Result using a massive softmax layer that converts

This preview shows page 11 - 13 out of 15 pages.

result using a massive softmax layer that converts raw scores into a probabilitydistribution across words. With this method, the prediction's uncertainty is representedas a probability distribution across all conceivable outcomes, assuming there are afinite number of them.Forecasting “missing” frames in a video, missing patches in apicture, or missing segments in a voice signal in CV, on the other hand, entailspredicting high-dimensional continuous objects rather than discrete outcomes. There isan unlimited amount of video frames that might theoretically follow a particular videoclip. It is impossible to clearly express all potential video frames and assign them aprediction score. Adequate probability distributions over high-dimensional continuousspaces, such as the set of all potential video frames, may never be represented byalgorithms.Fig 2. Modeling the uncertainty in prediction
62.4. What are the Limitations?
Gazi University Computer Engineering Department
2.5. What are the differences from supervised/unsupervised learning?

Upload your study docs or become a

Course Hero member to access this document

Upload your study docs or become a

Course Hero member to access this document

End of preview. Want to read all 15 pages?

Upload your study docs or become a

Course Hero member to access this document

Term
Summer
Professor
NoProfessor
Tags
Machine Learning, Gazi University Computer Engineering Department

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture