Training & Optimization

Check your understanding

4 questions · answer all to submit

1. 1.In machine learning, what does a loss function fundamentally represent?

   AA quantitative metric reflecting the discrepancy between predictions and targets.BThe total count of hidden layers configured within a given neural network.CThe aggregate volume of data instances present in the training dataset.DThe hyperparameter governing the magnitude of weight updates during optimization.

2. 2.What specifically defines the 'learning rate' within the context of training a machine learning model?

   AA particular class of non-linear functions applied to the outputs of neurons.BThe architectural configuration detailing neuron count within each hidden layer.CThe total count of complete iterations over the entire training dataset.DA scalar factor determining the step size taken by the optimizer to adjust weights.

3. 3.What fundamental computation occurs during the 'backward pass' phase of a neural network training iteration?

   AThe calculation of gradients, indicating how the loss changes concerning each weight.BThe procedure of randomizing the order of samples within the current training dataset.CThe dynamic modification of the neural network's structural design or layer setup.DThe process where the neural network generates its initial prediction from input data.

4. 4.What term refers to one complete iteration through the entirety of a training dataset in machine learning?

   AA mini-batch, representing a tiny subset of the full training data for processing.BAn epoch, signifying a full traversal over all available training examples in a cycle.CA checkpoint, indicating a saved state of the model parameters during the training period.DA gradient, which is a vector indicating the steepest direction of parameter adjustment.