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 | 31 |
- Gausian Density Estimation
- SQL 데이터 분석 첫걸음
- 밀도 기반 이상탐지
- One-Sided Selection
- 국비지원교육
- Sequence data
- Digital Pathology
- Inatance segmentation
- multi modal
- PM4Py
- 프로세스 마이닝
- Generative modeling
- 병리 AI
- Tomek links
- Process Mining
- Condensed neares neighbor rule
- Meta heuristic
- auto encoder
- Random Undersampling
- GAN
- Grad-CAM
- Petri net
- 딥러닝
- 거리 기반 이상탐지
- Text generation
- Fixed Learning
- Clustering 기반 이상탐지
- OCSVM
- XAI
- Data Imbalance
- Today
- Total
목록전체 글 (52)
Create Opportunities
Lv. 2 간단해 보였지만, while x=y: 정도밖에 생각을 못했다 정답 풀이 구현을 보고 놀람. 정답 풀이 def solution(x, y, n): answer = 0 s = set() s.add(x) while s: if y in s: return answer nxt = set() # new set for i in s: if i+n
Lv. 2 내 풀이 (오답) # n 미터짜리 벽 -> n 개의 구역으로 나눔. (1~n) # 페인트를 다시 칠해야 할 구역 정함. # 페인트 롤러의 길이 m 미터 def solution(n, m, section): answer = len(section) start = section[0] for i in range(len(section)): if m >= section[i+1] - start + 1: answer -= 1 else: start = section[i+1] return answer 왜 틀렸나? 이상한 것에 집착해서. : (n-m+1) 번째 이후의 페인트 여부는 좀 다르게 봐야한다고 생각했음. 정답 풀이 def solution(n, m, section): answer = 1 start = sec..
Lv. 1 내 풀이 (정답) def solution(keymap, targets): result = [] for word in targets: click_sum = 0 for spell in word: flag = False click = 101 for key in keymap: if spell in key: click = min(key.index(spell)+1, click) flag = True # 가능? if not flag: click_sum = -1; break click_sum += click result.append(click_sum) return result 1. targets의 단어를 순차적으로 순회. 2. 그 단어의 spell을 순차적으로 순회 3. keymap의 단어들도 순차적으로 순..
Lv. 1 내 풀이 (정답) def solution(cards1, cards2, goal): for word in goal: if word == cards1[0]: cards1.pop(0) elif word == cards2[0]: cards2.pop(0) else: answer = 'No'; return answer if not cards1: cards1.append('temp') if not cards2: cards2.append('temp') answer = 'Yes' return answer 원하는 단어 배열(goal)을 순회하면서 cards1과 card2 리스트의 첫 번째 원소와 동일한지 확인하고, 동일하다면 해당 card의 원소를 삭제한다. 만약 둘 다 없다? 그러면 answer는 'No' 가..
22.07.19. 제목 : Graph Convolution Networks for Cell Segmentation [PDF] Graph Convolution Networks for Cell Segmentation | Semantic Scholar Two graph-based convolution methods for cell segmentation to improve analysis of immunostained slides are proposed using advanced deep-learning, spectral-based, and spatial-based graph signal processing approaches to learn features. Graph signal processing www..
지난 방학, 연구실에서의 병리 AI는 너무 어렵게 느껴졌다. 정리했던 내용을 블로그에 옮깁니다. 22.07.05. 제목 : Graph Representation Forecasting of Patient’s Medical Conditions: Toward a Digital Twin Graph representation forecasting of patient's medical conditions: towards a digital twin Objective: Modern medicine needs to shift from a wait and react, curative discipline to a preventative, interdisciplinary science aiming at providing pe..
딥러닝 주요 개념들의 시작이 되었던 논문을 읽어볼 수 있는 좋은 기회라고 생각하였다. 딥러닝의 주요 개념을 잘 모르고 사용하는 경우가 많았다. 모르는 개념에 대해서 전부 찾아보고 확실히 이해하는 것이 쉬운 일은 아니었다. 본 논문은 Batch Normalization의 시작과 그 내용을 다룬다. Overfitting 방지를 위해서 해당 기법을 사용하는 정도만을 알고 있었고 이유는 모르는 상태였기에, 원리를 깊게 이해하고 싶은 마음이 컸다. Highlights 딥러닝의 학습에 있어서 파라미터가 업데이트됨에 따라 데이터의 입력 분포가 변경되는 Internal covariate shift 현상은 불가피하다. 이를 해결하고자 하는 Batch Normalization은 Hidden layer의 입력 분포 또한 조..
가장 많이 사용되는 eXplainable AI인 SHAP에 대한 내용이다. SHAP To explain the prediction of a data point "x" by computing the contribution of each feature to the prediction. SHAP은 Cooperative(=Coalitional) Game Theory에 기반이 된다. 쉽게 말해서, 각 변수들이 Player가 되는 것이다. 이들 각각이 y에 어떠한 기여를 했는가! 를 보고자 하는 것이다. SHAP을 이해하기 위해서는 shapley value를 먼저 알아야 한다. 이득을 공평히 나누자 는 것이 핵심 아이디어이다. SHAP의 수식은 아래와 같다. SHAP은 LIME은 지키지 못했던 세 가지 원칙을 지킨..