1600 points by literate_coder 6 months ago flag hide 12 comments
hnuser1 6 months ago next
Fantastic work! I've been looking to build something similar. Could you share what libraries and frameworks you used?
hnuser2 6 months ago next
@hnuser1 TensorFlow.js is a good choice! Did you implement any user-based or item-based collaborative filtering?
author 6 months ago prev next
Thanks! I used Node.js for the backend and React for the frontend. I relied heavily on TensorFlow.js for the recommendation engine.
author 6 months ago prev next
I used a hybrid approach, combining both user-based and item-based collaborative filtering.
hnuser3 6 months ago prev next
Impressive. How did you handle cold-start problem for new users/items?
author 6 months ago next
@hnuser3 For new users, I used a basic content-based filtering method. For new items, I added them to a pool that would be recommended to a small set of users with similar existing interests.
hnuser4 6 months ago prev next
Did you run the recommendation engine on your own server or is there a cloud solution that you prefer?
author 6 months ago next
@hnuser4 I ran my recommendation engine on my own server since I had full control over the infrastructure and costs.
hnuser5 6 months ago prev next
Great work! Are there any specific datasets or books you'd recommend before I start my own learning journey on these topics?
author 6 months ago next
@hnuser5 I'd recommend checking out the Machine Learning Mastery blog and books written by Jason Brownlee. I learned a lot from there!
hnuser6 6 months ago prev next
Could you give more details on the scalability and performance of your solution?
author 6 months ago next
@hnuser6 I used a microservices based architecture. Each service has its own cache and is auto-scalable based on demand. With proper load balancing and caching, I was able to achieve great performance.