Articles, Blog TensorFlow and Deep Learning without a PhD, Part 1 (Google Cloud Next ’17) Related posts: Kubeflow: Machine Learning on Kubernetes (AI Adventures) Train an Image Classifier with TensorFlow for Poets – Machine Learning Recipes #6 Deep Learning Frameworks Compared Let’s Write a Decision Tree Classifier from Scratch – Machine Learning Recipes #8 #GoogleNext17 cloud Cloud NEXT convolutional network deep machine learning dense network developers GCP google cloud machine learning network neural netowrk design neural network recognition accuracy software developers software engineering tensorflow Post navigation Train an Image Classifier with TensorFlow for Poets – Machine Learning Recipes #6All-knowing Watchers | 전지적 구경 시점 [Gag Concert / 2019.06.01] 100 Comments Prof. Dr. Jens Dittrich April 19, 2017 at 12:00 pm Reply awesome! Tej Kiran April 20, 2017 at 12:33 pm Reply That is so cool!!! Awesome training!!👍🏽 Bob Mickus April 20, 2017 at 5:59 pm Reply Martin, you have a superb teaching style. A terrific way of walking-through how neural networks work. I was so drawn in and engaged, that I hardly noticed that 55 minutes had passed. Thank you for really bolstering my understanding and knowledge of neural nets and CNNs. Hope to see more from you! Rohit N April 21, 2017 at 10:39 pm Reply Woah ! this is so cool ! Russ Abbott April 22, 2017 at 4:49 am Reply I don't understand the training approach with multiple layers. With one layer one knows the correct answer, but what do you do with multiple layers? How do you know what the gradient is, i.e., which way down is? I couldn't find that in the video. Thanks. Agung Setiaji April 24, 2017 at 6:39 am Reply Martin can you share the python code fully, i wanna try by myself ujjwal tamang April 28, 2017 at 6:42 pm Reply DON'T LET THE FALSE TENSOR MOVEMENT TO BE TRASHED BECAUSE THEY CAN BE MEANING FULL FOR THE ANOTHER TENSOR MOVEMENT…! OR ANOTHER FUNCTION PUT ALL THE MOVEMENT IN A SYMMETRY SO ALL THE TENSOR BIT CAN BE REMEMBERED AND THOUGHTFUL FOR THE HIGHER PROGRAMMING DIMENSION AND MORE INTELLIGENT PROGRAMMING…! LIKE POLYFUNCTIONS OF SINGLE UNIT ACTING AND DECIDING IN A SAME TIME IN A VERSATILITY IN A SINGLE UNIT FOR THE AUTOMATIC DIVERSION AND CONVERSION, DUAL OR MULTIPLE FUNCTIONALITY..! IN ANOTHER WORD NEW LIFE WHICH CAN THINK AND ACT INDEPENDENTLY…! BECAUSE ALL OF THAT WE ARE GIVING THE INPUT THEORY FOR THE PROGRAM FOR TENSOR MOVEMENT BUT MAY BE LETTING ITSELF LEARN FROM THE MISTAKE CAN LET IT DECIDES AND MAY BE THINK OR CREATE IT SELF…! I MEAN THE PROGRAM SHOULD ALSO CAPABLE OF CREATING FROM THE MISTAKE BUT NOT ONLY TO BE TRACKED IN A RIGHT PATH..! Sitthykun LY April 28, 2017 at 9:11 pm Reply That is amazing m.ali petek May 2, 2017 at 4:46 am Reply There is a problem with sound sudheer amara May 3, 2017 at 2:45 am Reply Hi Martin, Thanks it's really help ful. Can you please share the IDE that you are using and how to open tensorflow dashboard Nicky Lim May 3, 2017 at 4:12 am Reply May i know what tools were used to visualize the training? thanks! avatar098 May 4, 2017 at 5:41 pm Reply Thank you for posting these conference videos. This was incredibly helpful and I wish to attend these conferences next time when I get the opportunity. Sanyat Hoque May 6, 2017 at 1:56 am Reply You guys are just awesome ! Tam Gaming May 6, 2017 at 8:58 am Reply Why not make the picture big and the teacher small instead- its so hard to follow what he says, when he explains something without seeing the picture. Riley Lynn May 6, 2017 at 2:47 pm Reply You could chain a side neural network based on the first learning sequence to train the dropout for the second network. Having a random dropout seems absurd since you already have tensor information extracted from your data set. Shunsuke May 8, 2017 at 2:01 am Reply MNIST is a bit cliche, but I really love this lecture!It's concise and visually clear. Highly recommended. Thanks for putting this up, Google. Egils Jugans May 9, 2017 at 1:06 am Reply 4:32 instructions unclear. pulled my wiener wat next Sebastian Bohnen May 9, 2017 at 11:16 am Reply ha ha brian777ify May 9, 2017 at 9:44 pm Reply Fantastic lecture Martin. Makes everything so clear. One of the best tutorials I've seen on any subject. Muhammad Zakarya May 12, 2017 at 9:09 pm Reply really good Ansel Castro Cabrera May 19, 2017 at 5:06 pm Reply but the RELU is not differentiable so how do you compute the derivates for computing the gradients? Synergism, Inc., May 20, 2017 at 10:31 am Reply Interesting demonstration of the simplicity of the Tensor Flow. However, the real world data in not necessarily the correct one to ensure accuracy of fit. What if that real-world data is incorrect/false due to such factors as the human cognitive dissonances, uncontrolled variables and faulty classifications (human errors) to begin with? In this view a theoretical mathematical data set could be more appropriate to ensure purity and the right fit. Luis Miguel Villalba Mazzey May 23, 2017 at 5:54 pm Reply That title is so Sheldon Cooper. Knowledge_Seeker June 3, 2017 at 9:08 pm Reply Subtitles please. Jie June 4, 2017 at 9:38 pm Reply I love this beautiful lecture, very clear. Thank you. tobeornottobe June 9, 2017 at 5:23 pm Reply Thank you for a great overview of Machine Learning and Tensor Flow. TARINEE PRASAD June 10, 2017 at 8:17 pm Reply This guy is amazing… i Wish I had a professor like this in college 🙁 Nguyen Duc Bang June 19, 2017 at 5:05 pm Reply Hi Martin,Could you explain the convolutional neural network again in your example?You choose one weight and stride it across the whole image. Am I right?What is the value of this weight? I read in the other materials and they said that we will use a small matrix and use dot product to find out the convolutional matrix -> use ReLU -> use Max pooling for the next layer,…Which one is correct here?Thank you so much Chandra P Utomo June 21, 2017 at 2:04 am Reply Great Intro! What's the IDE he used? Koushik Khan June 22, 2017 at 5:37 pm Reply Best ANN video I have ever seen. Thanks a lot sir. Enjector June 25, 2017 at 6:27 am Reply Excellent explanation, really enjoyed your video. Thank you! Abhijit Annaldas July 2, 2017 at 4:34 pm Reply LOL… Awesome sense of humour and a great talk!https://youtu.be/u4alGiomYP4?t=1689 Education only July 4, 2017 at 6:53 am Reply did you see that they usig macosx for security bcoz google dont believe in other oses. lol. Xiao Q July 6, 2017 at 8:28 pm Reply I can't believe I finally understand this. Thank you! Amazing video. Tom Ashley July 15, 2017 at 2:44 am Reply This greatly helped jumpstart me. Thank you. Dondrey Taylor July 16, 2017 at 5:10 am Reply Wow, great explanation. But I'm not going to lie, I think I still need a PhD lol Vitaliy Gyrya July 19, 2017 at 8:50 pm Reply I don't know what everyone is raving about here.The presentation is far from being clear. Way too jumpy. Some concepts are not properly introduced and have to be deduced.– That woodoo with made up issues of adding matrices with different dimensions is just that – made up issue! All because speaker decided to jump to matrix multiplication. – Also, what's the point of scaling if you already have a bias for each neuron which after exponentiation acts like scaling?– The words like "this" should not be allowed during the presentation as it is often not clear what "this" is.– At 9:12 "network will output some probability" – probability?! This concept wrt network was never introduced.– 10:21: what's the point of exponentiating something just to take log later?!– 10:21: with that definition the network that outputs all 1s is the one that minimize entropy. Nkdms. July 20, 2017 at 11:41 pm Reply Va-rάι-able 😛 LOL Gaurav Singh July 22, 2017 at 5:37 am Reply This is just pure GOLD !!!! Ansh Chauhan July 25, 2017 at 5:46 am Reply Martin is an excellent teacher, but this is the 3rd or 4th time I'm seeing the same presentation given by him. We want the next level Martin cinegraphics July 26, 2017 at 12:54 am Reply It's actually not that well explained, unless you already have experience with neural networks. Here are some things that should be improved:– explain each variable and matrix variables with more details, especially on the second screen of python– take a bit of time to explain various runs. for example what's the difference in parameters between sess.run(train_step, feed_dict=train_data) and sess.run([accuracy, cross_entropy], feed=train_data), even if it's just for the display purposes. why are the different parameter names used (feed_dict vs feed), etc.– naming of the variables should be better. X, Y, Y_ is not very intuitive. Zeeshan Ali Khan July 26, 2017 at 8:04 am Reply can we use same procedure for speech recognition? Ridahoan August 10, 2017 at 12:15 am Reply Nice intro to Tensorflow! I found the run through of a single problem helpful.A bit of a nit to pick, though — shouldn't that 99% accuracy have been tested on a final final test set that had never been seen — how many informal model tweaking iterations occurred after peeking at the accuracy on the test set? Perhaps the ending model would not do so well on a truly novel test set. Not really important here, except that we should never forget that the whole point is to generalize to unseen data, which may be drawn from a different distribution. And it is a pain in the butt to not peek. Speechrezz August 12, 2017 at 12:54 am Reply Thank you for making convolutional neural networks clearer for me! (and teaching me how to use TensorFlow) flamingxombie August 13, 2017 at 6:32 am Reply More "democratization" of ANNs for the masses (thank you, NVIDIA, thank you google). Next thing you know, we'll liken them to commodities like toilet paper. Tao Li August 16, 2017 at 5:47 pm Reply 🐮 Qiao Hu August 31, 2017 at 4:59 pm Reply Fantastic presentation Martin! Just one question: where can I find the training and test images that you used in your tutorial? Karthik Arumugham September 5, 2017 at 3:51 am Reply Thank you Martin. One of the best tutorials on TensorFlow! Btw did you use Tensorboard for the realtime visualization? Márton Balassa September 7, 2017 at 12:10 pm Reply wow I want to be AI coder now Earthcomputer September 7, 2017 at 5:55 pm Reply This video was a great way for me to get up to date with my newfound machine learning skills after taking Andrew Ng's online course. It just amazes me how radical some of this stuff is since 2011! Shubham Mittal September 14, 2017 at 11:34 am Reply Very nice sir 🙂 David Porter September 17, 2017 at 12:33 pm Reply 2:00 brains don't have an L Jin Kang September 18, 2017 at 10:55 pm Reply very very clear. 0x0055 0x0054 September 26, 2017 at 10:05 pm Reply In 1:31 the softmax function: What is the purposes for the two absolute values? Deepak Yadav October 5, 2017 at 9:41 pm Reply when does a tensorflow model converge? Rachel Harrison October 7, 2017 at 11:05 am Reply This lecture was really easy to digest. Thank you!! biorpg October 11, 2017 at 5:29 am Reply Is referring to "shooting your neurons" as "dropout" a reference to LSD? ghanshyam sahu October 25, 2017 at 5:47 am Reply just amazing Rational Israel October 27, 2017 at 6:47 am Reply Is the room cold? Irakli Koiava November 1, 2017 at 3:27 pm Reply 22:10– In reality if you want to reach the bottom of the mountains very quickly you should take long steps. 😀 joseph pareti November 12, 2017 at 9:38 am Reply the best tutorial on ML I have ever seen Martin Görner November 21, 2017 at 8:22 pm Reply The next video in the series in online: https://youtu.be/vaL1I2BD_xY "Tensorflow, deep learning and modern convolutional neural nets". We build from scratch a neural network that can spot airplanes in aerial imagery and also cover recent (TF 1.4) Tensorflow high-level APIs like tf.layers, tf.estimator and the Dataset API. Developers that already know some basics (relu, softmax, dropout, …) I recommend you start there to see how a real deep model is built using the latest best practices for convnet design. Techno Elite November 27, 2017 at 4:28 pm Reply Recommended Top Data Analytics and Deep learning courses:this awesome post from website Kupons Hub helps you to learn Deep Learning quicklyMachine Learning, Data Science, Deep Learning, Artificial Intelligence A-Z Courseshttp://kuponshub.com/machine-learning-data-science-deep-learning-artificial-intelligence-a-z-courses/ Hlophe Nkosinathi December 7, 2017 at 8:17 am Reply I would like to get the jupyter notebook or python code for this presentation where can I get it Manan Kalariya December 14, 2017 at 5:42 am Reply Nice tutorial about machine learning though you need to have PhD to get that incredibly amazing stuff. Djane Rey Mabelin December 19, 2017 at 1:13 pm Reply This video alone was soooo useful. Here is what I was able to do https://github.com/djaney/ml-studies/blob/master/06_conv.py Charlie Li December 24, 2017 at 12:54 pm Reply When I changed batch size from 100 to 50 the program does not work. But the program worked fine for batch size > 100. Weird behavior. jeds December 29, 2017 at 5:28 pm Reply Best explanation , really easy of understand !!! Big thanks !!! can someone tell me the tool name he is using? those graphs of the error and the converge I dont see it those in the tensorflow I've installed 문선형 January 4, 2018 at 12:47 am Reply metrix 666x666x666의 3D space로 이우어진 것이 큐빅..이라고 난 생각을 합니다. 문선형 January 4, 2018 at 12:47 am Reply 여기에 사람과 생명은 인자로서 변수로 작용을 하고 이것이 신경회로망을 구축하는 하나의 본보기가 되도록 노력을 합니다. 문선형 January 4, 2018 at 12:50 am Reply 뉴런은 NEW RUN달립니다. 그리고 인간의 역사는 달리기 에서 시작을 하였고 이것은 추적의 역사을 가져온 것이 현재의 사회라 나는 봅니다. 이것이 뉴런의 생성목적이고 이것이 이루어지도록 생명체의 신경회로망이 구축이 된 것이라 봅니다. 양육강식이나 사회생활을 하는 개미의 습성 이나 혹은 거북이 의 장수 비결등 입니다… 신경이 둔 하면 오래 살고 신경이나 반사신경이 빠르면 그만큼 더 욱더 빨리 노화해 가는 습성을 가지고 있다고 봅니다. 이것이 NEW RUN의 LIFE CYCLE라고 봅니다. whoislewys January 18, 2018 at 9:38 pm Reply About the image in 45:33, shouldn't the first convolutional layer have dimensions of 24 x 24 x 4? If each patch is 5 x 5, you can scan this patch across 23 possible places in the x direction, and 23 places in the y direction, correct? Or does the padding='same' make it so that the first patch's position is with 4×5 pixels on off to the left of the image ('looking at padding'), and with 1×5 pixels on the beginning of the actual image? Cedric Poutong January 22, 2018 at 3:57 pm Reply Very amazing teaching! my grand-mother also could becomme Data Scientist. Great. Thanks a lot and I hope to hear you more and more. Woulb be the same thing if I want to do a simple regression wit mixed data? satyam shekhar January 22, 2018 at 6:59 pm Reply great video!!!! helped a lot Saurabh Prakash January 23, 2018 at 8:57 pm Reply Thanks for the video, at 8:46, the shape of W should be [784,10]. Zihe Cheng February 3, 2018 at 10:58 am Reply This is the most helpful tutorial that I have ever seen. It combines the theory and the practice together. The explanation is also very clear. Dan C February 8, 2018 at 5:53 pm Reply Excellent presentation and technology. Ray VR February 22, 2018 at 8:51 am Reply One of the clearest well organized tutorials even for a beginner. Hugo Chiang February 27, 2018 at 7:17 am Reply For some reason Martin's scripts run a lot faster than my replicated jupyter notebook code. Can anyone offer some insight? Yanmin Tao February 27, 2018 at 4:29 pm Reply The accuracy you referred: how is this calculated? USONOFAV March 29, 2018 at 1:57 am Reply Tensorflow is overrated. terpenstien May 1, 2018 at 7:37 am Reply I do not understand why now this is being taught when it's been know for 2 decades. This tutorial has no current application and nowhere to go because were actually very much past this. Mike Reynolds May 10, 2018 at 12:05 am Reply I've been casually watching machine learning tutorials for over a year and this is by FAR the clearest explanation of how a Convolutional Neural Network works, out of around two dozen that I've seen. vj vargs May 18, 2018 at 10:09 pm Reply and not hot dog..very important it's a hot dog and not hot dog… it's tecchology -Jian yang Karl Pages May 25, 2018 at 4:04 am Reply Awesome 🙂 Thanks to everyone for this enlightening vid a guy June 6, 2018 at 8:12 pm Reply wow Eli Spizzichino June 11, 2018 at 1:07 pm Reply What I would never have expected it's that he got >98% accuracy without making any "shape" correlation (that's almost magic to my eyes). CNN definitely important but maybe plays a bigger role with more difficult datasets. Himanshu Soni June 24, 2018 at 6:18 pm Reply Yeah Martin, It was really good one. Bilal Khan June 27, 2018 at 10:49 am Reply fantastic introduction monoham1 June 28, 2018 at 3:55 am Reply you might not need a phd but a high school certificate in maths and 3 years working in programming is certainly not enough santiago marco July 2, 2018 at 5:52 pm Reply Confusing explanations itshgirish July 6, 2018 at 11:53 am Reply 9:05: Should it be – Σ Y . log (Y-hat) ? itshgirish July 9, 2018 at 12:43 pm Reply 45:05 – Could someone pls explain W1[5,5,1,4] ? … i dont understand whats a patch of 5,5 and applying 4 of those to the images. im ur mother father gentleman July 31, 2018 at 2:26 pm Reply We don't need to have PhD just for utilizing Tensorflow/DL, however that doesn't mean we can learn them within 1 hours without having any pre-knowledge or reading a text before. I'm also an novice not having any real tensorflow coding myself, but have read a related text, still am confusing at many parts. So I visited yt and watched this. I'm not sure it's perfect or not. But this video was very very helpful to me. Thanks Mr. scarf. 🙂 Alexander Aroff September 23, 2018 at 8:31 pm Reply Hey, shouldn‘t the probabilities sum up to 1? @10:20 Vijay Prabhakaran December 9, 2018 at 2:06 am Reply Very nice talk conveying the big picture and the general intuition. I do not think anyone can convey and address everything from the big picture to the nitty gritty in a 1 hr talk, given that I think this was a excellent introduction. Nice work Martin ! shanaka jayatilake December 15, 2018 at 11:16 pm Reply Such a great presentation. Pham Xuan Trung December 27, 2018 at 3:15 pm Reply Very clear and easy to understand. The lecture gives by Gooogle is alway great and incredible! Pubudu Goonetilleke June 14, 2019 at 6:37 am Reply This is a great presentation. Thanks for sharing. How can I use this with my own color images instead of using minst data set ? Can I create my own color mnist data set (how) ? Francis Thibault September 5, 2019 at 6:51 pm Reply Great course, it helps a lot! Leave a Reply Cancel Save my name, email, and website in this browser for the next time I comment.