# 0608 Meeting

## Rosetta - lightgbm

<!-- 10% lightgbm -->
<!--  -->


### 30%, train: 21196s, val: 1467s
被early stop


100趴的都正在跑 (前處理64G夠, 但model train到一定程度後需256G ram)


## Concall ppt
> https://1drv.ms/p/s!Aigohf1HXs8UpVUdSzFqDrXHP1K8?e=dUeQB3

解釋attentive rnn input的attention
> https://github.com/johnny12150/RecSys2019_trivago_session/blob/master/attentive_rnn_input.ipynb


## 專題生
SR-GNN

## Session-based/ Sequential Recommendation survey
> https://airtable.com/shrH8vpKlF8U0stSz