计算机视觉代写 | Intro to Machine Learning

本次美国代写是关于Python计算机视觉相关的assignment

• Previously: we assume knowledge about the world is given

• Automated discovery of patterns in data and acting on it

• Data → pattern → model → action/decision

• Model: a simplification of some phenomenon

• MDP: transition function

• Sensor stochasticity: Bayes net

Data analytics

Making decisions

Finding patterns in data

• In Machine Learning we no longer make the assumption that models are given

• Too complicated

• Too many factors – don’t know what is important

• Model is not static – needs to be regularly updated

• Instead: give the agent a bunch of data about the phenomena and let the agent reverse-engineer a model

Types of Learning

• Supervised learning

• Give correct answer for each instance

• Learn a function from examples of inputs/outputs

• Take away the supervision and interpolate to new instances

Example data: inputi Expected_outputi

F(input) Predicted_outputi

Check expected_outputagainst predicted_ouput

If they differ, update f()

Eat at a restaurant?

• Features (observables):

• Alternate: suitable alternate restaurant nearby (y/n

• Bar: A bar to wait in (y/n)

• Fri/Sat: it’s a Friday or Saturday (y/n)

• Hungry: y/n

• Price: price range ($, $$, $$$)

• Raining: y/n

• Reservation: we made a reservation (y/n)

• Type: french, italian, thai, burger

• WaitEstimate: 0-10, 10-30, 30-60, >60

• Patrons: none, some, full

Training set

Supervised learning is function approximation