우리의 사명은 전 세계의 정보를 체계화하여 보편적으로 액세스 가능하고 유용하게 만드는 것입니다. AI는 우리가 사용자, 고객 및 전세계의 문제를 해결하는 놀라운 새로운 방식으로이를 가능하게합니다.
AI는 내가 좋아하는 사람들의 사진을 검색하거나, 언어 장벽을 허물고, 개인 디지털 보조기로 일을 끝내도록 돕는 등 일상 생활을 쉽게합니다. 그러나 그것은 또한 오래된 문제를보고 우리가 일하고 사는 방식을 변화시키는 새로운 방법을 우리에게 제공하고 있습니다. 우리는 모든 사람들이 접근 할 수있을 때 가장 큰 영향이 올 것이라고 생각합니다.
네이버가 인공지능 기반의 콘텐츠 추천 시스템 에어스에 이어, 개인화 상품 추천 시스템인 ‘에이아이템즈(AiTEMS, Ai+items)’를 자체 개발해 모바일 쇼핑판에 시범 적용했다고 9월1일 알렸다.
에이아이템즈는 쇼핑, 검색, 뉴스, 네이버TV, 네이버스포츠 등 여러 네이버 서비스 도메인에서 활동한 이용자 개인의 행동 데이터를 기반으로, 관심사나 취향을 분석해 상품을 추천하는 인공지능 추천 시스템이다.
에이아이템즈는 네이버의 여러 서비스에서 이뤄지는 사용자의 행동 이력과 상품의 메타 정보를 분석해 상품으로 추천한다. 상품메타정보란 상품명, 카테고리명, 가격, 판매처명, 상품 아이디 등의 텍스트 정보와 상품 이미지 정보 등을 말한다. 이미지 정보는 텍스트 정보와 결합해 활용될 수 있도록 컨볼루션 신경망(CNN) 기술을 바탕으로 벡터로 변환 후 사용된다. 네이버에서 사용자의 구매 경험이 없어도 네이버의 여러 서비스를 이용한 이력이 있다면 이용자 개인에 적합한 상품을 추천할 수 있다. 예컨대 사용자가 네이버에서 ‘야구장’을 검색하거나 관련 영상 등을 시청했다면 에이아이템즈는 사용자가 야구에 관심이 높다고 판단하고 에이아이템즈 추천 쇼핑 영역에서 다양한 야구 용품을 추천하는 식이다.
추천에 활용하는 사용자 이력은 쇼핑 관련 로그와 그 외 검색, 콘텐츠 클릭 및 구독 활동 로그 등을 포함한다. 네이버는 “개인정보 이슈가 없도록 동의받은 범위 안에서 활용한다”라고 알렸다.
네이버는 모바일 쇼핑판 첫 화면 중하단에 에이아이템즈가 적용된 ‘에이아이템즈 추천 쇼핑(베타)’ 영역을 구성했으며, 모바일 쇼핑홈과 모바일 쇼핑 MY 탭을 통해서도 해당 서비스를 제공한다. 에이아이템즈는 현재 패션 분야에 선 적용되었으며, 점차 범위를 확대해나갈 예정이다.
Call for application for Machine Learning Camp Jeju 2017
If you have studied machine learning/deep learning and TensorFlow, you probably want to implement a non-trivial and large-scale system for real use. We invite you to the month-long Machine Learning Camp Jeju 2017, where you can make that dream a reality.
For a full month in beautiful Jeju Island, you and other participants will train a deep learning model using TensorFlow from start-to-finish. Jeff Dean (Google Senior Fellow via Hangout), Rajat Monga (Google/TensorFlow Director (TBC)), and Prof. KyungHyun Cho (NYU) will give us keynote talks. Plus, you will have access to experienced mentors including Namju Kim (Head of Research for Kakao Brain), Sung Kim (HKUST), Lucy Park (TF-KR), Donghyun Kwak (TF-KR), Terry Taewoong Um (TF-KR), and many more. We hope you take advantage of this wonderful opportunity.
Those selected as participants will be provided with one round-trip airfare (up to 300 USD) to Jeju Island (South Korea), room and board at Jeju National University, USD 1,000 in stipends (can be used for the airfare, etc.) and USD 500 to 1,000 in Google Cloud Credit. In addition to these benefits, participants will gain valuable and practical experience in the field of deep learning. We look forward to your application!
Mentor Recruitment: If you’re interested in sharing your experiences and expertise with the camp, please contact us at mljejucamp@googlegroups.com. You will serve as personal mentors to 1 to 2 participants, holding 2 to 3 on/offline meetings a week to help them successfully complete their projects. While it is possible for you to provide online-only mentoring, we suggest you visit Jeju Island to meet with your mentees in person. We will provide round-trip airfare (up to USD 300) to Jeju Island and up to five (5) days of room and board.
(Information regarding schedule, program and benefits are subject to change as we are in the process of finalizing the details. We will have more information later.)
Camp Overview
Date: July 3rd through 28th, 2017 (Check-in date: July 2nd)
Participants: 20
Location: Jeju National University / Kakao Space.1
Organizers: TensorFlow Korea User Group, Kakao, Google, Smart Grid CK Center in Jeju National Univ, Jeju Center for Creative Economy and Innovation, Jeju Local Government (Subject to change)
“I will implement paper X from 2016 NIPS Conference using TensorFlow and apply idea Y to the implementation”
“My goal is to add idea X to existing TensorFlow model Y and apply it to dataset Z” (Please justify why you are interested in the particular paper, model, dataset, etc. Write your proposal as detailed as possible as it will be the primary criteria to select participants.)
Basic Tasks (but not limited to)
Each participant will implement own deep learning related ideas and recently published ideas (in ICML, ICLR, NIPS, etc.) in TensorFlow. Or adapt already implemented ideas to new dataset. Participants will propose in the application.
Participate in camp program. (10AM-5PM on weekday basis from July 1 to July 30)
Deep learning and TensorFlow expert mentors will advise each participant.
Release the code on the github as Open Source.
Camp rules
participants may be dismissed from the camp for the following or similar reasons:
Repeatedly engage in behavior that negatively impacts other participants' work
Spend unreasonable amount of time on non-camp related tasks.
More than 3 missed camp days without proper notice.
Important dates
Application due: April 20 (AOE time zone)
Notification: May 10
Mentor assignment and online discussion: June 1
Camp starts: July 3
About Jeju
Located just off the coast of the Korean peninsula, Jeju Island is the largest volcanic island in Korea. Also known as Asia's Hawaii, the island is rife with beautiful sceneries and getaway resorts. Jeju boasts several natural treasures including Mount Halla, the country's highest peak, Trail Olle that winds around the rocky coastline, and Sunrise Peak, a dormant volcano ideal for catching sunrises and sunsets. You will be able to fully experience the island's charm for as long as a month without a visa. For more information, please visit at: http://www.visitjeju.net/en/index.jto.
About Camp
Deep Learning Camp Jeju 2017 is a month-long program (July 3-28, 2017) where participants gain hands-on experience with TensorFlow through individual-based projects with the goal of implementing new deep learning related ideas, and/or already-published ideas. We are looking for approximately 20 participants. More than a dozen industry experts with strong backgrounds in deep learning and TensorFlow implementation will serve as project mentors to guide participants.
FAQs
Q: What are we doing during the one month program?
A: Basically, we design a deep learning model and fully implement using TensorFlow. It is also possible to (re) implement a published paper (by others) and adapt it for new datasets. Based on this, each participants will propose their own plans in their application.
Q. Can I apply for partial participation? (i.e. only weekends)
A: Unfortunately No.
Q: Is this only for students?
A: No. Anyone who can spend one full month in Jeju Island in Korea, and work from 10AM-5PM during the weekdays can apply.
Q: Is this a contest?
A: No, this is not a contest. Individuals will have different projects.
Q: Is this a training or teaching program?
A: This is not specifically a teaching event. The applicant should have good understandings on programming, machine learning/deep learning, and TensorFlow. However, we will provide mentors to assist you on your project.
Q: Will data for training be provided or it is up to participants?
A: We will provide some public data sets but participants can also utilize their own data.
Q: Should I bring my laptop?
A: We won’t provide PCs. You need to bring your laptop. However, we will provide cloud server credits.
Q: Training takes a lot of time and computing power. Does the camp provide any support?
A: We will provide cloud server credits.
Q: What are the criteria for selecting applicants?
A: There is no formal criteria, but we are looking for interesting and feasible projects.
Q: What if the attendee cannot complete the work that they submitted?
A: There is no penalty, but mentors will guide each participant to success.
A: It’s OK. As long as the project is interesting, we will consider it. You can also propose a small portion of a larger project that you wish to work on for a month.
Q: Is this only for deep learning? Can I propose a reinforcement learning project?
A: Yes, reinforcement learning is acceptable. Feel free to include other types of interesting machine learning projects.
Q: Can I participate in a keynote session or open seminars even though I am not a camp participant?
A: Yes, the keynote and open seminars are open to everyone. We will have a separate announcement regarding keynote sessions and open seminars.
Q: Will you also be recruiting staff members for the camp?
A: Sorry, but we have no current plans to recruit staff members.
Q: I do have more questions. Where should I contact?
