While for a t-SNE the axis should always be in the same scale (as is now), when we plot independent variables on the two axis, it does not make a whole lot of sense to require that (and leave a lot of unused space in the viewport, see screenshot).
Consider a car dataset that has horsepower and weight, it does not make sense that a delta of 1 in both variables must have the same delta in onscreen pixels.
While for a t-SNE the axis should always be in the same scale (as is now), when we plot independent variables on the two axis, it does not make a whole lot of sense to require that (and leave a lot of unused space in the viewport, see screenshot).
Consider a car dataset that has horsepower and weight, it does not make sense that a delta of 1 in both variables must have the same delta in onscreen pixels.