I have been playing with neural nets since the 80s, and J since around 2010 or so. This was the first I had heard of ELM, and having a J implementation, I dove in. My interest is mainly academic, since I do not work in the ML community, but I have been self-studying AI/ML for decades. I would be curious to how J performs against a Julia or Python implementation of the same algorithm, since it is interpreted, but very fast. I personally like the way J makes me think and work. I leave the JQT open on my desktop at work and at home for more than just a calculator, but less than a full-blown programming tool, although it is.