# Quantum machine learning in context

•  Part 1: Quantum Expectation What can we reasonably expect from quantum resources?
• Ask them what sorts of things they know about QC?
• What are some common claims in media?
• What can we actually expect?
• Part 2: Key concepts in Quantum computing
• What do we need to make a classical computer?
• what do we need for a quantum computer
• qubit (bit)
• superposition
• operations
• entanglement
•  Part 3: Quantum Programing What does a quantum algorithim look like?
• How is a quantum algorithim like a writing desk?
• How do we represent algorithims now?
• circuit diagrams are common in theory
• but are hard to translate to coding/simulations
• Many options, Q# allows us to program the way we think at an algorithimic level.
• How many quantum algorthims do we know?
• < 20, only about ~5 that are known to have some potentially useful advantages
• The one you have probably heard of
• The ones you should know about
• The real anwser is TBD
• dj: smaller questions= less information needed out
• Part 4: Where can we apply quantum to ML?

• deep learning
• reinforcement learning
• Solve: A.x = b if A is invertable then x = A^{-1}b
 A. x> = b> ->

HHL can do this in exponentially less time than classical

But what do you do with x?

If you try to read out x, you will do log worse than classical

 So either we have to make x> very small, or only need to know some restricted info about x>
• looks like linear least squares fitting from ML! b-> Data, A->Model, x->parameters

N.B. only works for A that are easily simulateable

Can train faster restricted boltzman machines faster on a quantum computer, but encoding the problem you blow away

N.B. All of these make great subroutines in larger algorithims, but we need structure for problems that does not blow away advantages of translating between classical and quantum.

• Part 1: quantum resources, what are they?
○ Qubit, what it is <Whiteboard>
○ Superposition <Whiteboard>
○ Entanglement <Whiteboard>

• Part 2: Computing?
○ How does this make a quantum computer
○ What can quantum computers do
§ Deutch Jorza
§ Grovers
§ Shor
§ Others
○ How are they beating classical?
§ Something of an open question
§ Better at noisy functions
§ Improvement is heavily state dependent
§ Not strictly better at deterministic functions
○ Common Misconceptions
§ Superposition does not mean in two states at once
§ Quantum teleportation teleports information, not anything physical
§ All quantum systems are non-deterministic, results are probabilistic

§ http://www.smbc-comics.com/comic/the-talk-3
§ DWAVE is not a quantum computer
□ It is not proven to be quantum, this is still debated
□ Even if it is quantum it is analogue
□ It is not a computer, it is an annealer
§ Quantum crypto, just like classical crypto is not infallible. Just because the protocols work doesn’t mean the implementation does
§ No such thing as quantum safe, just because it hasn’t been broken with any of the grand total of three algorithms doesn’t make it secure
§ There may exist algorithms that quantum computers are not good at solving, this does not mean that classical computers will also be bad at solving them

• Part 3: QKD
○ How can quantum resources make things secure?
○ QKD protocols?
§ SPACE!!!!!
○ Why is this not enough?
§ SIDECHANNELS.
§ Poor implementations