constant images - Algolit Introduction to Machine Learning

20160423 110658

Sun, 24 Apr 2016 09:47:31 +0000

20160423 110712

Sun, 24 Apr 2016 09:47:50 +0000

20160423 110735

Sun, 24 Apr 2016 09:48:15 +0000

Yann explainng the iterative improvements a model such as the Perceptron makes

20160423 110847

Sun, 24 Apr 2016 09:48:35 +0000

Linear Regression

20160423 111408

Sun, 24 Apr 2016 09:48:52 +0000

20160423 111411

Sun, 24 Apr 2016 09:49:12 +0000

20160423 111416

Sun, 24 Apr 2016 09:49:31 +0000

Iterative improvement model performed, going down the mountain

20160423 111533

Sun, 24 Apr 2016 09:49:54 +0000

Linear Regression, iterative improvement if the change allows for less errors, you keep it

20160423 111544

Sun, 24 Apr 2016 09:50:19 +0000

local minimum vs global minimum

20160423 112058

Sun, 24 Apr 2016 09:50:42 +0000

Leaping & non-linear Regression

20160423 114359

Sun, 24 Apr 2016 09:51:02 +0000

20160423 124429

Sun, 24 Apr 2016 09:51:25 +0000

Layers of a neural network explained

20160423 124503

Sun, 24 Apr 2016 09:51:44 +0000

Neural network example: combining 2 linear separators to make a non-linear separator

20160423 175030

Sun, 24 Apr 2016 09:52:05 +0000

20160423 175049

Sun, 24 Apr 2016 09:52:29 +0000

word2vec - measure distance between 'Belgium' and 'Brussels', high chance that same distance applied to 'France' will give you 'Paris'

20160423 175107

Sun, 24 Apr 2016 09:52:49 +0000

Perceptron explained: neurons * weight, sum of all, threshold, fire!

20160423 175117

Sun, 24 Apr 2016 09:53:07 +0000

Evaluation: train error < test error + ((square root) no of parameters / no of examples)

20160423 175133

Sun, 24 Apr 2016 09:53:28 +0000

20160423 175156

Sun, 24 Apr 2016 09:53:47 +0000

The Perceptron 'mountain' slopes

20160423 175205

Sun, 24 Apr 2016 09:54:06 +0000

Neural network mountain slopes probably look like this