Learn TensorFlow and Deep Learning fundamentals with Python (code-first introduction) Part 2/2

You’ve made it to part 2 of the longest code-first learn TensorFlow and deep learning fundamentals video series on KZbin!
This part continues right where part one left off so get that Google Colab window open and get ready to write plenty more TensorFlow code.
Sign up for the full course - dbourke.link/ZTMTFcourse
Get all of the code/materials on GitHub - www.github.com/mrdbourke/tens...
Ask a question - github.com/mrdbourke/tensorfl...
See part 1 - • Learn TensorFlow and D...
TensorFlow Python documentation - www.tensorflow.org/api_docs/p...
Connect elsewhere:
Web - www.mrdbourke.com
Livestreams on Twitch - / mrdbourke
Get email updates on my work - www.mrdbourke.com/newsletter
Timestamps:
0:00 - Intro/hello/have you watched part 1? If not, you should
0:55 - 66. Non-linearity part 1 (straight lines and non-straight lines)
10:33 - 67. Non-linearity part 2 (building our first neural network with a non-linear activation function)
16:21 - 68. Non-linearity part 3 (upgrading our non-linear model with more layers)
26:40 - 69. Non-linearity part 4 (modelling our non-linear data)
35:18 - 70. Non-linearity part 5 (reproducing our non-linear functions from scratch)
49:45 - 71. Getting great results in less time by tweaking the learning rate
1:04:32 - 72. Using the history object to plot a model’s loss curves
1:10:43 - 73. Using callbacks to find a model’s ideal learning rate
1:28:16 - 74. Training and evaluating a model with an ideal learning rate
1:37:37 - [Keynote] 75. Introducing more classification methods
1:43:41 - 76. Finding the accuracy of our model
1:47:59 - 77. Creating our first confusion matrix
1:56:27 - 78. Making our confusion matrix prettier
2:10:28 - 79. Multi-class classification part 1 (preparing data)
2:21:04 - 80. Multi-class classification part 2 (becoming one with the data)
2:28:13 - 81. Multi-class classification part 3 (building a multi-class model)
2:43:52 - 82. Multi-class classification part 4 (improving our multi-class model)
2:56:35 - 83. Multi-class classification part 5 (normalised vs non-normalised)
3:00:48 - 84. Multi-class classification part 6 (finding the ideal learning rate)
3:11:27 - 85. Multi-class classification part 7 (evaluating our model)
3:25:34 - 86. Multi-class classification part 8 (creating a confusion matrix)
3:30:00 - 87. Multi-class classification part 9 (visualising random samples)
3:40:42 - 88. What patterns is our model learning?
#tensorflow #deeplearning #machinelearning

[@kushalacharya4048]

The best part about Daniel's teaching is that he doesn't make his students feel like they are alone in their learning and becomes their companion by showing them how he would himself figure things out. This gives tremendous hope and confidence especially to the students who are just starting out in this field which is vast and ever-growing. Great job mate. You rock!!!

[@mrdbourke]

Thank you for the kind words Kushal! You’ve got this legend

[@drm8164]

@@mrdbourke Is it all I need to get ready for Tensorflow certification examen ??? I will do it next week !

[@paulntalo1425]

It's been an amazing tensorflow learning for the last two weeks to code along part 1 and 2 tutorial videos with ease. Thank you so much Daniel

[@sesshomarudogdemon7794]

Just wanted to say thank you for working hard on providing us with all these deep and informative videos on TensorFlow and Deep Learning.

[@mrdbourke]

You’re welcome my friend, thank you for the support

[@mahletalem]

Hey Daniel, I couldn't thank you enough for getting me started with TF. I am so grateful for all the work you put into the creation of your videos. Maybe, hopefully, someday I'll get to pay it forward. Thank you!

[@quickbuck7135]

Can't thank you enough Daniel for these amazing 2 videos, coded along with you and learnt so much... massive respect

[@subhadeepchatterjee1528]

I am gonna be totally honest with you, these were the most productive 14 hours of my life, Thanks Daniel for being an excellent teacher : )

[@mrdbourke]

Woah! That’s a massive compliment! Thank you Subhadeep

[@user-kf2ez2bq2k]

@@mrdbourke Йылдыз

[@jagadeeshamudala3331]

@@user-kf2ez2bq2k ee eh

[@zakariabouidane6011]

Thank you so much Daniel. This is actually the best course I had on TensorFlow. It was very helpful so I am very thankful for the hard work and the big effort you put to make it that successful.

[@danish5326]

Just finished the 14 hours course ! Thankyou for making such a awesome tutorial Daniel.

[@mrdbourke]

You're welcome Danish! Glad you enjoyed, thank you for the kind words

[@user-jy2vf1bn7n]

Great instructor, beautifully put together. I went on and registered for the full course in ZTM and I am watching everything from the beginning again.

[@productint7660]

This is absolutely fantastic. Can't tell you how valuable it has been for me.

[@sashanktalakola122]

Just finished the 2nd part thank you Daniel for this wonderful set of videos really appreciate your content

[@bipin_the_great]

Thank you Daniel. The course syllabus is great but most importantly the way you teach is amazing and your voice is so sweet and balanced. It is like step by step guidance and every machine learning beginners must watch this series.

[@mrdbourke]

Thank you for the kind words Bipin!

[@doffn.6053]

i have come this all way loving the course, and i will be coming back again after finishing the whole thing ... thank you mister daniel.

[@user-kt2lf4bz3j]

Amazing tutorial Daniel couldn't ever thank you enough for this. This was an amazing journey with you and thank you for the amazing experience and knowledge you have shared

[@johnnychristoffersen7453]

Thank you Daniel for an excellent crash course in TensorFlow and machine learning. I like your approach and the structure of the video. Thank you again, and keep up the good work.

[@James-ys2dd]

This series is the business! Thanks Dan looking forward to more of this!

[@mrdbourke]

Enjoy James!

[@b.k.7363]

Thank you Daniel, it was an amazing series of lectures and I am really grateful to you for teaching us not only how to work with DL, but also how to think when working with DL. It was one of the most influential lectures of my career.

[@eidmone8684]

After two weeks, I finally finished the whole course! It was a great tutorial Daniel, keep up the good work!!

[@mrdbourke]

Massive effort legend! Thank you for the kind words and keep learning!

[@Ayesha-wf3sy]

I have just completed this course. It's just amazing. I luckily found this amazing content on youtube which I really need as I am going to start my final year project. Thanks for all this

[@subhabratanath4822]

Can't give you enough thanks sir, I will remember these two videos for a long time because I start my deep learning journey from here. And I think now I have a good habit to write code as much as I can rather than copying. Thank you very much sir for your effort. Love from Kolkata(India).

[@mrdbourke]

Thank you so much legend! Glad to hear you’re enjoying them :) Keep learning

[@roypearlmusic]

One of the best (if not THE...) courses I found on youtube. I'm definitely gonna take the whole course. Well done! and thanks man for this great material 🙏🏾

[@egehanasal710]

One of the best 14 hours of my life. Thank you...

[@anukarthika4ever]

Great work Daniel👍👍👍thank you for making this great course!!!

[@1234567PokemGaming]

Thank you so much, Daniel. I just finished the 14 hours course for a month with 1 hour daily.

[@rshekarLan]

Thank you Daniel, that is great lecture and hands on. Look forward to join in actual course

[@GoredGored]

Thank you Daniel. I plotted accuracy per learning rate instead. Also I extracted the dataframe (history.history) index value where accuracy is the maximum. I then filtered the whole row using the iloc method. That gives me the best learning rate for the maximum accuracy. After that I just used that learning rate, build a new model and now I am getting higher accuracy even after 10 epochs (some where in the 90s). And I am following your typing away for every new model instead of what I used to do, copy-paste and modify. Now I make a simple mistake I delete the whole thing and start again. You completely demystified the whole thing for me and great way of teaching. Thank you again.

[@TheTiz90]

Thank you, Daniel!! You are a great teacher, never forget that! :)

[@roboticol6280]

I finally completed the entire tutorial! Thank you so much for such informative and helpful content, I understood pretty much all of it and this is a great starting point for my machine learning future!

[@mrdbourke]

Massive effort legend! Keep learning

[@ElmoLikeGD]

i finished the whole 14 hour course and it is excellent! thx Daniel!

[@dreamdipesh]

Thank you for this pair of wonderful tutorials. It was very easy to follow along.

[@zeazhang1669]

It's really valuable 14 hours for me, appreciate to you, super Daniel!

[@user-zb7pc3bi6z]

It's been a long ride... Thank you so much for this tutorial 😊

[@Hellowow0112]

Best tutorial of tensorflow (part 1 and 2) on KZbin 😇👌

[@mrdbourke]

Thank you Rahul! That’s very kind of you

[@sciWithSaj]

The series was awesome thanks a lot learnt lodes of new concepts built lots of models.

[@sunnymoon5262]

If you want to learn deep learning with hands on coding this is one of the best course !

[@mrdbourke]

Thank you Sunny! I appreciate it!

[@manUNITED986]

You deserve so many more views on this, thanks my man.

[@72SteveyJay]

watched and studied the complete 14 hours, thx a lot , greetings from Belgium

[@mrdbourke]

Glad you enjoyed my friend! All the best from Australia

[@wozdog7425]

this is the best deep learning course I have come across so far! Simple with lots of down to earth explanations, tonnes of coding & very engaging ... to sum it up in a single word .... B E A U T I F U L :-) I am signing up to the 51hr udemy course on tensorflow you have done - onya mate! BTW never heard an Aussie say 'integer' the way you do !!!! :-)

[@haze13_]

amazing course, watched all 14 hours and understood it all thoroughly. Now I'm buying the full course on udemy

[@phonethiriyadana]

already started part 1 and these will be my main target for March DL Learning :)

[@mrdbourke]

Enjoy! A big March!

[@ricardofigueiredo3244]

Daniel I finished your 14 hours right now. thank you a lot for your training videos i think it was the most engaging learning I've ever had, I could learn a little deep what is behind the scenes of machine learning and deep learning, and know how to work with tensors... i'm follow you on Linkedin... I'm happy to have this first start with you one think i will keep remembering how brought me to this world... thank you thank you thank you, cant thank you more.... see you

[@mrdbourke]

Thank you so much Ricardo! Glad to hear you're enjoying the course! Keep learning legend!

[@castle_of_jokes]

hi daniel just wanted to say that you're the GOAT when it comes to ml

[@patrickboucher892]

merci Daniel. What a huge amount of work, amazing!

[@mrdbourke]

You’re welcome Patrick! Glad you’re enjoying

[@ThePtaher]

Hi Daniel - I just wanted to say THANK YOU. I watched the entire two parts and I learned a lot. 👌 well done

[@mrdbourke]

Thank you Taher! I really appreciate it!

[@anoop_bhat]

Thanks Daniel, you are the Best!!

[@tonz0123]

Thanks a lot Finally I completed it all :) loved it.

[@mrdbourke]

Woohoo! Massive effort Tanmoy! I’m glad you enjoyed :)

[@kaosarahmed6560]

Thank you for this awesome tutorial

[@sueellen360]

wow first comment! Thank you for everything you're doing for the community, been following you since you have just a couple thousands followers. I want to buy it, maybe next month or when I'm less busy, to learn and to support you for all the great free content.

[@mrdbourke]

You’re fast! The video isn’t even public yet! PS Thank you for all the kind words and long-time support

[@sueellen360]

@@mrdbourke but I read the descriptions lol

[@dul8846]

This video is amazing video .thank you so much for your teaching

[@basketballpajama4339]

Now, I'm feeling good that I have completed whole course.

[@ashishk81]

this is BEST course i have ever completed !!!! i really enjoy this course by doing it side by side I like the way he cover small details in model building and clear all my doubts Thanks a lot for making this course interesting Daniel !!:)

[@mrdbourke]

Stoked to hear you enjoyed it Ashish! Keep learning my friend

[@Yannik_Wertz]

Daniel, thank you alot for this course - it's amazing! You are an impressive teacher, it really feels like we're exploring everything together. There are many courses I clicked on, however you got me with your roadmap video. My only complain is that I couldn't keep up coding side by side, your typing speed was to fast for me. However this is a really personal thing. Will continue on Udemy :)

[@mrdbourke]

Friends, here are some helpful links: • 🤓Sign up for the full course (60+ hours of TensorFlow) - dbourke.link/ZTMTFcourse • 💻Get all of the code/materials on GitHub - www.github.com/mrdbourke/tensorflow-deep-learning/ • 📖Read the course materials in beautiful book form - www.learntensorflow.io • ❓Ask a question - github.com/mrdbourke/tensorflow-deep-learning/discussions • 💬Get live captions (if using Google Chrome) - support.google.com/chrome/answer/10538231?hl=en Happy Machine Learning!

[@JoshKonoff1]

This is literally the best film I've ever seen (beating out my childhood favorite, Terminator). Huge respect!

[@totallytubularriverrentals4082]

epic, bro. Super fun.

[@bigsmerdo]

just finished, thanks!

[@amirabbasrezaei4055]

Finally finished! I dont know what should I do after Daniel help me

[@sachingavande1095]

Hey, I am mechanical engineer and I had fear of coding as I saw your video, I can overcome that fear , thank you so much for that. You had explained very well , thank you so much for great session.

[@mrdbourke]

So glad to hear Sachin! Keep learning my friend

[@sachingavande1095]

Will you provide some resources where I can learn, real time object detection like the way you teach.

[@mrdbourke]

Are you dreaming in tensors yet?

[@sesshomarudogdemon7794]

definitely

[@zoroXgamings]

Dreaming and making them true also 😎

[@thenewculture2359]

just finishedd

[@JoshKonoff1]

nightmaring

[@taurohkea2169]

Master, I can't visualize my dream with matplotlib

[@surajmeena6797]

you're excellent teacher

[@markyu9933]

for the confusion matrix, instead of the long code you can also use seaborn heatmap. But it is also a good way to practice matplotlib using matshow.

[@vambylamby83]

I've spent like 2 weeks to complete the 14 hours of tutorial, thank you Daniel for your interesting lecture, I like the way you encouraged us to try out different ways of building the model before the "next video".

[@mrdbourke]

Massive effort legend! Excellent persistence. Keep at it!

[@vambylamby83]

@@mrdbourke OMG!!! Thank you for your encouragement mate, will keep practicing!! The journey doesn't end at here.

[@zyadmagdy5708]

Amazing video I really enjoyed these14 hours

[@siddharthchauhan3404]

excellent video tutorials, learnt a lot.

[@mrdbourke]

Stoked you enjoyed Siddharth! Thank you mate

[@user-jj3we9jv9i]

Phew. Completed it finally

[@mrdbourke]

Huge effort!!! Happy machine learning my friend!

[@hedayetulislam8211]

Its one of the great course !!!!!!

[@gtalpc59]

Hey Daniel. It would be helpful if you can make a video on distributions and different types of scaling like minmax, standard etc and when to use it. I am yet to crack that concept. I tried to read through the references that you mentioned in the video. but i am sure, if i could hear Daniel's translation about that concept, its much clearer. Thanks very much for explaining all the hard concepts in a crisp, interesting and concise language. :D

[@constantinci]

This was an amazing experience Daniel. I build NN myself a long ago when nobody even thought about its use the way as nowadays (Java, c++). Now having a "bit" more of my spare time I decided to go back to my adventure, this time with python, but at a certain moment I loss my hope to find something really useful and comprehensive. Apart of your 13 hours everything else is just a nightmare trash. Thank you. BTW you should pass somewhere some donation account. PS. I wonder if still do your 3 days fasting. I am asking because I am a regular one 🙂

[@saikatsom4450]

Hey, Daniel just completed this second module and it was really good, interesting and easy to understand. Huge respect for this effort. The way you unwrap every concept is just brilliant. I really appreciate it. If possible please mention the link of the next module.

[@mrdbourke]

Hey Saikat! Glad you enjoyed! You can find the rest of the videos/sections (another 40+ hours) on Zero to Mastery - dbourke.link/ZTMTFcourse

[@user-nb5te8ui6r]

Thank you Daniel

[@timarcher52]

These are 2 great courses! Thanks for putting them together! Does the additional content on the zero to mastery site pick up where these left off, or did you change anything in the course than what's on youtube in part 1 and 2?

[@mrdbourke]

You’re welcome Tim! And yes, the ZTM course continues right at the end of the 2nd video. There’s another 20+ hours worth of material there (more coming soon) The KZbin videos cover notebooks 00, 01, 02 But there will be 11 total, see more here: www.github.com/mrdbourke/tensorflow-deep-learning/

[@CodeSuccessChronicle]

Okay Here me!! Completed part 1, Starting part 2.

[@tanujpahuja4749]

This looks great.

[@mrdbourke]

Thank you!

[@anoldn2398]

watched Full 14hours. This is so great Daniel. Is it possible to have one on unsupervised learning as well!

[@ComputerScienceWithBas]

This is the best video ever fr !

[@mrdbourke]

Glad you like it!

[@moisesmiguel897]

Thanks for putting the course on Udemy too!

[@mehdisoleymani6012]

Thanks a lot for your great courses, is it possible for you to explain my question? How should we add non-image features to our CNN model (features like cat and dog prices) to our flatten layer? Does the CNN model new added features belong to which input image?

[@m7moodaliabuzied]

Great job. thank you

[@ShivanS]

Thanks for doing this Dan. Will be dreaming of tensors soon. Also can't wait for the bloopers reel.

[@mrdbourke]

You’re welcome Shivan! I’ve got tensor goggles now... everything gets transformed into multiple dimensions

[@urnakundu3157]

Could you provide one video on using tensorflow for data preprocessing where we will use the basic tensorflow functions and will not rely on tensorflow transoform

[@TheSgrizli]

This guy is a legend

[@mrdbourke]

[@lekanadenusi462]

Best 14 hours of my life.

[@mathecarus9595]

Thanks a lot!!

[@space_ace7710]

Completed !!

[@parthshah1563]

You are insane Man..!

[@prajwalsharma477]

completed both of the video i can't express how lkucky i was i click on you video .Thank you Daniel and hoping for more cousre for free for like student

[@user-pl3ep6rh4l]

idk how but i completed the 14H in one day and a half , that's means nothing except the tutor is so amazing

[@mrdbourke]

What a huge effort!!! Well done legend and happy machine learning! PS thank you for the kind words :)

[@user-pl3ep6rh4l]

@mrdbourke bro you are the best , keep going ✨️✨️✨️

[@user-to7dp2uc3o]

Great!!!

[@user-gv9mi6gd2e]

i got a problem on 2:41:23 after changing the loss function to SparseCategoricalCrossentropy(), the error didn't changed and i can't find the solution in documentation or on forums

[@thisaramadawalage5395]

Thank you!

[@mrdbourke]

You’re welcome legend!

[@martinbadoy5827]

Thanks a zillion for this great resource! Now I can write tf.keras.layers.Dense in my sleep :)

[@mrdbourke]

Hahaha! Get ready to dream in Tensors!

[@burakduran7719]

what an amazing and motivating 14 hours! thank you! Do you plan to do something related to CNN & RNN implementations and publish it here?

[@mrdbourke]

Thank you Burak, glad you enjoyed! You can see the full version on Zero to Mastery (including RNN’s and CNN’s) - dbourke.link/ZTMTFcourse

[@subhanbasha8813]

Hello Daniel, Is it recommend to code all the machine learning algorithms from scratch so that I can learn math behind it or just understand and start to code?

[@mrdbourke]

Depends what you’re wanting to do. Learning the algorithms from scratch isn’t 100% required but knowing how to wouldn’t hurt.

[@maheenamin7707]

hy Daniel your course is amazing... but now while fitting any model(say CNN), a new error named "InvalidArgumentError: Graph execution error: Detected at node 'binary_crossentropy/logistic_loss/mul' defined at (most recent call last):" occurs... I don't know why it is occuring now... if you can help, please do ASAP... As I am student and working on my final year project... My assessments are near and i have tried alot to figure it out but i failed.... I believe you will help me.

[@jaylonnichols5504]

Hi Daniel, Quick question in practice do you create a new model every time you try a experiment

[@mrdbourke]

No you don't necessarily need to create a model with each new experiment. You could keep going with a previous model. Creating a new model each time is just one way to keep track of various experiments. For me, I've found one model = one experiment is a good way to keep track of things.

[@sergey902]

Daniel, hi! It's been a long time since you posted motivational videos where you talk about achieving goals. Can you ever make a video like this? It is also very interesting to watch and listen about your nutrition and training. Make a video about something that is not related to programming, please :-) You're a cool guy!

[@mrdbourke]

Thank you Sergey! Thank you for the idea! Anything specific around training you’d like to see?

[@sergey902]

@@mrdbourke No, nothing specific. I'm on a ketogenic diet, just like you, and I've been practicing interval fasting for over a year. It would be interesting to know what you eat, what food you buy, and what you cook.

[@drm8164]

merci chef

[@sairampenjarla]

for a 100% accuracy in circles binary classification: tf.random.set_seed(42) model5 = tf.keras.Sequential([ tf.keras.layers.Dense(100, activation='relu'), tf.keras.layers.Dense(20, activation='relu'), tf.keras.layers.Dense(1, activation='sigmoid') ]) model5.compile(loss = tf.keras.losses.BinaryCrossentropy(), optimizer = tf.keras.optimizers.Adam(), metrics = ["accuracy"]) history = model5.fit(x_train,y_train,epochs = 50, verbose = 0) plot_decision_boundary(model5,x_train,y_train)

[@mrdbourke]

Nice!

[@chrisvagkidis3865]

Amazing tutorial Daniel, thank you so much! If I could offer something to this course, maybe it would be this: When you are plotting the history object to find the ideal learning rate, instead of calculating the learning rate for the x-axis as "lrs = 1e-3*(10**(tf.range(epochs)/20))", you can simply use the variable directly from the history object as "history.history['lr']". Hope this helps!

[@mrdbourke]

Thank you for the kind words and the feedback Chris!

[@esrasaraiera1292]

thanks alot

[@mrdbourke]

You’re welcome esra!

[@YaSoFine]

Daniel where is the extension section ? You said you have put some resources in the extension. (to go deeper into how the calculations goes on in the deeper layers)

[@mrdbourke]

Hey legend! You can find all the extra materials here: learntensorflow.io as well as on the course GitHub (see the links in the description)