1
00:00:00,078 --> 00:00:04,493
[MUSIC]

2
00:00:04,493 --> 00:00:08,978
So in summary, we've presented
this concept of ridge regression,

3
00:00:08,978 --> 00:00:13,420
which is a regularized form of
standard linear regression.

4
00:00:13,420 --> 00:00:16,990
It allows us to account for
having lots and

5
00:00:16,990 --> 00:00:19,800
lots of features in a very
straightforward way,

6
00:00:19,800 --> 00:00:24,750
both intuitively and algorithmically,
as we've explored in this module.

7
00:00:26,230 --> 00:00:29,390
And what ridge regression is
allowing us to do is automatically

8
00:00:29,390 --> 00:00:31,870
perform this bias variance tradeoff.

9
00:00:31,870 --> 00:00:34,240
So we thought about how to
perform ridge regression for

10
00:00:34,240 --> 00:00:38,530
a specific value of lambda, and
then we talked about this method of cross

11
00:00:38,530 --> 00:00:43,420
validation in order to select
the actual lambda we're gonna use for

12
00:00:43,420 --> 00:00:46,630
our models that we would
use to make predictions.

13
00:00:46,630 --> 00:00:48,090
So in summary,

14
00:00:48,090 --> 00:00:53,030
we've described why ridge regression
might be a reasonable thing to do.

15
00:00:53,030 --> 00:00:58,450
Motivating that the magnitude term
that ridge regression introduces,

16
00:00:58,450 --> 00:00:59,900
the magnitude of the coefficients.

17
00:00:59,900 --> 00:01:03,230
Penalizing that makes
sense from the standpoint

18
00:01:03,230 --> 00:01:07,510
of over-fitted models tend to have
very large magnitude coefficients.

19
00:01:08,970 --> 00:01:13,340
Then we talked about the actual
ridge regression objective and

20
00:01:13,340 --> 00:01:17,445
thinking about how it's balancing fit
with the magnitude of these coefficients.

21
00:01:17,445 --> 00:01:20,380
And we talked about how to fit the model

22
00:01:20,380 --> 00:01:23,900
both as a closed form solution
as well as creating a descent.

23
00:01:23,900 --> 00:01:28,490
And then how to choose our value of lambda
using cross validation, and that method

24
00:01:28,490 --> 00:01:33,240
generalizes well beyond regression,
let alone just ridge regression.

25
00:01:33,240 --> 00:01:37,815
And then finally, we talked about
how to deal with the intercept term,

26
00:01:37,815 --> 00:01:40,419
if you wanna handle that specifically.

27
00:01:40,419 --> 00:01:44,509
[MUSIC]