Watch out for Julia

I first heard of Julia from Kaggle’s “First Steps With Julia” competition and started looking around some of its features while on julialang.org and at that point I thought that it was maybe a replacement for R or even a more hardcore version of R developed especially for scientists.

But recently I read a blog post from Microsoft’s Cortana Intelligence and Machine Learning Blog on Julia – “Julia – A Fresh Approach to Numerical Computing” and I have to say that I feel like the tide will turn in a year or so and we will see a rapid adoption of Julia in the Machine Learning community.

The blog post from Microsoft’s blog was authored by one of Julia’s co-creators, Viral B. Shah, who pointed one of Julia’s major promises, the one to eliminate the so-called “two language problem” where you would use one high-level language for prototyping and drop down into C code in order to squeeze maximum performance for the production code.

Julia overcomes this because there is no penalty in terms of performance for using either high-level or abstract constructs, essentially bridging the gap of coding differences between an engineer and a researcher.

There are many more good things to be read about Julia, like GPU support for ML just to name one. I recommend reading the articles linked in this post for more information.