Main takeaways

• In practical implementation more robust neural architectures are needed – in the presented case here we discussed an option with multiple outputs 
• Neural network was experimentally demonstrated to successfully tune concrete charge states 

Further thinking

What is the most common reason for neural network misclassification that was experimentally observed?

a. Extra observed transition that yields over-counting of the electron number.
b. Extra observed transition that yields under-counting of the electron number.
c. Missed transition that yields over-counting of the electron number.
d. Missed transition that yields under-counting of the electron number.

Solution

Further reading

Experimental implementation of charge tuning discussed in this lecture: https://journals.aps.org/prapplied/abstract/10.1103/PhysRevApplied.13.054019