One of the articles seems to be miss-linked to a redit post. This seems to be the correct link for “Loss landscape visualization 1 -- Seeing sticky plateau”
No problem. Thank you for all the amazing links! I look forward to every week.
I was particularly interested in this article so I tracked it down. It reminded me of what I learned from this related excellent video from Welch Labs: https://youtu.be/NrO20Jb
That visualization of loss landscape in 3D is inherently difficult, so I was curious what approach this article took.
Consistently one of the best data science newsletters out there. The weekly curation saves so much time sifting through the noise — always a great mix of practical ML content and cutting-edge research. Really appreciate the work that goes into each issue!
One of the articles seems to be miss-linked to a redit post. This seems to be the correct link for “Loss landscape visualization 1 -- Seeing sticky plateau”
https://kindxiaoming.github.io/blog/2026/loss-visualization-1/
Hi David - great catch! Thank you so much for a) catching it and b) putting the correct link.
I've updated the post.
Really appreciate you letting me know :)
No problem. Thank you for all the amazing links! I look forward to every week.
I was particularly interested in this article so I tracked it down. It reminded me of what I learned from this related excellent video from Welch Labs: https://youtu.be/NrO20Jb
That visualization of loss landscape in 3D is inherently difficult, so I was curious what approach this article took.
Consistently one of the best data science newsletters out there. The weekly curation saves so much time sifting through the noise — always a great mix of practical ML content and cutting-edge research. Really appreciate the work that goes into each issue!