Lesson 6: Practical Deep Learning for Coders 2022
HTML-код
- Опубликовано: 1 авг 2024
- 00:00 Review
02:09 TwoR model
04:43 How to create a decision tree
07:02 Gini
10:54 Making a submission
15:52 Bagging
19:06 Random forest introduction
20:09 Creating a random forest
22:38 Feature importance
26:37 Adding trees
29:32 What is OOB
32:08 Model interpretation
35:47 Removing the redundant features
35:59 What does Partial dependence do
39:22 Can you explain why a particular prediction is made
46:07 Can you overfit a random forest
49:03 What is gradient boosting
51:56 Introducing walkthrus
54:28 What does fastkaggle do
1:02:52 fastcore.parallel
1:04:12 item_tfms=Resize(480, method='squish')
1:06:20 Fine-tuning project
1:07:22 Criteria for evaluating models
1:10:22 Should we submit as soon as we can
1:15:15 How to automate the process of sharing kaggle notebooks
1:20:17 AutoML
1:24:16 Why the first model run so slow on Kaggle GPUs
1:27:53 How much better can a new novel architecture improve the accuracy
1:28:33 Convnext
1:31:10 How to iterate the model with padding
1:32:01 What does our data augmentation do to images
1:34:12 How to iterate the model with larger images
1:36:08 pandas indexing
1:38:16 What data-augmentation does tta use?
Transcript thanks to fmussari, gagan, bencoman, mike.moloch on forums.fast.ai
Timestamps based on notes by daniel on forums.fast.ai
The quality of this content is unreal, thank you so much for you contribution to open education
Thank you for the great content. Loving this lecture.
Thanks Jeremy. Great Tutorial
This content is sooo good. Thanks!
My pleasure!
Thank you Jeremy.
danke schon !
Hey Jeremy, fantastic lessons. Looking forward to working through part 2!
Quick question: When working with Random Forests and XGBoost regressors/classifiers, is there ever any accuracy advantage to using ordinal encoding over one-hot encoding? I realize there can be speed advantages when the number of categories in a column grows large, but if speed isn't a factor, do we even have a reason to play around with ordinal encoding? (Sorry if this was answered somewhere and I missed it!)
Yes, there can be a difference in some circumstances. Ordinal encoding assumes that there is an inherent ordering to the categories which may not have any meaning in real life. For example, mapping "Bad", "Acceptable" & "Good" to 1, 2 & 3 makes sense, but mapping "Irish", "Scottish" & "Welch" to an ordinal list doesn't make much sense in most contexts. Some models will be more or less sensitive to factors like this. When in doubt, try both and see which works better.
I:d like to take a moment to thank Zakia for that question about the notebooks.
Great, Thanks! But what if our data is biased, is random forests still good?
Jeremy, these videos are brilliant. Thank you so much for creating them.
I've heard you mention the zoom walkthrough videos a few times. Are they available to watch anywhere?
Yes, they're on the forum (forums.fast.ai).
My man Jeremy I am looking forward to competing against you
Where it says Chapter 8 of the book, its now Chapter 9 I think.
At about 35 mins in you jumped to some notebook that I can't find and you don't explain where it is. The written text is much clearer than the verbal explanations. It would be really helpful if your notebooks were numbered corresponding to the lessons. E.g. "Lesson 5b: Why you should use a framework"
I believe this notebook is chapter 9 of the fast AI book
vocab[idxs] blew my mind!