Machine Learning Systems - 1 Introduction

[giscus[bot]]

Machine Learning Systems - 1  Introduction

https://harvard-edge.github.io/cs249r_book/contents/introduction/introduction.html

[emmanuel2406]

Am excited to dive into a comprehensive overview of this rich vein within CS ... I can see so much potential for impact on the world!

[profvjreddi]

Hi! 👋 I'm thrilled that you're excited to explore the book. Happy reading, and I hope you find it both informative and inspiring! Keep me posted.

[PandaKungF]

I am in the very beginning of my ML implementation journey. I am very excited to read your book. Thank you for your work and sharing it!

[profvjreddi]

Welcome! I'm excited to have you join us. ML is a fast-paced but highly rewarding subject to study. In this book, we've tried to get down to the key concepts; it is a work in progress though since the field is still evolving. If you have any questions or just want to share your thoughts as you go along, don't hesitate to reach out. Your feedback also means a lot and can help make the book even better. So keep us posted. Feel free to post any thoughts or questions in the discussions section, which's here: https://github.com/harvard-edge/cs249r_book/discussions

Happy reading and enjoy the learning process!

[Mjrovai]

This is a Fantastic Introduction to the book. ML systems are definitely the bones of AI. Congratulations, and thanks for such amazing thoughts.

[SYLee1996]

Thanks for sharing, this seems like a great learning resource!

[profvjreddi]

Thank you, @SYLee1996! It’s a work in progress, and I’m improving it. Are there any specific areas or topics you feel could benefit from deeper exploration? If anything comes to mind, please let me know—I truly value constructive feedback, and it’s what keeps me motivated!

[GreggatElle]

Excellent work coming out of the man cave VJ :-)

[profvjreddi]

Haha, sometimes you’ve just gotta retreat to the cave to get things done!

[Imran-Iqbal]

Awesome! Fantastic! Terrific!

[profvjreddi]

Thank you for your enthusiasm and support. They motivate me to keep working on it! 🙏

[Nandalib]

I am thrilled to explore your book. Thanks for the comprehensive overview focusing on both theoretical as well as practical domains. Referring to this information ''This interconnectedness can create either virtuous or vicious cycles. In a virtuous cycle, high-quality data enables effective learning,.....
'' Is that really possible that we could be above to improve the accuracy after quantization? In my case, I have achieved SMAPE of 7%, marking a 3% improvement compared to the original unoptimized model.

[profvjreddi]

Thank you for the feedback! I always appreciate knowing that people are reading and learning from my work because that’s what motivates me.

'' Is that really possible that we could be above to improve the accuracy after quantization?

To answer your question, quantization introduces small noise or perturbations to the weights and activations due to reduced precision. This makes it act as a form of regularization, which helps prevent overfitting, particularly in over-parameterized models, which is often the case with many models. The models that overfit the training data often struggle to generalize well to unseen data, so regularization helps cope with this issue. I hope that helps!

[Nandalib]

Thank you so much for your feedback! and I appreciate the opportunity to connect and share ideas.