Machine Learning: Regression に戻る

4.8

星

4,554件の評価

•

853件のレビュー

Case Study - Predicting Housing Prices
In our first case study, predicting house prices, 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.
In this course, you will explore regularized linear regression models for the task of prediction and feature selection. You will be able to handle very large sets of features and select between models of various complexity. You will also analyze the impact of aspects of your data -- such as outliers -- on your selected models and predictions. To fit these models, you will implement optimization algorithms that scale to large datasets.
Learning Outcomes: By the end of this course, you will be able to:
-Describe the input and output of a regression model.
-Compare and contrast bias and variance when modeling data.
-Estimate model parameters using optimization algorithms.
-Tune parameters with cross validation.
-Analyze the performance of the model.
-Describe the notion of sparsity and how LASSO leads to sparse solutions.
-Deploy methods to select between models.
-Exploit the model to form predictions.
-Build a regression model to predict prices using a housing dataset.
-Implement these techniques in Python....

Mar 17, 2016

I really enjoyed all the concepts and implementations I did along this course....except during the Lasso module. I found this module harder than the others but very interesting as well. Great course!

Jan 27, 2016

I really like the top-down approach of this specialization. The iPython code assignments are very well structured. They are presented in a step-by-step manner while still being challenging and fun!

フィルター：

by Saiprasad B

•Jan 07, 2017

very interesting environment to learn the subject.

by Konstantin B

•Nov 18, 2017

Too much math...

by James F

•Nov 23, 2016

I'm a statistician, not a programmer. There is so much detail and explanation about the statistics and concepts behind how it works, but there is hardly ever an actual lesson on the code used or needed to implement the algorithms. When trying to fumble my way through the code, I found, on several occasions, the code in the self-directed lessons to be incomplete (I'm referring to pieces of code that were obviously meant to be there, but were missing), causing hours and hours of anguish and turmoil. I feel like there should be a lot more time spent on the actual coding and learning how to implement it within the code (similar to the 1st course), rather than spending an exuberant amount of time going through derivations and no time on actual coding and how to implement it within the programming language.

If you are a software designer/engineer or programmer, then you should be fine as long as you pay attention to the very long lessons and derivations and can fix the broken code that you are given. There are other mistakes within the quizzes as well, which make them near impossible to pass. For example, it is unclear which model you need to use to calculate in order to get the correct square foot. On other occasions, the question actually specifies to use the model from (3), whereas it actually wants you to use the model from (4) instead to get the correct answer. This course needs to have better quality checks to ensure needlessly lost time is minimized.

by Peter H

•Aug 28, 2016

There are two frustrations with this iteration of the series. One: the quiz questions are often opaquely worded. Instead of being tested on the material just learned, it seemed like the objective was to learn to decode test questions.

2 and the most glaring omission, was that when students are asked to provide functions, only some are provided with a follow up test to ensure the function is working properly. If the output is syntactically correct but provides incorrect output then you're moving forward blindly after that. Then add the quiz questions from problem One above, and you're just wasting your time after that and building up frustration. Validating your code as you move along seems like a pretty rudimentary process to impart to students and when the teachers don't practice it themselves, there are bound to be problems.

I like the intent of the course, and considering my outsider background to computer science, the mathematics etc, I did learn a fair bit. Not enough to justify the increasing frustration I was feeling toward the end of this course. I have no intention of taking any more at this point, not from these authors.

by Matthew K

•Jul 26, 2019

The course is well structured and organized; however, there is too much focus on the complex mathematical formulas and notation. The concepts are not terribly advanced but the math involvement makes it easy to get lost. The math is obviously necessary, but I just wished the lecturer had spent more time on the concepts than trying to explain what each of the subscripts, superscripts, and greek letters meant. There were many 7 minute lectures in which 5-6 minutes would be confusing math and 1-2 minutes would be actual conceptual talk. I was able to understand what was going on, but I felt it would have stuck much better if more time was spent discussing and reiterating the concepts. The math involvement could come from the coding assignments.

by Amol N

•Feb 18, 2016

Pace is extremely slow. The instructor writes and talks simultaneously. The words are put so slow that it puts me off too sleep at times. I love taking courses where the instructor speaks at the right pace and keeps you involved. Carlos, the co instructor. One of the perfect MOOC is Calculus One by Jim Fowler

by Wayne P

•Jan 09, 2019

Great concepts but material presented is very theoretical with minimal practical examples. As such it is easy to get lost unless you have advanced mathematics skills.

by Mesum R H

•Dec 09, 2017

Too Statistical depth. Could have explained in a more exampled manner rather than deriving a maths equation class. We are not Phd Maths & Statistics

by Eric Z

•Jul 05, 2018

The material is not very clear and I have to keep going through it and seek clarification from other resources.

by ashish s g

•Feb 15, 2017

Very good course material. However, Graphlab is no longer free to use for commercial purpose.

by Ignacio A d l T P

•Feb 27, 2018

PLEASE REVIEW EVERY QUIZ, in several of them I had to input a different answer from what I thought was the correct answer after VERY carefully following instructions, reading and re-reading, executing, looking for alternatives, incorrectly graded quiz answers significantly have slowed me and tested my willingness to continue. If the quizzes need to grow to 14-20 questions so that the exercises become more "step by step" that would be OK, since the whole purpose of taking this for someone with 10-12 years of professional experience is to become confident that I have understood the concepts, when I have to guess responses my confidence on my understanding of the concepts gets strongly tested. I chose your specialization because it is project oriented, has use cases and breaks down every course into very detailed concepts, it is awesome to have been able to deepen my understanding of regression through this course but it could have taken me a fourth of the time and have been an achievement and something fun to work on if the quizzes were correct versus a chore and a source of stress.

If you need further information please reach me at

by Omar A C T

•May 30, 2016

this was a really boring course not for the contet bu the teacher i fell bored every video because the theacher was really slow in everything tha she was showing, it is realy dificult to get focussed in the real topics when the teacher spend a lot of time explaining things at the end wont be evaluated. As an example I am not english native speaker but a had to put the playback speed to 1.50x in order to not get bored in all videos, it was really dificult to follow the teacher at the normal velocity , i just got sleep every video. and as a record i really like this topic so it is the tacher, I took the first course and it was a good experience but this one is owfull

by adam h

•Mar 09, 2016

gets way too in-depth with the math behind regression, to the point that it deters from the learning process. was hoping to learn better methods of interpreting or enacting regression, not the inner workings of the algorithms.

assignments got overly complex with confusing instructions. there are definitely some leaps made in the assumptions of what students' python capabilities are. vague instructions caused more frustration than desire to continue learning.

will continue in the specialization, but will not hesitate to drop out if instruction continues like this.

very disappointed.

by Monika K

•May 03, 2016

I've spent a bit of time going through the Specialisation (paid for one course here) and other courses online that offer Machine Learning with Python. I looked at books too. I've come to the conclusion that it's unforgivable to teach it using graphlab (that you have to pay for after free licence expiry) when everyone else teaches scikit learn (sklearn) for good reason.The tools used on this course are also not very good.

Everyone else teaches using text editors - for a good reason, you learn how to code properly.

The lessons are also dry and there are far too many of them.

by Eugene K

•Feb 10, 2017

If you are considering this specialization I would recommend the Andrew Ng course instead and the main reason is that it isn't depend on proprietary ML framework. Despite the good lectures, the assignments don't help you develop the knowledge required for ML developer role.

Taking in consideration the permanent postponing the courses delivery, from summer 2016 to summer 2017, finally the most interesting part of the specialization was cancelled. I'm completely disappointed with the specialization learning expirience.

by William S

•May 03, 2016

This course is structured around a specific and costly Python library called Dato. It is possible to do the homework without it, but it is EXTREMELY difficult to do so. If the course wasn't structured around using Dato, it would be a lot simpler and a easier to complete the assignments. Also, a lot of the mathematical notation was written in a kind of psuedo Python code that made things confusing sometimes.

by Mats W

•Dec 17, 2016

The lecturers try to keep the instructions basic and pedagogical. Pretty good. Everything in this revolves around a tool graphlab create. Not so great, I think. It is not free (you get a one year licence) and hides all the action from the user. I don't like that the course then makes me feel that I must rely on a specific product to solve problems.

by Konstantin K

•Jun 19, 2016

I was not aible to complete this course for free. That was very disappointing! Universities like Stanford and John Hopkins find the opportunity to offer similar courses free of charge to peoople who want to learn. From University of Washington I have expected the same. Your bad!

Best regards

Konstantin

by Ehsan M

•Mar 11, 2018

The teachers have a great success in developing Tori, but, the teaching is not good. The way machine learning is presented is mixed, and all over the place.

Not worth to put time on

by Andreas

•Jan 04, 2017

This specialization is delayed for months now - very annoying! Don't give them money!

by Adrien L

•Feb 02, 2017

No good without the missing course and capstone projects

by Ken C

•Feb 04, 2017

Not happy about course 5 & 6 got cancelled.