Design a neural network to solve the XOR problem, i.e. the network should output 1 if only one of the two binary input variables is 1, and 0 otherwise (see in Figure 1a). Use the hyperbolic tangent, or tanh, activation function in all nodes (Figure 1b), which ranges in [-1, +1]. 1X y=1 0 0 y=0 (a) O y=0 input 1 y=1 X OR Figure 1 NOT AND Note that (A XOR B) can be expressed as (A OR B) AND NOT (A AND B), as illustrated below (see in Figure 2): AND 9.29 Figure 2 (b) tanh(x) Convert to 0/1 In the diagrams shown in Figure 3, we filled in most of the tanh units' parameters. Fill in the remaining parameters, keeping in mind that tanh outputs +1/-1, not 0/1. Note that we need to appropriately change the second layer (the AND node) to take +1/-1 as inputs. Also, we must add an extra last layer to convert the final output from +1/-1 to 0/1. Hint: Assume that tanh outputs -1 for any input x < -2, +1 for any input x +2, and 0 for x = 0.