API Calls

We ahve created simply interactive API calls to use our system for building a business recommender system.

Notebooks

We have created two notebooks for a clear visualization of our training, evaluations, and downstream applications

Structure of System

The following illsutrates the structure of our system for ease of API calling and modification of code:

rsdb/
├── data/
├── configs/
├── features/
│   ├── featuring.py
├── math_formulation/
├── models/
│   ├── fpmc/
│   ├── tldf/
├── preprocess/
│   ├── data_preprocessing.py
├── eval/
│   ├── eval_processing.py
├── recommendation.py
├── train.py

Setting-up Training

Create a environment to work on:

conda env create

Running a training job with blf model(basic latent factor model):

python rsdb/train.py --action "train" --model "blf"

Running a training job with tdlf model:

python rsdb/train.py --action "train" --model "tdlf"

Running a tunning job with fpmc model:

python rsdb/train.py --action "tune" --model "fpmc"

Our system supports customized tunning through our yaml configs system, so all hyperparamters of tunning and training job can be tracked in the configs system. With the config system, we can tune and choose the hyperparameter that we want to use.