Example notebook PFSP

[leonlan]

Closes #106.

[leonlan]

This is almost done. I'll have another read tomorrow morning and then this can be reviewed.

[N-Wouda]

Notebook:

• "This notebook is largely based on Stützle and Ruiz (2018)", I would rewrite as "implements ideas of". They came up with ideas, but it's your implementation.

• with open("tai50_20_8.txt", 'r') as fi: -> can we make this into a classmethod on Data?

• Do the colours mean anything in the Gantt chart?

• Gantt chart y-label: "Machine", not "Machines"

Both destroy operators rely on the n_remove parameter, which should inferred from the **kwargs argument. We also set a default value of 2 in case it’s not provided as keyword argument.

Or:

def random_removal(state, rnd_state, n_remove=2):
    ...

And then mention n_remove can be passed as a kwarg to ALNS.iterate. Testing if kwargs has a thing is a bit weird - either pass them along, or have defaults for parameters your need.

• NEH ordering - what does NEH mean?

• "Autofitting acceptance criteria" section, first parag: list doesn't render well. I think there needs to be a white line separating it from the text.

• pct_diff = 100 * (objective - data.bkv) / bkv: this only works because the with open() part pollutes the global namespace (so bkv also exists, but perhaps should not). I'd use data.<value> throughout. Maybe write DATA to make its 'globalness' stand out a bit more.

• Most of the notebook uses SA, but tuning suddenly uses RRT (without explanation). That feels a bit weird; maybe stick to SA throughout.

• print(objectives) is somewhat unnecessary if we'll get a plot in the next cell already.

• Text(0.5, 0, 'n_destroy') you can avoid by putting a semicolon after the last statement in the cell, like so: ax.set_xlabel("n_destroy");.

What happens if you use 0.01 or 0.05 for $\alpha$? I think 0.1 is fairly large still, and can imagine (much) smaller values work even better.

[N-Wouda]

This LGTM, and I think most of what I remark on above can be updated really quickly.

[leonlan]

@N-Wouda I addressed all comments in the latest version! Can you merge this if you're OK with this?

[N-Wouda]

Yeah, I think it's OK now. I fixed two small typo's.

[N-Wouda]

This is a very nice addition to the examples. We have a wide range of examples now, which should help people get started quickly.