Hi everybody, that’s my first post in this huge community!
I’m writing this post since last week I’ve completed the ML course 2018 and I’ve just started the DL course 2018 but after the fourth lesson I started wondering if it could be better to stop the 2018 part1 right now in order to start the 2019 version.
Since I’ve understood that the part1 is mostly practical, I believe that learn to use the new fast.ai+pytorch is by far a better approach (I’ve noted that there are many differences between fast.ai v1 and v0.7!).
So I’m planning to complete DL course part1 v2019 and for part2 (which I believe it’ll be more technical) I’ll use the v2018 version.
What do you think about this approach, could it be good?
i prefer you to go with the part1 v2019 as it covers in depth of deep learning where you can learn about the classification and collab,nlp,and other concepts in a well structured format and jeremy does that in a cool and nice way… and i think only at collab filter concept you need some machine learning except that part every thing is completely new and you can learn by following the lecctures…
Presently i am working on the deep learning part 1 from the past 2 weeks and i have completed the 5 parts and now i m in 6 video …i will give you the flow… Video 1: in this video you learn about computer vision using fastai vision module where you get to know how to classify the images using the cnn which is pretrained and it gets the imagenet weights… Video 2:In this video you get to know …
how you can create your own dataset by running the simple script in your browser and there are set of methods in library which will give you more ease to get the images data(note:only images you can get from the script) and we get the script from the notebook) Video 3:you get to learn the nlp and other stuff Video 4:you get to learn the tabular,collab filter modules… Video 5:how the background learning works you get to know that… and deep dive into the actual math and processes…running behind the model … Video 6:deep dive into the architechture…
I believe the public ETA for 2019 Part2 is in June/July. You should do 2019 Part1, which uses the fastai library v1, and after that do 2019 Part2 when it becomes publicly available.
In fact, if you find that you have time to wait for 2019 Part2 after you finish Part1, then you should revisit 2019 Part1 again, and go deeper with your understanding of the code in the library, to try new things with it! : )
For what it’s worth, I’m doing 2019 part 2 now and also watching the 2018 part 2 videos at the same time. They’re pretty complimentary (the content is different).
So if you finish 2019 part 1 and are craving more content before part 2 is released, watching part 2 from last year for the concepts is worth it (with the caveat that the exact code isn’t really applicable anymore).
But maybe completing 2019 part1 could help me to learn fast.ai v1 with which then I could try to replicate the last year code. I don’t know, but I think it’s worth trying!
using fastai vision module where you get to know how to classify the images using the cnn which is pretrained and it gets the imagenet weights…