Since we originally proposed the need for a first-class language, compiler and ecosystem for machine learning (ML) - a view that is increasingly shared by many, there have been plenty of interesting developments in the field. Not only have the tradeoffs in existing systems, such as TensorFlow and PyTorch, not been resolved, but they are clearer than ever now that both frameworks contain distinct "static graph" and "eager execution" interfaces. Meanwhile, the idea of ML models fundamentally being differentiable algorithms – often called differentiable programming – has caught on.

Where current frameworks fall short, several exciting new projects have sprung up that dispense with graphs entirely, to bring differentiable programming to the mainstream. Myia, by the Theano team, differentiates and compiles a subset of Python to high-performance GPU code. Swift for TensorFlow extends Swift so that compatible functions can be compiled to TensorFlow graphs. And finally, the Flux ecosystem is extending Julia’s compiler with a number of ML-focused tools, including first-class gradients, just-in-time CUDA kernel compilation, automatic batching and support for new hardware such as TPUs.

This talk will demonstrate how Julia is increasingly becoming a natural language for machine learning, the kind of libraries and applications the Julia community is building, the contributions from India (there are many!), and our plans going forward.

Julia is a programming language for data science and numerical computing. Julia is a high-level, high-performance dynamic language, it uses just-in-time compilation to provide fast machine code - dynamic code runs at about half the speed of C, and orders of magnitude faster than traditional numerical computing tools.

Julia borrows two main ideas from Lisp lore:

1. Multiple-dispatch: a method is identified by a name and the types of all of its arguments (traditional OOP only uses the type of a single argument) multiple-dispatch allows for a natural programming model suitable for mathematics and general purpose programming. Interestingly, the same feature is responsible for Julia's ability to generate really fast machine code.
2. Macros: Julia has syntax that is familiar to users of other technical computing environments, however, Julia expressions are homoiconic -- Julia code can be represented as Julia data structures, and transformed from one form to another via hygienic macros. There are some very interesting uses of macros to create domain-specific languages for effective programming. 

Julia is a modern language that combines the productivity of Python, R, Matlab, and SAS with the performance of C, Fortran and Java. In doing so, it solves the two language problem, where specialists write algorithms in a scripting language and then programmers deploy it in a systems language.

In this talk, I will share our experience about the evolution of the Julia community - how it grew from 3 contributors to 500, and from 0 to 1000 packages in the last 3 years.

In this talk, I will talk about our motivation for creating Julia. Julia is a high-level, high-performance dynamic programming language. It provides a sophisticated compiler, distributed parallel execution, numerical accuracy, and an extensive mathematical function library. Julia’s Base library, largely written in Julia itself, also integrates mature, best-of-breed open source C and Fortran libraries for linear algebra, random number generation, signal processing, and string processing. In addition, the Julia developer community is contributing a number of external packages through Julia’s built-in package manager at a rapid pace. This is why Julia is seeing rapid adoption in universities for teaching and research, as well as in businesses.

I will discuss what makes Julia fast. Julia's ability to combine these levels of performance and productivity in a single language stems from the choice of a number of features that work well with each other:

1. An expressive parametric type system, allowing optional type annotations;
2. Multiple dispatch using those types to select implementations;
3. A dynamic dataflow type inference algorithm allowing types of most expressions to be inferred;
4. Aggressive code specialization against run-time types;
5. Metaprogramming;
6. Just-In-Time compilation using the LLVM compiler framework; and
7. Careful design of the language and standard library to be amenable to type analysis;

I will also touch upon how the language design has made it possible to explore many parallel programming models in Julia, completely within the language.

See: http://www.julialang.org/