(1) minute to learn a basic 2016-09-26,30 scikit-learn regression (linear, decision trees, SVM, KNN) and integration methods (random forest, and the Adaboost GBRT)
https://blog.csdn.net / u010900574 / article / details / 52666291
turned out to be the worst performance of this calculation KNN algorithm works best
(2) 2018-12-26, several machine learning (knn, logistic regression, SVM, decision trees, random forest, the limit random trees, ensemble learning, Adaboost, GBDT)
https://blog.csdn.net/ fanzonghao / article / details / 85260775
(3) 2018-02-13, Machine Learning: Ten machine learning algorithm
https://zhuanlan.zhihu.com/p/33794257
(4) 2018-12-28, machine learning - Machine Learning | ML
https://easyai.tech/ai-definition/machine-learning/
15 Zhong classical machine learning algorithms
| algorithm | Training methods |
| Linear Regression | Supervised learning |
| Logistic regression | Supervised learning |
| Linear discriminant analysis | Supervised learning |
| Decision Tree | Supervised learning |
| Naive Bayes | Supervised learning |
| KNN | Supervised learning |
| Learning vector quantization | Supervised learning |
| SVM | Supervised learning |
| RF | Supervised learning |
| AdaBoost | Supervised learning |
| Gaussian mixture model | Unsupervised Learning |
| Boltzmann machine restrictions | Unsupervised Learning |
| K-means clustering | Unsupervised Learning |
| Expectation-maximization algorithm | Unsupervised Learning |
5) 2017-06-01 depth study notes - based on the traditional machine learning algorithms (LR, SVM, GBDT, RandomForest ) sentence matching method
https://blog.csdn.net/mpk_no1/article/details/72836042
from the accuracy, the effect of random forests best. Time above, SVM longest.
(6) 2016-08-04,8 kinds of common machine learning algorithms compare
https://www.leiphone.com/news/201608/WosBbsYqyfwcDNa4.html
usually: [GBDT> = SVM> = RF> = Adaboost> = Other ...]
(7) 2016-07-21, each classification method scenarios logistic regression, support vector machines, random forests, GBT, deep learning
https://www.quora.com/What-are-the-advantages-of Classification algorithms---different
(. 8) 2018-12-10, the LR, the SVM, the RF, GBDT, XGBoost LightGbm and Comparative
https://www.cnblogs.com/x739400043/p/10098659.html
(. 9) 2018-03 -07, machine learning (thirty-six) - XGBoost, LightGBM, the Parameter Server
https://antkillerfarm.github.io/ml/2018/03/07/Machine_Learning_36.html
(10) 2019-07-29, deep learning (thirty-seven) - CenterNet, Anchor-as Free, NN Quantization
https://blog.csdn.net/antkillerfarm/article/details/97623139
(11) 2019-05- 13, machine learning how to entry?
https://www.zhihu.com/question/20691338
(12) 2018-12-28, machine learning, artificial intelligence, depth of learning what is the relationship?
https://easyai.tech/ai-definition/machine-learning/
(13 is) 2019-09-27, machine learning
https://www.zhihu.com/topic/19559450/hot
2019-09-26, ALBERT A BERT for Supervised Self-Lite learning of Language Representations
https://arxiv.org/abs/1909.11942
(14) 2016, machine-learning learning machine
https://github.com/JustFollowUs/Machine-Learning
(15) Bradley Efron, Trevor Hastie, 2016-08, Computer Age Statistical Inference: Algorithms, Evidence and Data Science
https://web.stanford.edu/~hastie/CASI/
(16) 2019-06-05, academician Zhang Bo: artificial intelligence technology has entered the third generation
https://blog.csdn.net/cf2SudS8x8F0v/article/ details / 90986936
2019-06-08, [learning data collection] time series (time series) learning books
http://blog.sciencenet.cn/blog-107667-1183775.html
In the data analysis, because of the "small sample mathematical statistics" is not fully developed, making many new approach "based on a large sample statistical theory", in small samples did not show their advantages.
For example:
(1) when the number of "valid data"> 100,000, these new methods would be excellent performance;
(2) the number of "valid data" <200, particularly effective only when several tens or data, the performance of these new methods is not satisfactory.
(3) Mathematical Statistics in the "confidence interval" is the basis for a theory to explain these phenomena.
"Valid data" is sufficient to carry the data information of "independence" between two adjacent data.
Shannon sampling theorem, is one of the basic method of determining an "effective data".
Related Links:
[1] 2016-09-01, SVM Support Vector Machine program website
http://blog.sciencenet.cn/blog-107667-1000087.html
[2] 2016-09-01, Crosswavelet and Wavelet Coherence URL wavelet analysis program
http://blog.sciencenet.cn/blog-107667-1000091.html
[3] 2019-07-27, Weibull distribution Weibull distribution resource page collection
http://blog.sciencenet.cn/blog -107667-1191323.html
[4] 2016-09-01, ELM Extreme learning machines (ELM) program website
http://blog.sciencenet.cn/blog-107667-1000094.html
[5] 2019-09- 27, extreme value distribution Extreme Values distribution Links
http://blog.sciencenet.cn/blog-107667-1199726.html
[6] 2019-09-22, fuzzy math: Zade "ambiguity," Kalman "fuzzy of "(Bowen page collection)
http://blog.sciencenet.cn/blog-107667-1199064.html
[7] 2018-08-26, estimated Largest Lyapunov exponent matlab program to collect (URL)
http://blog.sciencenet.cn/blog-107667-1131215.html
[8] 2018-08-18, "big data" period, more eager to "small sample of mathematical statistics"
http://blog.sciencenet.cn/blog-107667-1129894.html
[9] 2017-07-11, Bradley Efron (Bradley Efron): 2005 national Medal of Science winners (statistically)
http://blog.sciencenet.cn/blog-107667-1065714.html
