Session-Based Recommendation with GNN and Time-Aware Memory Network

Algorithm 1

Session-based recommendation model with GNN and time-aware memory networks (SR-GTM).

Input: the input session and time information

Output: Top- recommendations list.

1: 

2: for l in do

3:  Learn the item representation: by Equations (1) and (2)

4: end for

5: Inner feature extraction:

6:  Calculate position information: by Equation (3)

7:  

8:   by Equation (4)

9:  Learn inner session representation: by Equation (5)

10: Outer feature extraction:

11:  Learn dwell time information: by Equations (6) and (7)

12:  

13:  Learn outer session representation: by Equations (8)–(10)

14: Learn session representation: by Equation (11)

15:  by Equation (12)

16:  by Equation (13)

17: Calculate the rating of items: by Equations (14) and (15)

18: Loss function is given by Equation (16)

19: Recommend top- items as the recommendations list