Quantum Computing & Information - Short Review On Current Research Trends Using Cartoons
Quantum Computing & Information - Short Review On Current Research Trends Using Cartoons
1
NSU Optics Lab (for NSUers) & Mahdy Research Academy
(Open for all Bangladeshi University Students)
DHAKA SAHARA
Future Possibilities: If Error & Quantum Noise can be
reduced
Topic of the seminar
|𝑞𝑢𝑎𝑛𝑡𝑢𝑚 𝑠𝑡𝑎𝑡𝑒ۧ
|𝑑𝑒𝑎𝑑 ۧ + |𝑎𝑙𝑖𝑣𝑒ۧ
=
2
𝐴𝑓𝑡𝑒𝑟 𝑚𝑒𝑎𝑠𝑢𝑟𝑚𝑒𝑛𝑡
?
2. Measurement in QM: Wave Packet & its reduction
17
2.1. Qubits
18
2.1. BIT VS QUBIT
NORMAL COMPUTER
HOME OFFICE
3.2. Cybersecurity in Quantum Era: Basic Cryptography
3.2. BB84 protocol: Creating secret key before the actual encryption
3.2. BB84 protocol: In Presence of a Hacker
3.2. E91 protocol: More Secured
Results
Both figures depict, Encryption of
Image on Alice’s end and
Decryption of Image on Bobs end
using chaotic parameters failed decryption
● We use knowledge distillation in quantum model just like a classical setup, but, we add the supervision from a
classical model instead of a quantum model.
● Why should this work? Since a classical model outputs a probability distribution, and so does a quantum model
upon performing measurement.
Experimental results on MNIST
● We perform baseline experiments on the students (quantum models) without using any knowledge distillation.
● We train multiple classical models to act as teachers during knowledge distillation.
● Then, we use each of the teacher models to train the student quantum models using knowledge distillation.
● We observe that using knowledge distillation in this way improves the performance of the quantum models.
Experimental results on FashionMNIST
● Similarly, We perform the experiments on the FashionMNIST dataset as well, and observe that using knowledge distillation in
this way improves the performance of the quantum models.
● Thus, we can conclude that using classical models, it is possible to improve the performance of quantum models. This
eliminates the necessity of building and training large quantum models to act as teachers for smaller quantum models.
Moreover, it ensures re-usability of classical models and eases the infancy of quantum machine learning and current resource
constraints.
THANK YOU