This didn’t seem right to me:
For instance, you can squeeze the values of a feature between 0 and 1 by dividing every individual feature by the max value of that feature column and then subtracting by the min value.
To ensure the whole 0…1 range is used, don’t you want to first subtract the minimum value in the column (baseline to 0), then divide by the resulting maximum value?
Effectively:
normalised = (value - min) / (max - min)