#forward-propagation

A network has inputs $\mathbf{x} = (0.9, 0.7)$ and three ReLU hidden units with weight rows $(-0.3, 0.15)$, $(0.32, -0.91)$ and $(0.37, 0.47)$ and a bias of 0.001 each. Its two output units have no bias and weight rows $(0.15, -0.46, 0.59)$ and $(0.10, 0.32, -0.79)$. A softmax turns the outputs into probabilities, and the true class is the first one. What is the cross-entropy loss (natural logarithm)? Round to 3 decimals.