250x250
반응형
Notice
Recent Posts
Recent Comments
Link
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 |
Tags
- chatGPT
- integrated gradient
- Airflow
- 공분산
- API
- 상관관계
- BigQuery
- TensorFlow
- API Gateway
- grad-cam
- login crawling
- correlation
- tensorflow text
- Retry
- Counterfactual Explanations
- top_k
- GenericGBQException
- session 유지
- UDF
- flask
- requests
- subdag
- youtube data
- gather_nd
- hadoop
- GCP
- XAI
- 유튜브 API
- airflow subdag
- spark udf
Archives
- Today
- Total
데이터과학 삼학년
Learning rate and Batch size 본문
반응형
Learning rate and Batch size
- learning rate 는 최적화를 위해 어느 정도의 step size를 가질 것이냐 결정
> 너무 작으면 training이 오래 걸릴 것이요, 너무 크면 학습이 잘되지 않을 것임
> 보통 0.2보다 작은 값이나 1 / (sqrt(feature 개수)) 로 정함

- batch size가 너무 작으면 training 하기에 적합한 데이터 셋이 없을 수 있음 → 그래디언트를 계산할 샘플수

>> batch size가 크면 learning rate를 작게 줄이는 것이 실험적으로 좋다는 논문
728x90
반응형
LIST
'Machine Learning' 카테고리의 다른 글
Dropout (0) | 2020.01.18 |
---|---|
Optimization (0) | 2020.01.18 |
Regularization (0) | 2020.01.18 |
Feature engineering 기초 & wide and deep model (0) | 2020.01.18 |
Distributed training (feat.GCP CMLE) (0) | 2020.01.12 |