DeepLearning

After adding large numbers of neurons, the network exhibits a significantly lower training loss during training. What possibly happened?

• The complexity of the network is too low, and underfitting occurs.

• The complexity of the network is too high, and overfitting occurs.

• The model is able to fit the trainings data well.

• The adaptation of the network by backpropagation fails.

What distinguishes Deep Learning from machine learning?

• Usage of unlabeled data.

• Deep Learning is a category of machine learning.

• Feature extraction is done by the model in Deep Learning, in machine learning pre-selected features are used.

• Deep Learning models used is usually more complex.

You realize that your model is underfitted. What measures can you take to prevent this?

• Increasing the complexity of the model (e.g., higher number of neurons or layers of the model).

• Stop the training earlier.

• Perform data augmentation.

• Increase the amount of data.

You only have unlabeled data available, what methods could you use?

• Traditional Approach

• Reinforcement learning

• Unsupervised learning

• Supervised learning

What is the name of the algorithm used to adjust the node weightings?

• Data augmentation

• Forward propagation

• Backpropagation

• Inference

Why is data augmentation beneficial?

• Improve the performance of underfitted models.

• This can lead to better performance of the models, e.g. by reducing overfitting.

• Increase the amount of data with little effort.

• Accelerates training.

What is the role of the loss function in the training process of a deep learning model?

• To measure the difference between the predicted and actual output.

• To compute the accuracy of the model on the training data.

• To regularize the weights of the neural network.

• To initialize the weights of the neural network.

What is the primary goal of the training process in deep learning?

• To make the model fit the training data perfectly.

• To minimize the loss function on the validation data.

• To generalize the model to unseen data.

• To maximize the accuracy of the model on the training data.