Neural Networks From Scratch
October 2, 2022•376 words
created_at: 2021-05-18
The best way to learn how something works is to build it from scratch. Unsurprisingly, the Neural Networks from Scratch book does exactly that.
I had the privilege to edit it alongside Harrison and Daniel, and wanted to write about my experiences editing a technical book.
On January 10th, 2020, Harrison, or sentdex, released a YouTube video announcing a Kickstarter campaign for funding Neural Networks From Scratch. On January 10th, I backed the Kickstarter 😂 This was the first time I would formally edit something, as well try my hand at understanding the inner workings of neural networks. Initially, hundreds of people were making suggestions on the Google Docs page and it was hard to find new suggestions; however, as traffic tapered off and I read more closely, I started to make substantial improvements.
I edited for hours each day and on January 26th, Harrison reached out to me!
It was so exciting to be contacted by a YouTuber I respected so much, and we began working more closely and I began editing on private pages before the book was released for edits from the larger Kickstarter backer pool. It was similar to being on the alpha team and pushing changes to the beta team. Working more closely also meant I needed to devote more time and work on their schedule. This was difficult at first, but soon I found an editing strategy that worked for me.
The majority of my edits through September were only grammar suggestions, but around October, I understood enough about neural networks to begin making some small changes there as well. Though many of my edits didn't cut it, I did make a few helpful suggestions and, to me, that was amazing. I went from only knowing how to throw Tensorflow or PyTorch at machine learning problems, to understanding the backend enough to improve it.
Near the release of NNFS, Harrison began to upload the NNFS YouTube series. Covering the same topics as the book, this tutorial series proved incredibly popular and well-made.
Through editing NNFS, I went from having a typical computer science student's understanding of machine learning to knowing enough to get first place in three AI hackathons.