Introduction to Machine Learning
Introduction
Prof. Kutty
Copyright By PowCoder代写 加微信 powcoder
Today’s Agenda
• Introduction: what is this class about
• Administrative: resources, grading etc.
• What is Machine Learning?
• Supervised Learning
Be Responsiblue!
• This course will adhere to the university mask policy. Per the university guidelines: anyone attending an in person component is required to wear a mask. If a person is not wearing a mask, we will ask you to put one on. If the person does not have a mask, there will be a limited number of extra masks available in the classroom. If the person refuses to wear a mask, the person may be asked to leave the class or the instructor (the course professors or staff) may choose to leave the classroom and resume the lecture remotely.
– TLDR: please wear a mask
• As they have throughout the past year and a half, policies around academic and public health are subject to change as this pandemic evolves. This course will follow all policies issued by the University, which are documented on the Campus Blueprint’s FAQ. These policies may change over the course of the term, so please review the Campus Blueprint’s FAQ for the most up to date information.
Course Overview
Machine Learning
“a set of methods that can automatically detect patterns in data, and then use the uncovered patterns to predict future data, or perform other kinds of decision making”
– Murphy 2012
no explicit programming of
Title learning
Machine Learning is ubiquitous
An analogy…
Plastic: A Toxic Love Story
”But it wasn’t clear to me just how plastic my world had become until I decided to go an entire day without touching anything plastic. The absurdity of this experiment became apparent about ten seconds into the appointed morning when I shuffled bleary-eyed into the bathroom: the toilet seat was plastic.”
Machine Learning is ubiquitous
Applications of Machine Learning
Convolutional
Network CNN project 2
classification
with a mask
recommender
collaborative
multi armed
bandit Reinforcement
”Now, when there are multiple stories related to your search, we’ll also organize the results by story so it’s easier to understand what’s most relevant and you can make a more informed decision on which specific articles to explore. ”
unsupervised learning clustering algorithm
probability theory graphical and me or y models
Toxicity in Machine Learning Use with caution!
Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I. and Fergus, R., 2013. Intriguing properties of neural networks.
Buolamwini, J., Gebru, T. [2018]
“Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification.”
In January 2020, was wrongfully arrested based on a flawed match from a facial recognition algorithm https://www.nytimes.com/2020/06/24/technology/facial-recognition- arrest.html
“HasAsthma(x) ⇒ LowerRisk(x)”
Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission
Caruana et. al. 2015
trade-off between interpretability and accuracy
What is this class about? according to the course catalog
Theory and implementation of state-of-the-art machine learning algorithms for large-scale real-world applications. Topics include supervised learning (regression, classification, kernel methods, neural networks, and regularization) and unsupervised learning (clustering, density estimation, and dimensionality reduction).
What this course is not
• focused only on applied machine learning
– we are interested in the origins of algorithm and their
mathematical interpretation
– be prepared for some math-heavy assignments!
• focused only on theoretical machine learning
– we are also interested in applying algorithms to datasets to get
hands-on experience with the algorithms
– be prepared for some programming-heavy assignments!
Warning – math ahead!!
• We will focus a lot of lecture on derivations to get a deeper understanding of the underlying algorithms
– optimization-based and statistics-heavy perspective!
• This course is not for you if you want to just learn to apply
packages to datasets!
• This course is for you if you want to develop a deep understanding of and learn to implement ML algorithms
Prerequisites
• (Enforced) MATH 214 or 217 or 296 or 417 or 419 or ROB 101
• (Advisory) STATS 250 or equivalent (STATS 412 is good)
• (Enforced) EECS 281
– some experience in Python/MATLAB is helpful – nontrivial level of programming is required.
see wolverineaccess for up to date details on prerequisites
Today’s Agenda
• Introduction: what is this class about
• Administrative: resources, grading etc.
• What is Machine Learning?
• Supervised Learning
Course Logistics
Course staff
• Professor:
(she/her/hers)
(he/him/his) (he/him/his) (he/him/his) (she/her/hers) (he/him/his) (she/her/hers) Jiamin Yang (she/her/hers)
• • • • • • •
Things I do (and love…)
EECS 445. Introduction to Machine Learning
DS-Eng CS-LSA
Resources: Canvas
For Logistics. Contains Calendar, Syllabus, Schedule, HWs, Projects, Lecture Recordings and Slides, Discussion, link to Piazza, Gradescope etc..
Campus Blueprint’s FAQ.
Course Schedule
Resources: Piazza
• For course material discussion only
– for logistic questions please email staff or profs directly
• Piazza can help you connect with other students in the class
• You can answer each other’s questions
• You can clarify confusion on course material
• Code of conduct: Be respectful of each other and course staff
It is our intention that students from all backgrounds and perspectives will be well served by this course, and that the diversity that students bring to this class will be viewed as an asset. We welcome individuals of all ages, backgrounds, beliefs, ethnicities, genders, gender identities, gender expressions, national origins, religious affiliations, sexual orientations, socioeconomic background, family education level, ability – and other visible and nonvisible differences. All members of this class are expected to contribute to a respectful, welcoming and inclusive environment for every other member of the class. Your suggestions are encouraged and appreciated. Further, we will gladly honor your request to address you by an alternate name or gender pronoun.
Resources: email
• Use piazza for collaborative learning
• For questions meant for course staff/professor contact us
– e.g., logistics, grading, extensions, scheduling etc. – multiple modes:
• Lectures, Discussions and Office hours (See Canvas – Google Calendar)
• Staff E-mail: • You may contact me directly
– include [EECS445 W22] in subject line
Resources: Lectures
• In person lectures (Sections 001, 002) attend either or both!
• Lecture recordings will be posted on canvas
• Annotated and unannotated slides will also be provided
• Weekly quizzes will be due at the end of the week (typically) on Sundays by 10pm ET.
– Most will be graded for correctness
– Quiz 0 is a bonus quiz
• Optional readings will be assigned to complement lectures
• What if you have questions?
– in person lectures/discussions – Office hours
– : Discussion Notes
Starting this week
We will cover:
– Review of prerequisites
• this week: Linear Algebra, Matrix Calculus, see notes for Probability Theory + numpy
– Review Lecture materials
– Give help/hints to homework problems
Additional live review sessions:
– Python tutorial coming soon: in person + recorded (see calendar for details)
– Project information sessions
– Exam review sessions
Resources: Office Hours
see calendar for modality
email professor for appointment
Resources: (optional) textbooks
• A Course in Machine Learning by III (available online)
• Pattern Recognition and Machine Learning by
(available online)
• Machine learning: A Probabilistic Perspective by
(available online)
• Hands-on Machine Learning with Scikit-Learn and TensorFlow by
• Mining of Massive Datasets by Leskovec, Rajaraman and Ullman
(available online)
• Fairness And Machine Learning by Barocas, Hardt and Narayanan (available online)
additional online resources may be linked to in notes or schedule. Canvas Files/ Course Info: EECS 445 Winter 2022 – Syllabus.pdf
Study group
• Collaborative learning helps! Form your study group early on!
• For homework, you may discuss concepts between the study group members, but you should write your own solution independently. Never share your solutions!
• In the homework submissions, you must put the names of your collaborators if you have discussed ideas in depth
• Please start assignments early. (Warning: cramming does not work!)
Grading policy
– Midterm (20%) – Final (25%)
• Assignments:
– 4 evenly weighted Homeworks (25%) – 2 evenly weighted Projects (20%)
• Quizzes* (9.5%)
• Course Evaluation (0.5%)
Feb 23, 2022, 7pm-9pm ET
for all sections
Email me directly if you have a conflict for the midterm and submit accommodation requests
BOTH ARE DUE BY January 19 at 10pm
April 21, 2022
scheduled by registrar 7pm-9pm ET
Assignments – Homeworks
• There will be 4 homework assignments.
• Goal: strengthen the understanding of the fundamental concepts,
mathematical formulations, algorithms, and the applications.
• Homework #1 will be out Wednesday (01/12) – due Wednesday (01/26) at 10pm ET
– Python tutorial likely next Thursday 01/13 – early submission point
Assignments – Projects
• There will be 2 projects
• Goal: open-ended project designed to strengthen the understanding of the fundamental concepts, while developing some intuition and practical knowledge in applying ML algorithms to real datasets.
• These projects will all include programming assignments to implement algorithms covered in the class and real datasets.
Late Policy
• Solutions to assignments will be posted ~72 hours after the due date.
• It is imperative that you complete your course assignments in a timely manner.
– conceptsareinterdependent,assignmentssolidifyyourunderstanding
• Deadlines are strict!
• For each assignment type (projects or homework) you have a total of three penalty- free late days to be used at your discretion
– usewiselyand(only)forexceptionalcircumstances
– countedlatedaysinincrementsofdaysstartingimmediately
– formoreextremelifeevents,wewillreferyoutostudentsupportservices
• requires documentation e.g., medical documentation that explicitly specifies both a timeline and states that you are unable to engage in your coursework
No late submissions will be accepted after three days have been used up
Submitting Assignments
• (Approximately) (bi-)weekly assignments are due at the dates noted (typically on Wednesdays 10pm ET) submitted on Gradescope.
• Late days
• Solutions released 3 days after due date
– always review!
• Regrades only via gradescope
• (Approximately) weekly quizzes based on lecture and discussion material are due at the dates noted (typically on Sundays 10pm ET). Quizzes will be submitted on Gradescope*.
• Most will be graded for correctness
• Quiz 0 is a bonus quiz
– I want to get to know you and your background – please take it!
– will be worth a (bonus) point only if you need it to meet the threshold
Course Evals
• Your voice matters!
• We appreciate and acknowledge your effort
– Students will receive full credit iff
• submit the final course evaluations and upload a screenshot
of completion
• midterm evaluations are not required it is strongly encouraged.
Next steps
• Familiarizeyourselfwithsyllabusandcanvasforresources • Seecanvasforrecordingsandnotes
• Accessweeklyquizzesongradescope–noteduedates!
• Reachoutifyouhaveanyquestions
• Ifatallpossible,attenddiscussionsin-person • Doalltheweeklyquizzes
• Getstartedearlyonassignments
Today’s Agenda
• Introduction: what is this class about
• Administrative: resources, grading etc.
• What is Machine Learning?
• Supervised Learning
Linear Classification: Intuition predicting the helpfulness of a review
Supervised Learning
review the review
Fa predict
Supervised Learning
Linear Binary Classifier
Supervised Learning – Classification
• Problem: predict whether a review is helpful or unhelpful Binary Classifier
– Features: star rating (1 – 5 stars),
length of review (max length of 200 words) – Labels: helpful/unhelpful
threshold using
10 helpful
Supervised Learning – Classification
• Problem: predict whether a review is helpful (+) or unhelpful (-) • Data:
– Features: star rating (1 – 5 stars),
length of review (max length of 200 words).
– Labels: helpful/unhelpful
Star rating
Review length
review i helpful review z
review I 1122
Supervised Learning – Classification
• Geometrically:
A. Predict +
B. Predict –
Star rating
Review length
star rating
https://docs.google.com/document/d/1JWGiT8gAD4-fNbq8j41xh1A-ruXlS_x7mwnOpkmN7Do/
review length
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com