... one^2.1

Note that because of the duality between B and B^*, we could also have considered $\beta$ to be the angle between $\bold{b_i^*}$ and the hyperplane spanned by $\bold{b_1^*},\ldots{},\bold{b_{i-1}^*},\bold{b_{i+1}^*},\ldots{}\bold{b_n^*}$.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... lattices^2.2

In fact, we use the same set of integer lattices used in the previous section for comparing the lazy and greedy selection mechanism.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... 2-moves^2.3

We use ``k-move'' to designate a row operation in which multiples of k-1 vectors are added simultaneously to another vector.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...)^3.1

Three additional stages are used for subset sum problems with $n \geq 90$ to help reduce error.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... decreased^3.2

The loop termination condition in Equation 3.7 is strictly a heuristic. Other terminating conditions could certainly be used.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.