-
Notifications
You must be signed in to change notification settings - Fork 87
Home
- This is a basic tutorial for how to use DasiyRec
- It includes two parts, i.e., hyper-parameter tuning and test
- Let's take NeuMF running on ML-1M as an example
Make sure you can open the GUI Command Generator
Step 1: Open TUNE
tab
Step 2: Fill in Basic Settings
and click the Submit
Button
![image](https://user-images.githubusercontent.com/15884764/182068966-76341e9d-1d79-48cf-8bec-a6c50f49cd8c.png)
The useless form will be automatically disabled according to the algorithm and processing methods you choose.
Step 3: Tick 'Algorithm-Specific Settings' and click 'Submit' Button
![image](https://user-images.githubusercontent.com/15884764/182069039-4f2a51b8-4d8c-4cc3-8574-b4389a0167a6.png)
Step 4: Set the hyper-parameter searching space and click 'Submit' Button
- For int/float, simply set ['Min. Value', 'Max. Value', 'step size']
- For int/float choice, directly set ['Value'], separated by
,
as default
![image](https://user-images.githubusercontent.com/15884764/182069678-ef996939-a567-481c-9436-9630c54507cd.png)
Step 5: Paste the generated command into the terminal to start tune parameter with Optuna
![image](https://user-images.githubusercontent.com/15884764/182069753-c5ad8e91-96f5-4176-b7c2-92937f7bc445.png)
Step 6: After tuning, the results are automatically saved under './tune_log/CL_neumf_ml-1m_origin_tloo.csv'
Step 7: Results on the validation set of 30 rounds are available, and select the parameter settings with the best NDCG@50. Note that topk is set as 10, then NDCG@10 will be the optimization metric for the hyper-parameter optimization via Optuna.
Step 1: OpenTEST
in the GUI Command Generator
Step 2: Fill in 'Basic Settings' (same as tune) and click 'Submit' Button
![image](https://user-images.githubusercontent.com/15884764/182068449-3ac80cb1-60ad-4b92-ba45-28c2bf45896a.png)
Step 3: Fill in 'Algorithm Specific Settings' with the best parameter settings and click 'Submit' Button
![image](https://user-images.githubusercontent.com/15884764/182070057-5e80383b-5c49-4270-b3b2-5987925e8e77.png)
Step 4: Paste the generated test command into the terminal to start test
![image](https://user-images.githubusercontent.com/15884764/182070139-d5efc543-319b-486f-9481-377fe925dfe4.png)
Step 5: After testing, the results are saved under './res/ml-1m/origin/tloo/CL_point_neumf.csv'