Question

4. Consider a MLP with one input, two layers, one neuron in each layer, and one output. The activation function of the first layer is $f^{(1)}(n) = 2n$ and the activation function of the second layer is $f^{(2)}(n) = 3n$. The initial weights of the first and the second layers are respectively $w^{(1)} = 0$ and $w^{(2)} = 0$. There are no bias terms, so $b^{(1)} = b^{(2)} = 0$ and they are kept zero during training. Assume that we present the data point with $x = 1$ and y = 2 to the network. Perform one step of the Stochastic Gradient Descent algorithm by using the backpropagation algorithm, assuming the learning rate $alpha = 0.5$. Use the MSE objective function, i.e. $J = (y - a^{(2)})^2$. This means that you should calculate the updated weights.

          4. Consider a MLP with one input, two layers, one neuron in each layer, and one output.
The activation function of the first layer is $f^{(1)}(n) = 2n$ and the activation function of
the second layer is $f^{(2)}(n) = 3n$. The initial weights of the first and the second layers are
respectively $w^{(1)} = 0$ and $w^{(2)} = 0$. There are no bias terms, so $b^{(1)} = b^{(2)} = 0$ and they
are kept zero during training. Assume that we present the data point with $x = 1$ and
y = 2 to the network. Perform one step of the Stochastic Gradient Descent algorithm
by using the backpropagation algorithm, assuming the learning rate $alpha = 0.5$. Use the
MSE objective function, i.e. $J = (y - a^{(2)})^2$. This means that you should calculate the
updated weights.

4. Consider a MLP with one input, two layers, one neuron in each layer, and one output.
The activation function of the first layer is f^(1)(n) = 2n and the activation function of
the second layer is f^(2)(n) = 3n. The initial weights of the first and the second layers are
respectively w^(1) = 0 and w^(2) = 0. There are no bias terms, so b^(1) = b^(2) = 0 and they
are kept zero during training. Assume that we present the data point with x = 1 and
y = 2 to the network. Perform one step of the Stochastic Gradient Descent algorithm
by using the backpropagation algorithm, assuming the learning rate alpha = 0.5. Use the
MSE objective function, i.e. J = (y - a^(2))^2. This means that you should calculate the
updated weights.

Added by Wesley P.

Question

Please give Ace some feedback