Dropout We now present the connection between dropout method and ridge regression (outlined in more detail in Wager et al.) To recap, dropout randomly drops units along with their input/output connections. We now want to apply this method to our simple setting. Let us define the indicator random variable Iij to be whether the j 'th neuron is present or not in predicting the response of the i 'th sample network for i 'th sample becomes j=1pIijjxij where Iij={011withprobabilitywithprobability1,1jp,1in, drawn independently from the training dataset. Note that E[Iij]=1, thus the output of the network is f(xi) on average. Question 2: Write down explicitly the loss function after using the dropout as a function of I=(Iij)in,jp denoted by L(,I). Dropout We now present the connection between dropout method and ridge regression (outlined in more detail in Wager et al.) To recap, dropout randomly drops units along with their input/output connections. We now want to apply this method to our simple setting. Let us define the indicator random variable Iij to be whether the j 'th neuron is present or not in predicting the response of the i 'th sample network for i 'th sample becomes j=1pIijjxij where Iij={011withprobabilitywithprobability1,1jp,1in, drawn independently from the training dataset. Note that E[Iij]=1, thus the output of the network is f(xi) on average. Question 2: Write down explicitly the loss function after using the dropout as a function of I=(Iij)in,jp denoted by L(,I)