Base concepts of quantum machine learning

1. Coursework
Quantum machine learning
Russian-Armenian University
Faculty of machine learning
Martin Ayvazyan
10/01/2019

2. Classical computers
• Adventages
• Accurate and speedy communication machine
• Make work easy and faster
• Disadvantages
• Many kind of numerical problems can’t be solved
efficiently using conventional computers.
• Integer factorization
• Search an element in unordered array
• FFT
• Finding eigenvectors and eigenvalues
• Matrix inversion

3. World of quantum mechanics
• From Real Numbers to Complex Numbers
• From Single States to Superpositions of States
• From Locality to Nonlocality
• From Deterministic Laws to Probabilistic Laws

4. • Quantum computing applies the knowledge of
quantum physics and quantum mechanics to
manipulate elemental particles such as electrons,
photons or ions, to create processing power.
• Binary computers are based on transistors, whereas
quantum computing uses quantum bits, qubits, to
perform calculations.
• Entanglement and superposition are used to
develop special circuits that offer options beyond
the classical computers’ binary options of one and
zero
Source: https://www.quora.com/In-quantum-computing-what-is-expected-to-be-the-symbol-for-the-superposition-of-0-and-1
Quantum information

5. • A unit of quantum information is the qubit, which is
continuous valued.
• A qubit cannot be (wholly) converted into classical bits; that
is, it cannot be "read". This is the no-teleportation theorem.
• An arbitrary qubit can neither be copied, nor destroyed. This
is the content of the no cloning theorem and the no-deleting
theorem.
• Due to the volatility of quantum systems and the
impossibility of copying states, the storing of quantum
information is much more difficult than storing classical
information.
Quantum information

6. Quantum parallelism
• Unlike classical computers, where parallel computation is
performed by having several processors linked together, in
a quantum computer a single quantum processor is able to
perform multiple computations on its own, by allowing it
to work on many computation at once.

7. Big proven speedups via quantum computing
• Breaking RSA
• Database search
• Crypto
• Classification
• Recommender systems

8. Useful quantum computation is hybrid

9. Machine Learning
• Supervised learning
• Unsupervised learning
 Deep learning
 DNN
 CNN
 RNN

10. Quantum Machine Learning
QNN

11. Quantum neural network / QNN
• QNN is an Artificial Neural Network(ANN) included in a system with
Quantum Computation.
• The reason researchers are attempting this is to develop more
efficient algorithms in pattern classification or machine learning than
what is available now in the capabilities of ANN.
• Quantum computation expands the computational behavior of
machines to exponential capacity, using quantum parallelism and the
effects of interference and entanglement.
• Quantum computation promises to expedite the training of classical
neural networks and the computation of big data applications,
generating faster results

12. Quantum neural network / QNN
• Qubit Neurons (Qurons)
• A quron is a qubit in which the two levels stand for active an resting neural
firing states.
• This allows for neural network to be in a superposition of firing patterns.
Source: https://www.slideshare.net/vcamro/quantum-computing-43800212

13. QNN advantages
• memory capacity is in exponential
• lower number of hidden neurons provide higher
performance
• learns fast linearly inseparable problems for single layerhigher
stability and consistency

14. CONCLUSION
• The Quantum Neural Network is still only theoretical. Since the three
major obstacles to making it real have each had recent research
breakthroughs, this powerful deep learning method will be possible
very soon. There are lots of areas for growth and change as these
breakthroughs become translated into practical experiments on real
quantum computers.
• Some areas where QNN can be useful are:
• Investigation of the potential similarities between the underlying mechanisms
of the enhanced memory storage capacity in black holes and in brain
networks.
• Connection of Ramsey theory and unsupervised learning.
• Investigation of cancer.

15. Thank you!