일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 |
- Array operations
- scatter
- 정규분포 MLE
- 최대가능도 추정법
- Numpy
- Python 유래
- Operation function
- subplot
- seaborn
- pivot table
- 부스트캠프 AI테크
- namedtuple
- type hints
- Comparisons
- groupby
- 가능도
- unstack
- linalg
- BOXPLOT
- 딥러닝
- boolean & fancy index
- 표집분포
- VSCode
- 카테고리분포 MLE
- python 문법
- dtype
- Numpy data I/O
- Python
- Python 특징
- ndarray
- Today
- Total
목록전체 글 (160)
또르르's 개발 Story

1️⃣ P Stage (프로젝트) 1) 이미지 분류 주어진 이미지를 가지고 Mask 착용, Gender, Age 총 18개의 class로 분류하는 문제 전체 사람 명 수 : 4,500 한 사람당 사진의 개수: 7 [마스크 착용 5장, 이상하게 착용(코스크, 턱스크) 1장, 미착용 1장] 이미지 크기: (384, 512) Git Repo Title Description Link Git Repo P Stage 01 Git Repo https://github.com/bcaitech1/p1-img-JAEWOOSUN 프로젝트 Title Description Link [Stage1-01] Baseline 작성 https://dororo21.tistory.com/114 [Stage1-02] Data Processin..

문제 정의 성능 높은 모델들에 대한 리뷰 해결 아이디어 논문 : https://arxiv.org/abs/2010.14061 진행 상황 1. Key Idea 이전 논문들에서는 Encoder에서 BERT를 사용하지만, Value Generation부분에서는 RNN Decoder를 사용하는 아이러니.. Purely Transformer-based framework를 사용 즉, Single BERT가 Encoder와 Decoder 모두에서 work → 이렇게 하면 prediction objective와 value generation objective가 BERT 하나만 optimize하게 됨 Encoder(BERT)에서 사용한 hidden states 값을 Decoder(BERT)에서 Re-use가 가능해짐 2...

CHAN-DST slot imbalance 문제를 해결하고자 adaptive objective를 도입. a contextual hierarchical attention network (CHAN)를 사용 : dislogue history에서 relevant context를 찾기 위함. → 각 턴의 발화로부터 word-level 관련 정보 검색 → contextual representation으로 encode → 모든 context표현을 turn-level관련 정보로 집계한 후 word-level 정보와 합친 output 생성. state transition prediction task Definition T : turn Ut : user utterance of turn t Rt : syste..