Show the mapping of elements of ANN with the elements of biological neural network.
Why activation function is used in Artificial neuron? Explain sigmoid, ReLU and tanh activation function along with plots showing their output values.
Explain back propagation in detail considering the following architecture where activation function at hidden layer is sigmoid. No activation function is applied at output layer. Derive equations for updating weight parameter:
Calculate gradient of weight parameter w5 (∂yp/∂w5) for the given neural network with two inputs x1=1 and x2=-1. The activation function at the output layer is : f1(x): x^3+2x^2+x+3 and activation function applied at the hidden layer is f2(x)=x^2+x+2. w1=2, w2=- 3, w3=1, w4=4, w5=2 and w6=-1
Determine the total number of parameters required for applying convolution considering the following values of a CNN: input layer (32,32,3), conv(F=5,S=1,K=6)
Determine the shape of output matrix after applying convolution for an image of size 19 x 19 x 3 that uses full padding of size 2, stride size 2, and 6 filters of size 5 x
List out four regularization techniques in deep learning. Describe any two of them.
List out three applications where RNN is required rather than CNN or ANN
Briefly describe different types of RNN based on total number of inputs-outputs
Circulate this solved paper with KaTeX formulas and 1-click AI step solvers to your batchmates on WhatsApp or Telegram.
Official Gujarat Technological University (GTU) examination paper and step-by-step solutions for Deep Learning and Neural Network (Summer 2024, B.E. · Computer Engineering, Sem 6). Features complete 70-mark regular & remedial examination pattern, official marking distribution across all 5 questions, and direct 1-click official PDF download.
Transcribed for student exam preparation from Gujarat Technological University official examination archives. Questions, syllabus guidelines, and curriculum marking schemes remain the intellectual property of Gujarat Technological University.