BIBA - Clustering - Deep Neural Networks

9 important questions on BIBA - Clustering - Deep Neural Networks

What is a deep neural netowrk?

A deep neural network is networks that have many hidden layers and neurons are connected with each other. The output of every node is input for the next one.

What is the advantage of Deep learning in contrast to Machine learning?

  • In ML you always need to know the predictor, however in DL we expect something to predict something else and you can run the model. The back propagation makes sure the model learns and it can handle large amount of data.
  • Can handle large amount of data
  • High predictive performance

Why does a DL model starts to overfit at one point and how do you counter this?

The model continuously updates itself and therefore in time it will overfit the data.

Overfitting regularization:
  • Early stopping
  • Drop out regularization
  • L1/L2 regularization
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What is a convolutional neural network (CNN)?

A network that is used for image recognition in self-driving cars and face recognition.

How does a convulational layers calculates based on the input image?

The input image is devided into pixels values. You slide a filter over the input image and multiply the pixel value with the filter value, then sum this output as athe output value.

What does the pooling layer?

Reduce the number of parameters and summarize the features.

What is max pooling and average pooling?

Max Pooling: Takes the maximum value of the filter
Average pooling: take the average value of filter per section

What type of learning is used in Convolutional Neural Network and is the implication?

  • Supervised learning, data labeling (e.g. Selfdriving car: people on the street, other cars etc,)
    • If you input labeled data not well, the output will be bad as well.
  • Unsupervised learning, find patterns in data, don't need labels.

What are challenges within convolutional neural network?

  • Ground truth / missing labels (expert labeling)
  • Transparency
  • Unbalanced datasets 

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo