Katkat
Back

Compute Conv2D parameter counts

TeslaMachine Learning Engineer · Technical Screen
EasyMachine Learning EngineerTechnical ScreenMachine Learning

Last updated

This question evaluates understanding of convolutional neural network parameterization, specifically how kernel dimensions, input/output channels and an optional bias term determine the number of learnable parameters.

For a 2D convolution layer, given input channels C_in, output channels C_out, kernel size (k_h, k_w), stride (s_h, s_w), padding (p_h, p_w), and an optional bias term, compute the total number of learnable parameters for the layer in both cases: with bias and without bias.