/tags/ml/ Recent content in ml on Shylock Hg Hugo -- gohugo.io en-us Sun, 04 Mar 2018 00:00:00 +0000 /post/2018/03/04/simple-tips-of-machine-learning/ Sun, 04 Mar 2018 00:00:00 +0000 /post/2018/03/04/simple-tips-of-machine-learning/ Tips of Machine Learning Data Quantity Representation Quality outliers missing features Irrelevant Features Feature selection Feature extraction Creating new features by gathering new data Algorithm Overfitting: too complex relative to the amount and noisiness of the training data. regularization simplify the model reducing the number of attributes gather more training data reduce noise Underfitting more complex model better features reducing the regularization Testing and Validating training set <–> validating set -finally-> testing set cross-validation If you make absolutely no assumption about data,then there is no reason to prefer one model over any other. /post/2018/03/03/overview-to-machine-learning/ Sat, 03 Mar 2018 00:00:00 +0000 /post/2018/03/03/overview-to-machine-learning/ Overview to Machine Learning Definition A computer program is said to learn from experience E with respect to some task T and some performance measure P,if its performance on T,as measured by P,improves with experience E. - Tom Mitchell,1997 So the essential of machine learning is improving P on T by learning from E,not just E. Premise to apply Machine Learning To complex for traditional algorithm. Adjust fluctuating environments. /post/2018/01/23/the-knn-algorithm-overview/ Tue, 23 Jan 2018 00:00:00 +0000 /post/2018/01/23/the-knn-algorithm-overview/ Overview The KNN algorithm compute and compare the L2 norm of vector,classify the the vector to the min L2 norm class. This means calculate the L2 norm of input vector & train data,classify the input vector to the min sum of L2.