Spring 2017
Quiz 7
Note: answers are bolded

Which of the following is able to approximate any continuous function to an arbitrary accuracy?
 A twolayer neural network (input layer, output layer) using a linear activation function.
 A twolayer neural network (input layer, output layer) using a nonlinear activation function.
 A threelayer neural network (input layer, hidden layer, output layer) using a linear activation function.
 A threelayer neural network (input layer, hidden layer, output layer) using a nonlinear activation function.

The use of sigmoid functions makes backpropagation possible because it is continuous and differentiable. Besides enabling backpropagation, the sigmoid function also makes neural network a:
 linear classifier
 nonlinear classifier

While training neural networks with at least one hidden layer (and using a nonlinear activation function), would the initialization of weight vectors have an impact on the performance of the neural network?
 No, because backpropagation using gradient descent would always find the best weights.
 Yes, Neural networks in the given configuration optimize a nonconvex objective function.
 No, Neural networks in the given configuration always optimize a convex objective function and will reach the minimum eventually.

Which of the following are reasons why one may prefer using onevsall (OvA) over allvsall (AvA) in the multiclass classification setting (Multiple choices may be correct)?
 OvA involves learning fewer classifiers than AvA
 OvA is able to learn problems that AvA cannot
 Each individual classifier for OvA receives a larger set of examples for training than for AvA (Assuming uniform label distribution)
 OvA makes weaker assumptions regarding the separability of the data than AvA does

In kclass classification, onevsall at least requires k classifiers for k different labels.
 True
 False
Dan Roth