3. Consider a deep network used for classification of given inputs into 5 mutually exclusive classes,...

Transcribed Image Text:

3. Consider a deep network used for classification of given inputs into 5 mutually exclusive classes, A, B, C, D and E. The Softmax function is used for getting the probabilities of each of the classes, and the Cost function used for training of this network is the negative log probability function C=-, log y,, t and y are standard nomenclature (t is 1 for the true class, 0 for the rest; y is the softmax output). For a given input, let the 5 inputs for softmax classification be (4, 2, 9, 1, 5). Find the corresponding 5 softmax probability values. [Marks: 8] [Marks: 6] Find the value of the cost function if the correct class for this input were (a) C and (b) E? 3. Consider a deep network used for classification of given inputs into 5 mutually exclusive classes, A, B, C, D and E. The Softmax function is used for getting the probabilities of each of the classes, and the Cost function used for training of this network is the negative log probability function C=-, log y,, t and y are standard nomenclature (t is 1 for the true class, 0 for the rest; y is the softmax output). For a given input, let the 5 inputs for softmax classification be (4, 2, 9, 1, 5). Find the corresponding 5 softmax probability values. [Marks: 8] [Marks: 6] Find the value of the cost function if the correct class for this input were (a) C and (b) E?