Hi Team, do we have any courses or pointers to learn how to build a neural network from the scratch with the handful data already present. Really appreciate your inputs on this…
When starting out, I found this free online book http://neuralnetworksanddeeplearning.com/
by Michael Nielsen to be very helpful to understand the entire structure of a deep feed forward neural network basic neurons all the way up. There is a code repository with the course with implementations of a basic network from scratch in python for classifying MNIST digits.
Thank you so much Carl :)) really appreciate it…
Welcome to the community @mageshmcc. As Carl suggested, the Neural Networks and Deep Learning book by Michael Nielsen is a great resource. You can also look into the Machine Learning Specialization by Prof. Andrew Ng where building neural networks is covered in Course 2 - Advanced Learning Algorithms.
Many thanks for the link, Carl.
Hi @mageshmcc
in addition also with hints towards LLMs after the previous explanations:
check out how how Replit built their own LLM from scratch using the databricks platform: [Training LLM] blog post
This thread could be worth a look: Best model and procedure to train a model that can answer law questions - #9 by Christian_Simonis
Best regards
Christian
Here also a great overview on deep learning bibliography by @paulinpaloalto. Feel free to check this out: Deep Learning Bibliography
Best regards
Christian
@Christian_Simonis
Hello Christian.
Thanks for the links and the information. I’m eager to read the Replit case.
Thank you for all the inputs. Really appreciate it… I also found a YouTube video from Andrej Karpathy… It really helped me a lot…
Here is the link: Let's build GPT: from scratch, in code, spelled out. - YouTube
Data Acquisition and then feeding it to ANN for the Output