Exam MAS-II Practice Question 14
Study guide: Exam MAS-II
A single-hidden-layer neural network for a quantitative response is trained by gradient descent on the objective
R_i(\theta) = \tfrac{1}{2}\left(y_i - f_\theta(x_i)\right)^2, \qquad f_\theta(x) = \beta_0 + \sum_{k=1}^{K}\beta_k A_k .For one training observation, y_i = 5.0, the current prediction is f_\theta(x_i) = 3.8, and the second hidden unit's activation is A_2 = 1.5. The current value of \beta_2 is 0.6, and the learning rate is \rho = 0.1.
Calculate the value of \beta_2 after one gradient-descent step on this observation alone.
- A) 0.42
- B) 0.72
- C) 0.78
- D) 0.96
- E) 2.40
Answer this question and see the worked solution