Md. Asif Uddin
I.5.X01

Seven of the eight patterns

numeric▲▲△

Three hidden units admit eight activation patterns on paper. Using the network of I.5.B01, determine how many of the eight actually occur for some input in the plane, name the one that does not, and prove by hand that it cannot.

Hint

Each unit is on where one linear inequality holds. “All three off” is three inequalities at once — add two of them together and see what they force.

Solution

Seven of eight occur. A scan of the plane at a spacing of 0.0050.005 finds witnesses for 001001, 010010, 011011, 100100, 101101, 110110 and 111111, and never finds 000000.

Why seven is the number to expect. Three units draw three lines. In general position, three lines cut the plane into (30)+(31)+(32)=1+3+3=7\binom{3}{0} + \binom{3}{1} + \binom{3}{2} = 1 + 3 + 3 = 7 regions (I.5.4). One region per pattern, so exactly one of the eight sign vectors is unrealised. The count is not an accident of these particular weights; it is the arrangement bound, met exactly.

Which one, and why. Suppose all three units were off at some x\vec{x}:

z1=x1+0.5x20.50,z2=x1+x2+0.250,z3=0.5x1x20z_1 = x_1 + 0.5x_2 - 0.5 \le 0, \quad z_2 = -x_1 + x_2 + 0.25 \le 0, \quad z_3 = 0.5x_1 - x_2 \le 0

Add the second and third:

z2+z3=0.5x1+0.250x10.5z_2 + z_3 = -0.5x_1 + 0.25 \le 0 \quad\Longrightarrow\quad x_1 \ge 0.5

From the third alone, x20.5x10.25x_2 \ge 0.5x_1 \ge 0.25. Substituting both into the first:

z1=x1+0.5x20.5  0.5+0.5(0.25)0.5=0.125>0z_1 = x_1 + 0.5x_2 - 0.5 \ \ge\ 0.5 + 0.5(0.25) - 0.5 = 0.125 > 0

which contradicts z10z_1 \le 0. The pattern 000000 is impossible.

What the impossibility means. There is no input this network maps to zero through its hidden layer, so the constant b(2)=0.75b^{(2)} = 0.75 is never the whole output. More generally, the set of realisable patterns is a property of the weights, and it is smaller than 2n2^{n} for every network with n>dn > d: at n=10n = 10 units in the plane, 5656 of 10241024 patterns occur, or 5.5%5.5\%. Counting units and exponentiating badly overstates what a layer can do — which is the content of I.5.T1 stated as a fraction.

Draws on