Course/[Coursera] Machine Learning

[Optimization] Recognition

1. Sliding Windows In Phto OCR problem, There are few segmentation of solving problem. 1. Text detection 2. Text segmentation To solve first problem, We use Sliding windows classification. In Sliding windows, We make some windows much smaller than origin photo. And start this windows in left high corner to right below corner by each step. And by setting detection of windows as 1, we can detect o..

[Optimization] Map Reduce

1. Oneline Learning In our real world, We need to process and do prediction for continous datasets. Then how we do with this data? Online Learning is same as stocahstic algorithm, it use previous parameter theta and update for new datasets. By using this, We can reduce computational expense and can predict changed people's taste. 2. Map Reduce In Map Reduce, We use multi computer and multi core ..

[Optimization] SGD

1. How do we deal with Large Dataset? It is significant that if we have more data, more accurate prediction will come. But Problem is computational expense. If we have 1,000,000,000 data sets, then how do we process in short time? 2. Batch Gradient Descent Batch Gardient Descent is what we use previously, by taking all one by one. 3. Stochastic Gradient Descent What stochastic Gradiend Descent d..

[Recommend System] Week9 Assignment

1. selectThreshold.m function [bestEpsilon bestF1] = selectThreshold(yval, pval) %SELECTTHRESHOLD Find the best threshold (epsilon) to use for selecting %outliers % [bestEpsilon bestF1] = SELECTTHRESHOLD(yval, pval) finds the best % threshold to use for selecting outliers based on the results from a % validation set (pval) and the ground truth (yval). % bestEpsilon = 0; bestF1 = 0; F1 = 0; steps..