ANNS-06

Overfitting: the model learns not only the [] features but also the [] patterns of the finite dataset

• general - large

• general - strange

• small - large

• small - strange

By [] the model’s size (complexity), the flexibility [], so it will become more accurate on training examples

• increasing - decreases

• • decreasing - decreases

• increasing - increases

• • decreasing - increases

overfitting generally occurs near the [] of training by backpropagation

• second quarter

• middle

• begining

• end

[] regularizes the network by modifying the loss function

• Weight decay

• Dropout

• Ensemble

• noise injection

Dropout drops certain [] according to a random probability.

• inputs

• neurons

• weights

• output neurons

Which technique helps to prevent overfitting by combining multiple models?

• Batch Normalization

• Data Augmentation

• Ensemble Learning

• Dropout