CCBR Computer Pracs

MIBR 2020 : deep supervised learning (Jan 4)

• Ex1_basic classification (fashion, handwritten digits)
• Ex2_classification_transfer_learning
• Ex3_DenoisingAutoencoder on handwritten digits
• Ex4_Unet for image to image learning (facades)
• Ex5_Standard image classification models (mobilenet)
• Ex6_Word Prediction (shakespeare)

More NLP samples and assignment https://github.com/shiva16/ccbrmllab

CCBR >> MIBR 2018 >> Computer lab / Jan 7

Codes and material used in the session

Further

Email: keerthi@ htic.iitm.ac.in

http://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html

Assignment for 2018 students

Describe a cool application for which deep learning could be used. Try to answer the following questions:

• Is this impossible if not for deep learning?
• Has someone already tried it? Who? When? What is the outcome?
• Which are the enabling prior deep networks which you could leverage if you are to build your application?
• What do you imagine is the data required to make your learner? What would be a good architecture to begin experimenting?
• What is the performance metric? How stringent is it?
• What are the technical challenges you foresee if you are to set out to build your application?