Taking AI from prototype to production - MFML Part 3
Vložit
- čas přidán 2. 08. 2024
- Making Friends with Machine Learning was an internal-only Google course specially created to inspire beginners and amuse experts. Today, it is available to everyone! This video is the third installment of a six hour session, covering the second half of "Lifecycle of an AI project" (Steps 6-12 of our 12 step applied AI guide).
The course is designed to give you the tools you need for effective participation in machine learning for solving business problems and for being a good citizen in an increasingly AI-fueled world. MFML is perfect for all humans; it focuses on conceptual understanding (rather than the mathematical and programming details) and guides you through the ideas that form the basis of successful approaches to machine learning. It has something for everyone!
Part 1 is available at bit.ly/mfml_part1
Part 2 is available at bit.ly/mfml_part2
To stay tuned for Part 4, don't forget to hit that that subscribe+notify button!
Looking for hands-on ML/AI tutorials? Here are some of my favorite 10 minute walkthroughs:
AutoML - console.cloud.google.com/?wal...
Vertex AI - bit.ly/kozvertex
AI notebooks - bit.ly/kozvertexnotebooks
ML for tabular data - bit.ly/kozvertextables
Text classification - bit.ly/kozvertextext
Image classification - bit.ly/kozverteximage
Video classification - bit.ly/kozvertexvideo - Věda a technologie
Index of steps 6-12
0:00:45 Step 6: Train models
0:25:14 Step 7: Tuning and debugging
0:43:45 Step 8: Validation
0:56:34 Step 9: Testing
1:27:25 Step 10: Production
1:53:11 Step 11: Launch
1:56:58 Step 12: Monitoring
Steps 0-5:
Step 0: Find an application where ML is useful - czcams.com/video/lIFLeHDanmA/video.html
Step 1: Set objective - czcams.com/video/lIFLeHDanmA/video.html
Step 2: Data engineering - czcams.com/video/lIFLeHDanmA/video.html
Step 3: Split data - czcams.com/video/lIFLeHDanmA/video.html
Step 4: Explore data - czcams.com/video/lIFLeHDanmA/video.html
Step 5: Get tools - czcams.com/video/lIFLeHDanmA/video.html
MACHINE LEARNING QUEEN 👑🙌
Most complex ML concepts and principles made simple - spot on as always and simply genial. Thanks Cassie!
My main takeaway is the debugging dataset - hardly any mention in the online material on this. Makes total sense that using the validation set for tuning will result in data leakage, overfitting and poorer generalization.
Finally.... Loving this series❤️
So brilliant! sharing this in the MLOps community now! thank you for putting that time into this!
I love this. You are amazing Cassie.
Did you publish the 12 steps classical statistics course that you talked about ?
Also, are you considering making a book out of this series ?
Was waiting for this for so long
Finally!!!! Will watch ASAP
Thank you for sharing and Merry Christmas!
Make sure you have 10 times as many instances as features
1:12:24:“The p-value is the probability of obtaining a sample at least as extreme as the one we just observed in a world where the null hypothesis is actually true.”
1:13:10: “A small p-value makes your H₀ [null hypothesis] look ridiculous.”
1:17:10: "A p-value is the probability of getting a test performance at least as good as ours if the model is actually garbage."
very informative
Awesome work! Are further videos going to be published?
more please
There's some AI/ML-specific stuff in the middle there, but an awful lot of that sounds much the same as any other well-planned system launch.
Beautiful person
She is impressive she has no ums, ah, or repetitions on over 6hrs of talk? she is stunning too. Damm
Wonder Where she got the graphics in the slides
"But today the reality is that you should be doing this"
☺️☺️☺️☺️☺️☺️☺️😘😘😘😘😘
This just proves my bigger C3P0 theory
speaking two hours even without drinking sth. hats off
It's nice to be seen and appreciated. Most people don't notice that.
@@kozyrkov It may not be healthy. I used to teach and not having water had health issues. Amazing energy and very expressive way of telling the facts. I could relate to the mistakes I was doing most often :)
I beg to disagree that pvalue is hard to describe
The coin tossing expt she conducted was a teachable moment. The audience member who picked up the coin said there was no tail marking!! That is surprising and low pvalue