much as we can perceive the future has in store. We can change our process flow easily by the implementation of finite state machines of finite state machines.
We think we are happy with our choices and will stick with dagger.jl and Pluto.jl. Not completely onboard with your statement that “Python at its evolutionary end” (sic), We like Python and know that evolution will always find a way.
If you caught me two years ago, I would have been at Feinberg. I’ll ask around. You might try to see if the leadership of the Biomedical Informatics and Data Science student group is responsive or could advertise for you: https://sites.northwestern.edu/bids/people/
Dr. Ted Codd’s tuple relational calculus introduced in the “A Relational Model of Data for Large Shared Data Banks" paper was before R.
The power of the tuple relational calculus is if data is normalized properly one adapt the “process flow” easily using joins.
One of the reasons Julia was created was to solve the “two language problem”. In data science, Python is a glue language for optimized C++ and Fortran libraries. If you have smaller datasets or are OK with writing the computational intensive parts in another language then Python can work. Julia has had the benefit of being designed much later and has incorporated very sound design principles. Over time, I believe the cutting edge work will move to languages better suited for the work.
hi @mkitti
that’s the secret sauce of the valley, chance encounters. Or at least it USED to be before the castles went up sorry to have missed you but I am SURE that NW helped you on your new adventure.
Thank you so much for the hints and we’ll be in Feinberg today after watching the joggers and the dogs in the park. It’s too cold to go to the lakefront right now.
thanks again
theakson
sorry to have missed you but I am SURE that NW helped you on your new adventure.