Machine Learning Foundations: A Case Study Approach

University of Washington via Coursera

Go to class Write review

Details

Go to class

Provider

Coursera
Pricing

Free Online Course (Audit)
Languages

English
Certificate

Paid Certificate Available
Duration & workload

18 hours 9 minutes
Sessions

On-Demand
Subtitles

Arabic, French, Portuguese, Chinese, Italian, German, Russian, English, Spanish, Korean, Thai, Indonesian, Kazakh, Hindi, Swedish, Greek, Ukrainian, Japanese, Polish, Dutch, Turkish, Hungarian, Bengali, Pashto, Urdu, Azerbaijani, Farsi

A Best Course of All Time

Found in

Part of

Machine Learning

1.0

Overview

Do you have data and wonder what it can tell you? Do you need a deeper understanding of the core ways in which machine learning can improve your business? Do you want to be able to converse with specialists about anything from regression and classification to deep learning and recommender systems? In this course, you will get hands-on experience with machine learning from a series of practical case-studies. At the end of the first course you will have studied how to predict house prices based on house-level features, analyze sentiment from user reviews, retrieve documents of interest, recommend products, and search for images. Through hands-on practice with these use cases, you will be able to apply machine learning methods in a wide range of domains. This first course treats the machine learning method as a black box. Using this abstraction, you will focus on understanding tasks of interest, matching these tasks to machine learning tools, and assessing the quality of the output. In subsequent courses, you will delve into the components of this black box by examining models and algorithms. Together, these pieces form the machine learning pipeline, which you will use in developing intelligent applications. Learning Outcomes: By the end of this course, you will be able to: -Identify potential applications of machine learning in practice. -Describe the core differences in analyses enabled by regression, classification, and clustering. -Select the appropriate machine learning task for a potential application. -Apply regression, classification, clustering, retrieval, recommender systems, and deep learning. -Represent your data as features to serve as input to machine learning models. -Assess the model quality in terms of relevant error metrics for each task. -Utilize a dataset to fit a model to analyze new data. -Build an end-to-end application that uses machine learning at its core. -Implement these techniques in Python.

Syllabus

Welcome

Machine learning is everywhere, but is often operating behind the scenes.
This introduction to the specialization provides you with insights into the power of machine learning, and the multitude of intelligent applications you personally will be able to develop and deploy upon completion.
We also discuss who we are, how we got here, and our view of the future of intelligent applications.

Regression: Predicting House Prices

This week you will build your first intelligent application that makes predictions from data.
We will explore this idea within the context of our first case study, predicting house prices, where you will create models that predict a continuous value (price) from input features (square footage, number of bedrooms and bathrooms,...).
This is just one of the many places where regression can be applied.Other applications range from predicting health outcomes in medicine, stock prices in finance, and power usage in high-performance computing, to analyzing which regulators are important for gene expression.
You will also examine how to analyze the performance of your predictive model and implement regression in practice using a Jupyter notebook.

Classification: Analyzing Sentiment

How do you guess whether a person felt positively or negatively about an experience, just from a short review they wrote?
In our second case study, analyzing sentiment, you will create models that predict a class (positive/negative sentiment) from input features (text of the reviews, user profile information,...).This task is an example of classification, one of the most widely used areas of machine learning, with a broad array of applications, including ad targeting, spam detection, medical diagnosis and image classification.
You will analyze the accuracy of your classifier, implement an actual classifier in a Jupyter notebook, and take a first stab at a core piece of the intelligent application you will build and deploy in your capstone.

Clustering and Similarity: Retrieving Documents

A reader is interested in a specific news article and you want to find a similar articles to recommend. What is the right notion of similarity? How do I automatically search over documents to find the one that is most similar? How do I quantitatively represent the documents in the first place?
In this third case study, retrieving documents, you will examine various document representations and an algorithm to retrieve the most similar subset. You will also consider structured representations of the documents that automatically group articles by similarity (e.g., document topic).
You will actually build an intelligent document retrieval system for Wikipedia entries in an Jupyter notebook.

Recommending Products

Ever wonder how Amazon forms its personalized product recommendations? How Netflix suggests movies to watch? How Pandora selects the next song to stream? How Facebook or LinkedIn finds people you might connect with? Underlying all of these technologies for personalized content is something called collaborative filtering.
You will learn how to build such a recommender system using a variety of techniques, and explore their tradeoffs.
One method we examine is matrix factorization, which learns features of users and products to form recommendations. In a Jupyter notebook, you will use these techniques to build a real song recommender system.

Deep Learning: Searching for Images

You’ve probably heard that Deep Learning is making news across the world as one of the most promising techniques in machine learning. Every industry is dedicating resources to unlock the deep learning potential, including for tasks such as image tagging, object recognition, speech recognition, and text analysis.
In our final case study, searching for images, you will learn how layers of neural networks provide very descriptive (non-linear) features that provide impressive performance in image classification and retrieval tasks. You will then construct deep features, a transfer learning technique that allows you to use deep learning very easily, even when you have little data to train the model.
Using iPhython notebooks, you will build an image classifier and an intelligent image retrieval system with deep learning.

Closing Remarks

In the conclusion of the course, we will describe the final stage in turning our machine learning tools into a service: deployment.
We will also discuss some open challenges that the field of machine learning still faces, and where we think machine learning is heading. We conclude with an overview of what's in store for you in the rest of the specialization, and the amazing intelligent applications that are ahead for us as we evolve machine learning.

Taught by

Carlos Guestrin and Emily Fox

Reviews

4.0 rating, based on 40 Class Central reviews

4.6 rating at Coursera based on 13489 ratings

Start your review of Machine Learning Foundations: A Case Study Approach

Gregory J Hamel ( Life Is Study) @greg

Machine Learning Foundations: A Case Study Approach is a 6-week introductory machine learning course offered by the University of Washington on Coursera. It is the first course in a 5-part Machine Learning specialization. The course provides a broad…

Machine Learning Foundations: A Case Study Approach is a 6-week introductory machine learning course offered by the University of Washington on Coursera. It is the first course in a 5-part Machine Learning specialization. The course provides a broad overview of key areas in machine learning, including regression, classification, clustering , recommender systems and deep learning, using short programming case studies as examples. The course assumes basic Python programming skills and it uses a software package called GraphLab that requires a 64-bit operating system running Python 2.7. Grades are based on periodic comprehension quizzes and short programming assignments.

The course covers a broad range of machine learning topics at a high level with the promise of drilling down into the details in future courses in the specialization. The lecturers have good chemistry, but they tend to get distracted when they are on screen together. The video and slide quality are very good and although the delivery is a little rough around the edges at times, the lectures are informative. The machine learning methods covered aren’t necessarily treated as complete black boxes, but the course intentionally avoids getting too deep into the details, putting the emphasis on conceptual understanding.

The weekly labs are contained in short IPython Notebooks—interactive text and code documents rendered in a web browser—that illustrate some simple models in GraphLab. The labs themselves are easy and don’t require much coding other than calling various built in GraphLab functions. The hardest part about the class is getting your programming environment set up in the first place. If you don’t have a new version of 64-bit Python 2.7, you can’t run GraphLab. It is relatively easy to get set up if you can use the recommended Anaconda Python distribution, but getting things set up manually on an existing Python installation may prove troublesome. The instructors provided some workarounds for doing the course without GraphLab or using GraphLab on Amazon’s cloud computing service; I wouldn’t take the course without getting GraphLab working in some form. Many students decried the use of a non-open source package for an open class; I think it is useful to be exposed to new tools and GraphLab seems cleaner than Python’s popular scikit-learn package. In this sort of course, the focus should be one concepts rather than syntax.

Machine Learning Foundations: A Case Study Approach achieves its goal of introducing machine learning at a high level without rushing or trying to cram too much into any particular week. What the professors lack in terms of polish they make up for with enthusiasm. Compatibility and setup issues will be a roadblock for some, but overcoming them is worth it.

I give Machine Learning Foundations: A Case Study Approach 4.5 out of 5 stars: Great.
Robert Stahr

I had already completed Andrew Ng's Machine Learning course (Coursera/Stanford), and a couple of courses in the Data Science specialization (Coursera/Johns Hopkins). Although I loved Andrew Ng's course, I was looking for something more in-depth and…

I had already completed Andrew Ng's Machine Learning course (Coursera/Stanford), and a couple of courses in the Data Science specialization (Coursera/Johns Hopkins). Although I loved Andrew Ng's course, I was looking for something more in-depth and a little more useful in my daily work than Octave or R, which are the languages used in these other Coursera courses. So when I saw this University of Washington specialization and read that they use Python, I was very excited.

I was a little less excited when I saw that the courses were not free like the other Coursera courses I had previously taken - but hey, it costs money to produce a course, so paying a little seems reasonable.

However, I soon realized that you don't actually program anything in python, you just slightly modify pre-chewed bits of code. And you don't really do much that is useful in a real job situation because everything is done via very high-level calls to a proprietary (and VERY expensive) python library called GraphLab. They "kindly" give you a one-year free license to the thing, but when you discover that one of the instructors is the co-founder of the company that produce and sell GraphLab, you realize that this is not really a course, it's an info-commercial that you had to pay for.

Don't waste your time or money on this course. Try Andrew Ng's course (on Coursera) or Sebastian Thrun's courses (on Udacity). They are much better and they are free.
Ericdo1810

This course is easily the best introductory course to Machine Learning one can get. Well-designed, beginner-friendly but also rich in content at the same time. There's a pretty nice balance between theory and practice. Basic, foundational machine le…

This course is easily the best introductory course to Machine Learning one can get. Well-designed, beginner-friendly but also rich in content at the same time. There's a pretty nice balance between theory and practice. Basic, foundational machine learning techniques are taught via GraphLab, a Python-based software specialized in analytics. The course instructor is also one of the founders of this great software.

If you are looking for some courses that are more difficult and challenging than Andrew Ng's Machine Learning course, this course is not for you.

We should understand that, this course is designed to hold learners' hands and walk them through understanding the very basic foundation of machine learning. With that objective in mind, this course did really well in familiarizing a complete beginner with all the terminologies and techniques in Machine learning: regression, classification, clustering, etc.

For people who look for complex algorithms and deep understanding of techniques, this course is not meant for them. However, as the instructors promise, the next courses in the specialization will delve deeper into those complex topics. Hence, there's a great deal of excitement to look into this specialization.

So, the theory part has been really beginner-friendly. For the practical, programming part, it's also very digestible. Remember, the target of this course is beginners, so if you're some expert Python user looking to learn some magics. you will be disappointed as there is not much advanced Python techniques here. However, there's some Python basic requirement, as the software they use is based on Python and you will need to know some basic data structures and iterators in Python to have it easy.

About the software Graphlab, some people complain that the course is trying to sell GraphLab. It is not doing so, and it is unfair to label the course as having such a perverse intention. The course instructor made clear that GraphLab is available for free, indefinitely, for academic purpose. Commercial use is forbidden by the license, which is obvious, and reasonable.

Looking forward to the next courses in this specialization, where the instructors promise that the concepts will be more challenging and complex but that also comes with the possibility of building exciting machine learning applications.
Gary Hess

EDIT: After finishing 4 courses in the specialization, my opinion has gone downhill. Here is my review of the whole specialization: https://www.class-central.com/certificate/machine-learning-specialization#reviews :END EDIT I completed this course…

EDIT: After finishing 4 courses in the specialization, my opinion has gone downhill. Here is my review of the whole specialization: https://www.class-central.com/certificate/machine-learning-specialization#reviews :END EDIT

I completed this course in mid-February. After the Andrew Ng course from Stanford/Coursera, it was relatively easy but I was happy to get a simple Python-based introduction to machine learning since Andrew Ng's course uses Matlab/Octave.

Compared to the big (free) MOOCs, it is a little disappointing to see so little activity in the forums. The approx. $75 price tag keeps enrollment down and I didn't manage to have a lot of interesting communication in the forums (compared to the bigger MOOCs).

I installed Graphlab on my machine and would NOT recommend trying this course unless you plan to use Graphlab. Seems like the folks who are trying to use scikit-learn for the course are having trouble getting the right answers.

Overall, I don't regret taking this course but remember that it is part of a 6-course series and the series is not yet complete. At this time there is also the second course in the series (Regression, which I just finished today and found to be somewhat more in-depth). The third course in the series is launching this week (Classification). I plan to take it but I will take a month break first and wait for the initial bugs to be worked out.

BTW: University of Washington does now appear on the Coursera certificate. I think there was some doubt about this earlier, but here is my certificate (on closer inspection, I noticed the name of my professor is "John Doe"...): https://www.coursera.org/account/accomplishments/verify/Z6ZXWN7D8US4
Igor Filippov

This is an introductory course, so don't expect any in-depth explanation. What it teaches you is "you can take this data and have this prediction", but course doesn't explain math behind the solutions.
The most frustrating for me is that staff isn't responsive on the forums. Questions can stay answered for weeks. Moreover, "Course by University of Washington" isn't quite right, since they "decided not to put UW logo on certificate".
Anonymous

This is supposed to be an introductory course to machine learning. It is not quite introductory in the sense of a gentle start from basics, but is geared towards providing some kind of introduction or overview of regression, classification, clusteri…

This is supposed to be an introductory course to machine learning. It is not quite introductory in the sense of a gentle start from basics, but is geared towards providing some kind of introduction or overview of regression, classification, clustering, recommendation systems and deep learning.

Half of the course consists of lectures on theory, and the other half consists of lecture-labs in Python using Graphlab. I found 4/5 of the lectures fine, sometimes with a bit of self-research added on - the two instructors, Emily and Carlos, were engaging and had good chemistry between them. The remaining 1/5 I did not understand despite trying to do self-research on the topic or issue.

In some cases the quiz questions and lab-assignment instructions were utterly poorly written. Some were downright confusing. Some just did not make sense. Some lab-assignments required students to make use of Graphlab commands not taught during the lecture-labs. Some quiz questions covered topics not at all covered during the lectures. There was a lot of frustration with these issues, some expressed as early as 7 months ago, but the material does not appear to have been updated to resolve them.

Unfortunately, despite having 2 instructors, 3 teaching staff and 29 mentors, most students who were facing technical issues with the datasets or Graphlab commands, or who did not understand the lectures or quiz questions, received no help from the instructions TS and mentors. There was not very much participation from them, and I never one post from the instructors and TS.

I have checked out Andrew Ng's and Udacity's ML courses, and preferred UW's out of the three. However the course materials need to be re-examined and improved on.
Anonymous

A lot of people is attacking the Course as a high-level, not deep. I must say that at the very beginning I thought the same way. However in the second course, they force you to develop your own routines in python. So there is no need to pay a license or anything in the real world.

I didn´t know about UW declining to sign the certificates, I wish we all could know the reason.

Keep studying, this course will take you far!
Anton Poznyakovskiy

This is an introductory course in the specialization, and a such, aims more for the breadth than for the depth. It gives a good overview of the areas of machine learning and motivates and explains them with the case studies mentioned in the title. I can't say that the case study approach is different from other data science courses that I have participated in, but the lecturers present the concepts of machine learning in a clearly explained and memorable way. The only thing that I disliked about the course (and the reason why I rate it only 4 out of 5) is that the programming assignments amount to modifying already existing code. This makes them much too easy in my opinion, and also reduces their learning outcome.
Tim Haines

This course is pretty basic, offering an overview, and some workshops to work through. I gave this course a low 2 star rating because the workshops use software called Graphlab. Graphlab is actually pretty cool, and is made by one of the course pr…

This course is pretty basic, offering an overview, and some workshops to work through.

I gave this course a low 2 star rating because the workshops use software called Graphlab. Graphlab is actually pretty cool, and is made by one of the course presenter's startups. However, Graphlab is not available for commercial use. Before Apple acquired Turi, the company that makes Graphlab, I've heard it was available at a special price of $400/month, but even that is no longer available.

I started and got a long way through this course with the intent of learning software that I can use in my daily work, only to find out that would be impossible. Frustrating waste of time.

(I'd already done Andrew Ng's Stanford course previously - which was great for learning the theory)
Ala Falaki

This course consist of 6 week, each week has two part.
Part 1: They discus about problem and algorithms we can use to solve the problem.
Part 2: After explaining ways that is possible to solve the problem, They try to implement the algorithm using GraphLab software.
sometimes in course you just feel that it is a GraphLab workshop ( Carlos, one of the instructors, is founder of Dato-GraphLab company ) but i don't think that it is a problem!
Altogether i think Carlos and Emily put too much effort for this course, and if you excited about Machine Learning, definitely you will enjoy this course.
Daniel Rosquete

It is a great course for beginneers in Machine Learning, to know what you could do.

Many people may say that you won´t do your own algorithms, and that you will always require a commercial license, but that is not true. Even when in this course it is absolutely true, the following courses are not like that.

It is awesome and the professors are great!
Jason Michael Cherry

For what it is, this class does a good job overviewing different analytical techniques in machine learning. It's light on the details (that's what the follow-up courses are for), but gives you a flavor on how this stuff works.
Amilkar Herrera

Pros: Well explained lessons and the case study approach is good to help you understand in what situations you might apply what you are learning. It's not too expensive and it focuses on understanding the concepts and not in the programming language.

Cons: It does not use native python, but a proprietary library, so the code you develop during the course cannot be used latter as you would need to buy a licence. The course is an introduction to the rest of the specialization so if you are not planning to take the whole specialization, this course will serve as a very general introduction to the topic.
Anonymous

Spent a lot of time learning how to use Graphlab. They start saying the benefits of using Python as it is open source and then we have to spend significant amount of time learning Graphlab, for which you would have to pay money for non-academic purposes. Seriously considering dropping the course because of this.
Solicitor Charlie Lee Scott Esq

Research such as Library of Congress areas Psychology

where conclusive society

Entail social aspect that a

matter within all matters of

Intervene interests, Our new

social society of security for

life that matter to man.
Suresh

Gives good basic foundation on Machine Learning for people who have absolutely no idea on what's ML is. Anyways got to complete the whole specialisation to try your own models.
Aissata Thiongane

the aim of this course is to learn artificial learning and the modeling of classification systems but also to study and ultimately be able to design prediction algorithms
Mikael
Dhawal Shah @dhawal
[email protected]