I noticed some mistranscriptions in the transcript (subtitles, captions) for the course videos. Below is the errata for the 1st week. I’ll post the errata for the other weeks in those groups.

The format for the errata is as follows. The approximate time is in braces, then the mistranscribed word or phrase follows, followed by a hyphen, then followed by the correct word or phrase. Like this:

[time] mistranscribed word or phrase - correct word or phrase

Video 1 - Welcome!

[0:28] Two sets - tool set

Video 5 - optimization objective

[0:19] Key means - K-means

[0:30] Creating dissent - gradient descent

[0:41] A means - K-means

[1:32] Constant century - cluster centroids

[2:00] Extent - x 10

[2:21] cluster androids - cluster centroids

[2:38] some from Michael’s one to him - sum from i equals 1 to m

[3:01] Kenyans - K-means

[4:33] For kings - K-means

[4:40] Centralism, U1 and U2 - centroids, mu1 and mu2

[4:58] Key news - K-means

[5:38] New one - mu one

[6:40] closest sanctuary - closest cluster centroid

[6:55] cluster central one - cluster centroid one

[7:07] sign it too close to central to - assign it to cluster controid 2

[7:20] album - algorithm

[7:23] sign points to cluster century - assign points to cluster centroid

[7:29] UK - mu k

[8:05] aside to it - assigned to it

[8:10] closest centroid - cluster centroid

[8:54] years - here

[10:08] duration - iteration

[10:25] great inter sent - gradient descent

[10:42] album - algorithm

[10:51] album - algorithm

Video 6 - Initializing K-means

In this video there appears to be a difference in timing between the subtitles and the audio.

[0:44] cluster central’s K - cluster centroids K

[1:43] blue cluster sent troy - blue cluster centroids

[1:56] pray end up - probably end up

[2:01] classes - clusters

[2:05] cost of central - cluster centroids

[2:57] kineys\g raw end - k-means will

[4:50] kings algorithm - k-means algorithm

[6:23] ranking means - run k-means

[6:52] cluster centuries - cluster centroids

[8:40] cost of centroids - cluster centroids

Video 8 - finding unusual events

[1:55] Flow - flaw

[3:42] Once - ones

[10:01] Users - usage

[10:04] Racial - ratios

[11:00] that hope - that helped

[11:29] Albums - algorithms

Video 9 - Gaussian

[7:44] Encodes - in code

Video 10 - anomaly detection algorithm

[0:17] end features - n features

[0:22] within - with n

[0:35] excise - Xis

[1:42] Probably - probability

[2:07] physical independence - statistical independence

[2:08] discuss - this class

[2:43] new 1 - mu 1

[4:12] fronts - runs

[5:00] in the nominee - an anomaly

[7:34] Album - algorithm

[8:17] Into the - in this

[8:52] meme - mean

[10:10] to - two

[10:18] Album - algorithm

Video 11 - developing and evaluating an anomaly detection system

[6:50] fictious - features

[10:04] [inaudible] - rates

Video 12 - anomaly detection vs. supervised learning

[0:07] Michael’s 1 - y = 1

[1:36] She - Here

[1:40] an obvious - anomalies

[4:38] In the nominate section - then anomaly detection

[5:41] Stretch - scratch

[5:42] Stretches - scratches

[5:51] Stretched - scratched

[6:00] Stretches - scratches

[6:03] Stretch - scratch

[6:07] Stretched - scratched

Video 13 - choosing which features to use

[0:42] anomaly - algorithm

[1:23] hissed a gram - histogram

[1:38] history graham - histogram

[1:48] hopes - helps

[1:56] hissed a gram - histogram

[2:27] hissed a gram - histogram

[3:59] hissed a gram - histogram

[4:33] hissed a gram - histogram

[4:40] hissed a gram - histogram

[4:52] hissed a gram - histogram

[5:08] hissed a gram - histogram

[5:41] excellent power of - x to the power of

[6:12] excellent - x in

[6:33] hissed gram - histogram

[7:01] hissed gram - histogram

[7:13] hissed gram - histogram

[8:05] trust validation set - cross validation set

[8:59] galaxy - Gaussian

[10:54] learning anomaly - learning algorithm