I currently work on an Alienware 15 and it is sweet.
Two requirements I had were: (i) NVIDIA 10-series discrete graphics, and (2) 32MB of RAM. I wasn't excited about the 4K display until I saw it, now I am never going back. Also the build quality is great.
Apple backed the wrong horse in the GPU wars a few years back, I think all of their machines still ship with ATI GPUs if the have a discrete GPU. I think for gaming and creative work ATI is OK but if you do neural networks or other heavy calculations software support is not universal for tools like Tensorflow, Torch, etc. as it is for NVIDIA.
Most of the time I would also tell you that you can get a lot more computer for your $ and not be complaining about drivers for your OS or you don't like the ports by getting a desktop machine.
I recently bought a tower machine built from refurb parts, put in 32MB of RAM and a GTX 1050 graphics card (largest that wouldn't need special considerations for power supply) for a machine that I mainly use as a media server but may also want to run some Tensorflow. What I paid for the 1050 card was crazy because the price has been driven up by bitcoiners.
My only bitch about ML on a laptop is that it is not necessarily the kind of thing you want to do on a machine where you close the lid. For instance on evening I closed the lid when the system was uploading a model to a server and the server never got it...