Close

Filter courses

Show courses Show filters
Close

Machine Learning for Semiconductor Quantum Devices

01 /Supervised Learning for quantum Dots 02 /Charge Stability Diagrams 03 /Classify CS Diagrams with Neural Networks 04 /Supervised Learning with Feedback 05 /Charge Tuning as a NN with Feedback 06 /Real-Life Experimental Applications 07 /Unsupervised Learning for Quantum Dots 08 /Analyse CS Diagrams with Clustering 09 /Analyse CS Diagrams with Unsupervised NNs 10 /Autoencoders in QD Experiments 11 /ML Fine Tuning of Quantum Dots 12 /Overhauser Field estimation 13 /Recap 14 /State of the Art Tuning

Analyse CS Diagrams with Unsupervised NNs

In this video we explain how to employ Neural Networks trained without supervision to identify features in Charge Stability Diagrams. We explain the notion of latent space, and show how to analyze it.

Main takeaways

  • Neural networks can be trained even without labels
  • Autoencoder is an example of a neural network generative model that can be trained on a label-free dataset
  • The latent space of an autoencoder can be used to identify relevant features in a dataset

Further thinking

True or False: The Encoder network in an Autoencoder is responsible for mapping an input to a latent space.

Solution

Further reading

More About Clustering - PCA and more: https://ml-lectures.org/docs/structuring_data/ml_without_neural_network.html

Kernel principal component analysis: https://link.springer.com/chapter/10.1007/BFb0020217

More about Unsupervised Learning - Autoencoders and more: https://ml-lectures.org/docs/unsupervised_learning/ml_unsupervised.html

Reducing the Dimensionality of Data with Neural Networks: Book Chapter: https://ml-lectures.org/docs/structuring_data/ml_without_neural_network.html