Think DSP 1.2 Spectral decomposition The most important topic in this book is spectral decomposition, which is the idea that any signal can be expressed as the sum of sinusoids with different frequencies. The most important mathematical idea in this book is the Discrete Fourier Transform, or DFT, which takes a signal and produces its […]

# Month: October 2018

## Setting up your goal

Single number evaluation metric Use a single real number evaluation metric ML Project : Loop(Idea -> Code -> Experiment) needs evaluation There are lots of evaluation metric. ex) Precision , Recall for classifier Combine these evaluation metric to new one. ex) F1 score which algorithm is better? for example, use the Average Satisficing and Optimizing […]

## Cyclical Learning Rate (CLR)

https://github.com/bckenstler/CLR https://www.datacamp.com/community/tutorials/cyclical-learning-neural-nets http://teleported.in/posts/cyclic-learning-rate/

## ML Strategy

ML strategy 데이터 모으기, 구조 개선하기 등 수많은 아이디어 중 가치 우선순위를 두는 전략이 필요 Orthogonalization 특정 결과를 위해 어떤 것을 튜닝할지 정하는 절차 -> orthogonalization(직교화) 여러 기능들에 대해 각각 method들이 있음 각 method들의 조정에 따라 나오는 결과가 독립되어있어야 조절하기 쉬움 만약 method들의 영향이 서로 중첩되면 조절하기 어려움 ex) tv 화면 조정 버튼 : 버튼 […]