Machine Learning Question. Consider a 2 Layer neural network with no bias units for regression problems where the input

[answerhappygod]

Machine Learning Question Consider A 2 Layer Neural Network With No Bias Units For Regression Problems Where The Input 1 (84.87 KiB) Viewed 94 times

Machine Learning Question. Consider a 2 Layer neural network with no bias units for regression problems where the input feature dimension is dand the output targed is one dimension. The regularized neural network loss function L(W) can be defined as: L(W) = {(t-o(XW))"(t - (XW))+{WTW where sigma is the sigmoid activation function, X is the data matrix in size Nx d/N: The number of data; d: the dimension or the number of features). t denotes the N target values in Nxl, and W=[w11,w21.....,Wd1j^T the weight matrix of size dx1, and 120 is given regularizer. Questions 1. For you to use Gradient Descent (GD) algorithm to minimize L(W) with respect to W, derive the matrix formula for the gradient BLW of the loss funtion L(W). Show your detailed derivation step by step and present your solution. 2. Write out the GD updating formula for the first parameter wil explicitly (not in matrix form) with a learning rate a > 0 Hint (x'e) a ЭХ 8( Ax) a(xx) ОХ = 2x