Neural Networks

Check your understanding

4 questions · answer all to submit

1. 1.Which fundamental operation does a single artificial neuron perform in a neural network architecture?

   AIt computes a complex polynomial regression function from its received input features directly.BIt performs a recursive Bayesian inference to update its internal state based on input probabilities.CIt determines the Euclidean distance between its input vector and a stored prototype vector.DIt calculates a weighted sum, applies a bias, and passes the result through an activation function.

2. 2.What is the primary architectural advantage of employing multiple layers in a neural network over a single-layer perceptron?

   AMulti-layer networks learn faster due to increased parameter count and computational complexity.BThe addition of more layers inherently mitigates the vanishing gradient problem within many deep learning models.CA single-layer model is restricted to linear separability, while depth enables learning non-linear boundaries.DMultiple layers primarily serve to increase overall memory footprint, which improves the data locality.

3. 3.What is the mathematical consequence of constructing a deep neural network without incorporating any nonlinear activation functions between layers?

   AThe network's representational capacity would be significantly increased due to simplified gradient flow.BTraining convergence speed would dramatically improve because of the reduced computational complexity.CThe entire network would mathematically degenerate into an equivalent single linear transformation.DThe network would be rendered incapable of learning any meaningful patterns or relationships.

4. 4.Within the standard neural network architecture, what fundamental role do 'hidden layers' serve?

   AThey are computational units deliberately obscured from direct access by the training hardware.BThey compute and learn progressively abstract intermediate representations from the raw input.CThey store predefined feature templates that are matched against the input data.DThey define the objective function used to evaluate the network's performance during training.