CAPP Lab.

Program

All the programs take place on August 2, 2022. All the timelines are shown in KST (UTC+9).

10:00AM-11:00AM	Invited Keynote: Dr. David Seunghyun Yoon (Adobe Research, San Jose) Title: Semantic Textual Understanding for Information Retrieval Abstract : Semantic textual understanding is a fundamental research topic engaging in many text-related tasks (semantic search, word/phrase/sentence-level similarity/representation). Recently, we have observed huge advancements in this area, leveraging pre-trained language models. In this talk, I’ll introduce recent techniques for building textual representation models, their strengths/limitations, and our research direction. Further, I’ll present our research that leverages a pre-trained textual encoder to retrieve textual information and generate a synthetic training dataset for a named entity recognition task.
11:00AM–12:00PM	Invited Keynote: Prof. Mingoo Seok (Columbia University, New York) Title: Tiny-ML: Sub-Microwatt Keyword Spotting and Emerging Ultra-Low-Power Intelligence Devices. Abstract : Wearable, embedded, mobile, and Internet of things (IoT) devices are expected to generate an unprecedented amount of data. Locally computing on those data is critical for many reasons, including latency, power consumption, and privacy and security. To enable this vision, for the last few years, the VLSI Design Lab at Comlumbia University has been working on VLSI systems with TinyML capablilty. TinyML is an emerging area with a high potential, which is performing on-device sensor data analytics at extremly low power, typically in the mW range and below. The power efficiency of silicon artificial intelligence is getting close to that of biological intelligences (100's of pW per neuron). Divisive normalization is very effective to tolerate background noise, even unknown noise types. Analog feature extraction can aovid power-hungry Nyquist-rate A/D and DSP. In addition, relaxing non-linearity in AMS circuits enables large power savings.
12:00PM–1:00PM	Lunch Break
01:00PM–02:00PM	Invited Keynote: Dr. Paul Hongsuck Seo (Google) Title: Multimodal Learning for Videos Abstract : Human perceives the world through multiple sensory systems (ex., vision, audition, touch, smell), which work together complementing each other and therefore it is natural to build a model processing multiple modalities simultaneously especially given rich multimodal online data such as modern video contents containing visual frames, audio signals, human speech and metadata. Moreover, the emergence of online video sharing platforms enabled the collection of large-scale video datasets for such tasks. In this talk, I will introduce our recent studies on multimodal learning in videos and explore the techniques that we can apply to improve the multimodal understanding capabilities of models. Through our extensive sets of experiments, we show significant improvements from incorporating multimodal input signals and effectiveness of those proposed techniques. the contract.
2:00PM–3:00PM	Student Talks Seungmin Lee (SNU), Stability-Plasticity Dilemma in Class-Incremental Learning Viduranga Munasinghe (SNU), Deep Learning-Based Training Dataset Generation To Improve Product Recognition Rate At A Checkout-Free Store.

About

SNU GoGE Workshop 2022 is a program hosted by the department of Electrical and Computer Engineering at Seoul National University, and supported by the Brain Korea program. The Brain Korea program is is a human resource development program initiated by the Korean government. The objective of BK21 FOUR is to produce the next generation of world class leaders in these fields by upgrading research infrastructure.

We sincerely thank our invited speakers, Dr. David Seunghyun Yoon, Dr. Paul Hongsuck Seo, and Prof. Mingoo Seok. We also thank all the participating students in the student talk session, where the students are from the research groups of Prof. Kyomin Jung, Prof. Bohyung Han, and Prof. Hyuk-jae Lee at SNU.

Contacts: Hyuk-jae Lee

SNU GoGE Workshop 2022
: Artificial Intelligence and Hardware

Attending Instructions

Invited Speakers

Program

About