I’m currently building less-sugar food recommendation using Tensor Flow recommendation. I faced a lot of issues. Is there anyone who wants to help me? 
Hi @icecreau ,
It would be helpful to have more information about the issues you are dealing with to start the discussion.
I’m currently working with data from MyFitnessPal that I found on Kaggle. The dataset contains information from multiple users about the foods they consume on specific dates and the sugar content of those foods. I’ve built a model based on a recommendation system approach, similar to those used for movies in TensorFlow website. My goal is to train the model to analyze a person’s food preferences and recommend similar options with lower sugar content. However, in the retrieval part, the model’s loss is quite high, and I’m unsure if the issue lies with the model itself or the fact that I’m only using 20,000 records out of the 512,000 available due to RAM limitations in Google Colab.
How many features does each example contain?
The dataset contains 587,187 days of food diary records logged by 9.9K MyFitnessPal users from September 2014 through April 2015. Each line is a tab-separated list of:
- Anonymized user ID
- Diary date
- List of food entries and nutrients (as JSON objects)
- Daily aggregate of nutrient intake and goal (as JSON objects).
I’ve extracted the data and only using user id, name of the food, and how much sugar in the food. For retrieval part, I only use user id and the food name.
20,000 records with three features each, that seems like sufficient data to create a useful model.