LG AI Research Attends AIIS Spring Retreat

LG AI Research attended the “AIIS Spring Retreat” held by the Artificial Intelligence Institute at Seoul National University.

Marking its third anniversary this year, AIIS Retreat is an event wherein renowned researchers at Seoul National University and leading AI companies in the industry meet to share and exchange the latest research trends. Byoung-Tak Zhang, head of the AI Institute at Seoul National University, explained that the event is dubbed “retreat,” meaning rest during the war, as it aims to gather companies on campus for a day to address AI issues beyond the competition.

This year’s AIIS Retreat was held on April 30 at the Hoam Faculty House and lawn square in Seoul National University. Participants include 3 major AI-related member companies such as LG AI Research, NAVER, and Hyperconnect and about 55 students from Seoul National University.

The participants gave presentations on free topics, including recently published papers, AI promotion plans of each company, and examples of current use of AI technology. From the LG AI Research, Woohyung Lim, head of the Data Intelligence Lab, gave a presentation on “Advancing AI for a Better Life” along with a congratulatory speech; researcher Sungha Choi gave a presentation on “RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening.”

A number of papers by various researchers were introduced through poster sessions, with LG AI Research publishing the following:

- Hyunwoo Kim, Online Continual Learning in Image Classification: An Empirical Survey
- Janghoon Han, Fine-grained Post-training for Improving Retrieval-based Dialogue Systems, NAACL 2021
- Sangmin Lee, Explaining Convolutional Neural Networks through Attribution-Based Input Sampling and Block-Wise Feature Aggregation, AAAI 2021
- Dongsub Shim, Online Class-Incremental Continual Learning with Adversarial Shapley Value, AAAI 2021
- Jongseong Jang, Integrated Grad-CAM: Sensitivity-aware Visual Explanation of Deep Convolutional Networks via Integrated Gradient-based Scoring, ICASSP 2021
- Yeonjeong Jeong, Ada-SISE: Adaptive Semantic Input Sampling for Efficient Explanation of Convolutional Neural Networks, ICASSP 2021

Dongsub Shim's poster 

Having formed a membership with the AI Institute at SNU and continued academic exchanges, LG AI Research plans to advance such industry-academe interactions and keep exploring areas of cooperation to address on-site AI issues.



You can watch the LG AI Research presentation video at AIIS Retreat via YouTube.
LG AI Research Youtube (Link)