Possible Performance Improvement: matmul vs dot vs einsum

[rickvanveen]

Is your feature request related to a problem? Please describe.
Not a problem just enhancement. Should check if there is a possible performance gain (probably mostly in the distance function) when changing any of the dot products to matmul or better einsums. Furthermore changing certain dot calls to matmul makes it more clear what is going on for anyone not that familiar with numpy.

Describe the solution you'd like
Change dot function to better einsum or matmul calls.

Describe alternatives you've considered
n/a

Additional context

• matmul is faster for larger matrices.
• dot is faster for smaller matrices