Boostcamp AI Tech (Day 036)
My assignment: 탐색적 데이터 분석 (landmark data)
Lightweight models
-
Decision making
- 연역적(deductive) 결정
- 귀납적(inductive) 결정
- data ($\rightarrow$ training) ($\rightarrow$ compression) $\rightarrow$ inference
-
Decision making machine
- 평균(mean)
- 가장 나이브한 머신러닝 모델
- (70,80,90) not decided $\rightarrow$ 80 decided
- $\Leftrightarrow$ model(data) $\rightarrow$ “cat”
- 분류기(classifier)
- 평균(mean)
-
Lightweight (경량화)
- 학습시킨 큰 모델을 작게 compress
- TinyML
-
Keywords
- Backbnone & dataset for model compression
- Edge device
- dumb and fast
- Edge intelligence
- edge training
- edge inference
- edge offloading
- edge caching
Optimization
-
Optimization & Decision
- Optimization problem
- MST, HamCyc, MVC
- decision problem을 여러 번 반복하며 근사, 귀납
- Decision problem
- DST, DHamCyc, DVC
- Yes / No
- 참고 자료(알고리즘)
- Optimization problem
-
Constraints
- Objective: maximize performance while $Cost_1 + Cost_2 + … \leq Constraint^*$
-
Model compression
- objective: performance
- constraints: costs