[Feed] 2021.07.17 tds

2021. 7. 17. 13:21madquer/Feed

https://towardsdatascience.com/kats-a-generalizable-framework-to-analyze-time-series-data-in-python-3c8d21efe057

 

Kats: a Generalizable Framework to Analyze Time Series Data in Python

Forecast, Detect Change Points, Anomalies, and Get Key Statistics in Time Series

towardsdatascience.com

Kats : time series analysis Framework by facebook

10 forecasting models in Kats

[Linear, Quadratic, ARIMA, SARIMA, Holt-Winters, Prophet, AR-Net, LSTM, Theta, VAR]

 

 

https://towardsdatascience.com/introduction-to-normalizing-flows-d002af262a4b

 

Introduction to Normalizing Flows

Why and how to implement normalizing flows over GANs and VAEs

towardsdatascience.com

Normalization for GANs and VAEs. Not yet.

https://towardsdatascience.com/hyperspherical-alternatives-to-softmax-6da03388fe3d

 

Hyperspherical Alternatives to Softmax

Can Hyperspherical Alternatives Compete Against Softmax?

towardsdatascience.com

HPN(Hyperspherical Prototype Network)

softmax : cross entropy 오차 최소화 = 올바른 출력의 크기 최대화

H = the number of hidden unit in last layer

C = the number of class

HPN : 클래스 수 대신 미리 정의된 고정 차원이 있는 출력 레이어

HPN 의 출력 단위 수가 클래스 수보다 훨씬 적을 수 있으므로 많은 수의 클래스에 효율적일 수 있다.

네트워크 출력에서 미리 정의된 프로토타입까지 제곱 코사인 거리를 최소화

차원이 다르므로 네트워크가 인스턴스를 올바른 클래스의 프로토타입으로 "rotate" 하는 방법을 배워야 한다.

손실함수는 올바른 클래스의 프로토타입만 필요하다.  프로토타입 직교(벡터) -> 덜 유사한 것 분리

but, HPN 은 softmax 에 비해 거의 이점이 없다.

https://towardsdatascience.com/matplotlib-linear-regression-animation-in-jupyter-notebook-2435b711bea2

 

Matplotlib Linear Regression Animation in Jupyter Notebook

Creating animation graph with matplotlib FuncAnimation in Jupyter Notebook

towardsdatascience.com

matplotlib.animation FuncAnimation -> visualize data with animation

https://towardsdatascience.com/stop-learning-data-science-to-find-purpose-and-find-purpose-to-learn-data-science-23f6f5e568ac

 

Stop Learning Data Science to Find Purpose and Find Purpose to Learn Data Science!

How I Flipped the Educational Model to Become a More Effective Data Scientist

towardsdatascience.com

Data science : tool 사용여부보다 실제 문제를 해결할 수 있는가에 초점

https://towardsdatascience.com/how-to-detect-random-walk-and-white-noise-in-time-series-forecasting-bdb5bbd4ef81

 

How to Detect Random Walk and White Noise in Time Series Forecasting

No matter how powerful, machine learning cannot predict everything. A well-known area where it can become pretty helpless is related to time series forecasting. Despite the availability of a large…

towardsdatascience.com

https://towardsdatascience.com/causal-reasoning-in-machine-learning-4f2a6e32fde9

 

Causal Reasoning in Machine Learning

An investigation through some of the main limitations Artificial Intelligence-powered systems are facing

towardsdatascience.com

기계학습과 인과관계

 

 

'madquer > Feed' 카테고리의 다른 글

[Feed] 2021.07.26  (0) 2021.07.26
[Feed] 2021.07.25  (0) 2021.07.25
[Feed] 2021.07.20  (0) 2021.07.20
[Feed] 2021.07.19  (0) 2021.07.19
[Feed] 2021.07.18 tds and etc  (0) 2021.07.18