search

Tristan-Chow / Recommendation-System-based-on-TensorFlow

Neural candidates generation network based on Youtube reommender paper
19 stars 8 forks source link

about DataInput.py #2

Open liupengfei66 opened 5 years ago

liupengfei66 commented 5 years ago

I think in Datainput.py line 39 - 41 should be for item in batch_data: for l in range(len(item[4])): watch_history_matrix[row_count][l] = item[4][l] row_count += 1 because I think watch_history_matrix is represent user watch video history,this is just for fixed the length, am I understand wrong? thank u.

Tristan-Chow commented 5 years ago

Sorry for late reply for my busy school stuff, yeah, your suggestion is right,I make a small mistake here when uploading files and have changed it. Because number of samples in my dataset is limited, my thought is to fix watch_history_length for each user and fill the matrix with movie_index.