What Makes a Good Feature? - Machine Learning Recipes #3

[@hejarshahabi114]

holy god, Ive been looking whole KZbin and torrent to find a best package for learning machine learning and I watched many videos, but believe me I am totally in love with your course, this is awesome, you explain it so simple it is like you teach me in my mother tongue. i wish you best like my man.

[@medupidups4741]

Me too this is clear and straight to the point

[@rac]

loving this series!

[@EmettSpeer]

I love your series and wish you would put out videos more often.

[@JBGordon]

+Emett Speer Thanks so much! Episode #4 is live: kzbin.info/www/bejne/bmXKooaYgch1fKc

[@sephirothjc]

I have limited knowledge of programming and computer science, yet I find this series very approachable and fun.

[@philipsalvo]

Josh, Thank you so much to you and your team for building this series! In particular, I really like your 'tl;dr' approach and keeping things grounded in accessible, real-world examples -- I can't wait to see what comes next!

[@mitchese1]

This is a really great series, please publish more often/sooner!

[@Technologynorth]

Really enjoying this series. The examples are quality, and you can tell you have put a lot of thought into them. Thanks for making the subject clear and interesting in each episode.

[@gbhall]

I could honestly watch Josh all day. He presents really well. Keep up the good quality content Josh! :)

[@jamiequigley5906]

Creepy smiles at the end of each sentence, "Smile MORE Josh" Marketing bellows!

[@michaelwhitlock7398]

Now I can't unsee it

[@andreasparasian7225]

Ik right, I mean yea this is very informational but he doesnt have to act like he is inhuman or sth...

[@giantneuralnetwork]

Loved the stacked histogram, nice way to visualize the different means of the distributions!

[@arkrou]

For feature selection: is having no individual prediction power (in your example, eyes) enough to tell whether those features have no value at all? Could they not have some non-linear joint predictive power with some other features?

[@ThereIsNoSpoon678]

Everyone complaining and wanting more episodes. Here I am in 2020 enjoying all 10 of them so far.

[@codingwithjoyk]

Very cool series & I appreciate the links to the examples and especially the "article that inspired". Extra links like that really help! Thank you!

[@sanjayakumarsahoo2293]

Awesome series, prevents myth that machine learning is difficult

[@Abdullah-mg5zl]

*quick summary of the video:* - let's say that your goal is develop a program that can distinguish between two breeds of dogs - what features do you want your example data to have? - you want the features to be the "distinguishing" features between the breeds, i.e. features that are very different between the two dog breeds - for example, if the two dog breeds tend to have very different heights, you want to use height as a feature in your training data - if on the other hand, the two dogs have about the same distribution of eye colors, you don't want to use eye color as a feature - you also don't want to use features that are highly correlated (i.e. that don't bring in new information) - you want to use simple features, as simple features will require less examples to get a decent classifier - you wanna be careful about adding too many features, especially if the features are not "distinguishing" features, they may just by chance be distinguishing in your example data, thus your classifier will start basing its predictions based on these faulty features *key thing to take away from the video:* Selecting features is extremely important. Select the simple, distinguishing features, that bring in new information (i.e. that aren't highly correlated). Thanks so much for these videos!

[@MattSiegel]

terrific episode! those heuristics for feature selection are invaluable. also, lol @ whoever produces the graphics: the frontmost dog *head tilt* XD

[@alxleiva]

WE WANT MORE EPISODES!!!!

[@ErnestGWilsonII]

Agreed

[@alsonyap]

Great job on the video! I can tell that you have taken feedback from previous videos and made adjustments. Thanks for the effort! Will be waiting for the next episode :)

[@luisleal4169]

How do you use categorycal features, for example, we are trying to train a classifier and one of the features is "State"(or maybe "city"), do you create a mapping table where every state(or city) gets a numerical representation? Or would you solve this at programming level: looping through the states(or cities) and finding a classifier for every state(or city) ?

[@net]

Patiently awaiting the next episode in two weeks!

[@EddieImada30]

Very good explanation! Looking forward for new episodes!

[@diegolima2098]

I just want you to know that I loved the article reference in this video. Please refer to more nice articles like this.

[@tkmallik86]

Awesome videos, probably the best ML course by far

[@KelvinKagia]

so awesome i think machine learning is the easiest topic i have ever come across having a good background in programming,probabbility and statistics

[@DerekEskens]

Great series so far. Considering independent features, for something like a dog, would capturing both height and weight be counter-productive since they are most likely interrelated?

[@rjankie]

like the way you are (story)telling about this subject. Makes it accessible for many People. looking forward to "future episodes". ;-)

[@TTarfreak]

This is explained in such a simple and practical way.Loving this series! :)

[@nandishajani]

I tried the same code but the graph is looking very ugly. There are no spaces between the bars. Can anyone please help?

[@amomasi9909]

This was incredibly useful, Josh. Thanks.

[@HubertRozmarynowski]

How to get that pretty data visualisation from matplotlib like in 2:05 ?

[@mrtnsnp]

I enjoy the series. Is it possible to provide the source code (and possibly data if not included in sklearn) so we can follow along and experiment while watching?

[@JBGordon]

+Maarten Sneep Good idea! I've been meaning to, but haven't found cycles yet. In the mean time, I'm trying to keep the code nice and short (~20 lines or less).

[@asiddiqi123]

Is it possible that we don't know about what features lie in Data and we do some processing and find features needed?

[@ameynaik2743]

Wasn't this episode a bit inconclusive? Does anyone know the next episode in this series which discusses about the features?

[@dzxxbj]

These videos are very helpful. Through them I know how to code and understand happily. How can I pass the gap between Little Code and real world problem?Hope for topics about this：）

[@deevioo]

What about ratio between height of a dog and the width of its head?

[@dipalimalviya6921]

hey Josh! I wanted to learn concepts and standard algorithms of machine learning ,please suggest me how i can do this?It will be so helpful. I also want more tutorial of this series.Thanks for this 9 episodes series

[@sukantakr1269]

Dipali Malviya you can watch prof Andrew's ml tutorial! very very helpful to learn algos.. I have also been some ppts made by our one ml research and developer prof!

[@brahmlife]

can you have sub features of features in your decision tree algorithm?

[@xsyed]

Waiting for more of this kind of stuff!!

[@dannyhunn6025]

Couldn't you use Latitude and Longitude to find Euclidean distance?

[@abdulrahmanalotaibiq8]

keep up the good work. Would you talk about image classification but not the simple one that's in the tutorial, please?

[@JBGordon]

+Abdulrahman Alotaibi Definitely planning on that down the road when we get to TensorFlow. Their tutorials are great btw! www.tensorflow.org/versions/r0.8/tutorials/index.html

[@abdulrahmanalotaibiq8]

Awesome and thanks for the reply. I really wanted to know how to preprocess the images before feeding them to the classifier. Every tutorial out there uses the classic MINST and I wish that use guys use something different. Thanks again and you are doing a great job.

[@giantneuralnetwork]

+Abdulrahman Alotaibi I'd checkout open-cv for image preprocessing. Great idea to distort/translate/skew/scale the training images to get a better general predictor!

[@sadiqsariq8037]

what is the difference between np.random.random() and np.random.randn() ?

[@diwashshrestha4996]

It great easy to understand series .we want more.

[@MattieCooper10000]

Exactly what I needed! Thank you so Much! Love your presentation!

[@fatihturgel]

Great work! Really love your series! :)

[@RM-lw2qi]

What does the greyhound = 500 line actually do?

[@shoun4148]

Juts makes the variable greyhound equals to a value of integer 500

[@dzxxbj]

Are there viedos like this tutorial's type?

[@husainzafar1648]

I think np.random.randn is not restricted in range (-1,1). So the error wouldn't be just +/- 4%. Btw excellent videos and loving your style of explaining!!

[@MrHuno92]

Is there a good read available to elaborate on the topic or follow the tutorial?

[@JBGordon]

+60SecondsTech Good question. I haven't done a write up (maybe will do in the future, that's a good idea). For an in depth treatment, Prof Andrew Ng’s course is an incredible resource: kzbin.info/www/bejne/i6vbip-YgJeWbac

[@mycount64]

simplified classifiers OK. what if feature ratio's are significant or their are sorting stations for the letters with varied release times for the letters... eh just confusing things. will the ML program figure out relationships between features if it is supervised or is it our job to figure out the relationships. Actually isn't the problem that if we have a system with 2500 features and do not know how they are significant then ml will figure it out?

[@AviPars]

mine doesn't overlap , is that normal?

[@theburntcrumpet8371]

I'm a bit late to the party here but I'd just like to say thanks to the Google Developer's channel for putting these videos out there.

[@Hieuiph]

thanks !!!! I hope you will explain in the future the "feature engineer" technique.

[@rajnishrajput6121]

So if i want to make a program that identify all the dogs in the world, I have to store all the data of all dogs in the world? like height, weight, speed hair etc.

[@Warri0r1992]

Hey Josh! I'm studying machine learning in my University since a couple of years, and i've found your tutorials very useful, altough i already know those things! The concepts are very clearly teached! Python looks really easy to understand, instead of R, for istance. Also, i'm looking for blogs or web sites to keep myself always updated on ML topics....do you have some advices? Keep it up, i can't wait for another video!

[@JBGordon]

+Warri0r1992 Thanks! My favorite blog is colah.github.io - it's a pretty incredible explanation of how neural networks work, and some assorted other topics.

[@TheIsrraaa]

Analytics Vydhia at fb or web

[@harendrasingh_22]

This is an awesome series ! The best thing ever in ML :P (Well not "the best" , but yeah ! ) !

[@DragoonGalaxy7]

what's the program you use to write the python code?

[@titas3419]

Sublime Text

[@stickmouse5002]

how does training work, my program marks an apple as an orange

[@newcoolvid27]

If you're using spyder and want a new window to show the plot, [Tools > Preferences > IPython console > Graphics > Graphics backend > Backend: Automatic] then restart Spyder.

[@pitbbe]

Awesome Series! please make more videos!

[@bimDe2024]

Complex in easiest words, thanks a lot

[@MuammarElKhatib]

Very clear explanation.

[@HighNorthStudios]

These vids are great! Thank you and keep it up!

[@andrescolon]

Good job on these videos!

[@Cupofshake]

The graph doesn't work, only gets this message.

[@bhavykhatri2669]

His smile is so motivating.

[@tekki.dev.]

im trying to learn two things at once here, python and machine learning, but i guess its not too hard as i already know c#,php etc... ML is also not very hard at first but gets little complicated as you go deep...

[@gaurav9839]

What if I don't what is dog and I need to identity it ... My program is just a toddler and it is learning from Internet ...

[@SirajRaval]

Hell yeah! If you guys like machine learning check out my new ML series on my channel.

[@user-ny8xt1ui3h]

Wonderful Series of ML! Someone recommends me another one? please

[@JeffWeakley]

Big Fan. thanks Josh. Much appreciated.

[@vladvlog8261]

Thanks for the video. Really appreciate it!

[@JeromeEtienne]

So clear! Thanks

[@twmicrosheep]

Keep up the great work！

[@akansha.da1iiitmk.ac.inaka290]

Doing it on Python 3? Don't want to pause the video and write? Find the code here: github.com/akanshajainn/Machine-Learning---Google-Developers

[@jinxblaze]

Wouldn't Latitude and Longitude give you accurate distances by some very simple calculations ?

[@riverland0072]

yes, but who has time for simple calculations when we have simpler distances...now you just made me try to remember log and lat maths from high school..it was not that simple

[@egor.okhterov]

Amazing. I understand everything perfectly :)

[@ErnestGWilsonII]

# Import Numpy import numpy as np # Import matplotlib import matplotlib.pyplot as plt # 500 of each dog greyhounds = 500 labs = 500 # Set the dog heights +/- 4" randomly grey_height = 28 + 4 * np.random.randn(greyhounds) lab_height = 24 + 4 * np.random.randn(labs) # Plot plt.hist([grey_height, lab_height], stacked=True, color=['r', 'b']) # Launch the results in a window plt.show()

[@dolomikal]

A bit light on the info here I think. Compared to the last episode.

[@amomasi9909]

#Hi guys. Code that works now so you can follow: import numpy as np import matplotlib.pyplot as plt greyhounds = 500 labs = 500 grey_height = 28 + 4 * np.random.randn(greyhounds) lab_height = 24 + 4 * np.random.randn(labs) plt.hist([grey_height, lab_height], stacked=True, color=['r', 'b']) plt.show

[@Unhacker]

I loved it, but my dog would prefer a squirrel detection algorithm.

[@fpspsyduck]

great video!

[@vivasjimmy]

we want more of these videos

[@TheTruthFadeswithTime28]

why does my bar graph look way worse aesthetically than yours?

[@andrecastro1179]

mine too, probably because of his magic Mac.

[@MamunSuper]

awesome lectures...)

[@avinashravi4606]

where i can Learn deeper about ML algorithm with statistics

[@SergioArroyoSailing]

Awesome 😀😀!!!

[@akarshrastogi3682]

The decision tree that he follows to decide when and how often he SMILES creeps me out..! 1:06

[@ElVerdaderoAbejorro]

I have created a github repo with all of the code for all of the recipes of this series. I've used Python3 for all recipes. I've also updated all of the libraries and have added some things to the code here and there. Check it out: github.com/TheCoinTosser/MachineLearningGoogleSeries

[@State_exam_preparation]

Awesome👏✊👍

[@husker3345]

Awesome, thanks :)

[@koushik7604]

Great series for machine learning.........only thing, please upload videos within a week.

[@denisedias7796]

Thank you

[@dude2260]

when the max height is 28 + 4 where does 35 comes from

[@jeremyheminger6882]

Yeah, I agree, its sketchy. I fed it print(10 + 4 * np.random.randn(10) + 4) to generate numbers only above 10 and less 18 but I occasionally am getting numbers like 9.6... and 18.5... [ 9.65793498 13.18266908 8.06937322 10.59406851 11.29406986 11.14891777 11.93107746 13.25606161 18.58717831 8.54857054] How does this function work?

[@JackalovichMokov]

has anyone else noticed that he never blinks

[@editmergedx6541]

5:00

[@sapnaroy6373]

At 2:16

[@breddowen1801]

Thank you a lot!!!!!!!

[@Dynamite_mohit]

Thankyou

[@harshseth3231]

Nice Tshirt ! :P

[@ReverseMe]

better remove inches

[@minigam3s479]

I wish i could look so happy ^^

[@TheVipul845]

Hi there, this is a very cool example for beginners. however, I am unable to get the histogram printed correctly, here's the image link: imgur.com/RGbzrxH I am running pycharm on ubuntu 20.04