What a language is good at depends a little on its basic syntax and semantics, but a whole lot more on what libraries are easily available. If you don't think of JavaScript as a math language then perhaps you need to meet Math.js.
Math.js is a simple extension to the standard Math object that you should be familiar with. For example
math.pow(2,3)
returns 2 raised to the power 3. Math.js also has a pow function but it can raise complex numbers, big numbers or matrices to a power.
One of the most important features of Math.js is that is supports a range of data types and, as far as possible all of its functions work with all of the available data types. You can use standard number, Bignumber i.e. arbitrary precision, complex numbers and matrices. You can also define units for physical calculations.
For example:
var a = math.complex(2, 3); var b = math.complex('4  2i');
defines complex numbers and you can work with them using
math.add(a,b);
and similar functions. You can also use the eval method to work out arithmetic expressions involving any of the data types without having to use the special functions. For example:
math.eval('a+b');
does complex addition if a and b are complex.
As well as the basic operators, you also have some more advanced function such as math.inv to find the inverse of a matrix and a small number of probability and stats functions.
It has to be admitted that the range of facilities doesn't put Math.js into same class as Python's NumPy, but it is heading in the right direction. You can use it in the browser or as a Node.js library and, of course, it has the advantage of working in the same way on the client or server. It is open source and you can alway help extend it.
The big problem with JavaScript and number crunching is speed. It would be interesting to see if Math.js could be optimized by introducing the asm.js data types. This would be a lot of work but the result would be something that worked at close to the speeds achevable by native code.
Detecting When A Neural Network Is Being Fooled 08/03/2017
Adversarial images are the biggest unsolved problem in AI at the moment and progress is being made but for all the wrong reasons. Now we have some progress on detecting when an image has been speciall [ ... ]

A Single Perturbation Can Fool Deep Learning 25/03/2017
If you have been reading our reports on adversarial images, the headline should come as no surprise. What is a surprise is the way that AI researchers are regarding such images as security threats rat [ ... ]
 More News 
