Are you going to post other lecture videos for this new semester? I have been following your teaching for the past two years and can’t wait for this new one.
Haha, yes yes, very slowly. I've uploaded this one because I couldn't look at the book I'm writing 🤢🤢🤢 you know, after a while you get a little disgusted by what you're doing. Anyhow, in a week you'll have part two. Then I'm not sure when I'll have the time to edit more videos.
I'll do my best, then 😅😅😅 The videos will be very similar to last year, but I won't add much (any?) special effects. The topics are the same, but I explain them sightly differently. So… yeah, I'm not sure how much value there's in editing and uploading them. What's your opinion?
Every time a person repeats something, he rethinks it. Therefore, even if it is the same material, it still says something that was not there before, so i think there is much value. I've watched courses over several years and have read most of the papers that were listed. I already know most of this material, but I always find something new (Deep learning is growing very fast). By the way, in the tiger puzzle, which is used for pretext task in the video, the combination number is predicted (the output predictor has the number of outputs equal to the number of combinations of fragments as I remember it)
Thank you for this video, Alfredo. Excellent as usual! Will your new book include topics on Self-supervised learning? Really hope it does! As a side note, I want to let you know that you're an inspiring person inside and outside of the deep learning field.
Yann wants me to include some joint-embedding SSL. I don't have a lecture myself about that, so that has to come first. And thanks for your kind words 😊😊😊
I don't want to upload the same content multiple times. Every new semester I change something, but not everything. But after 3 editions (Fall 21, Spring 22, and now Fall 22) there's enough diversity. Still, I'm spending all my time on writing the textbook. So, perhaps I'll have some time during the winter break.
