1
00:00:00,000 --> 00:00:11,862
Let's recap what we've learned this week.
First we spoke about search, indexing, and

2
00:00:11,862 --> 00:00:16,776
ranking.
But if you will notice we.

3
00:00:16,776 --> 00:00:24,200
Left out, an important part.
While alluding to, the difficulty.

4
00:00:24,200 --> 00:00:29,520
Of, assembling.
Results, which are very large.

5
00:00:30,080 --> 00:00:36,560
At the same time ranking them efficiently.
The homework problems.

6
00:00:36,840 --> 00:00:43,793
Are related to this pro, this issue, and I
also expect that you'll have some animated

7
00:00:43,793 --> 00:00:47,850
discussion on the forum on this particular
point.

8
00:00:47,850 --> 00:00:54,028
At the same time, please don't give away
the exact answers to the homework

9
00:00:54,028 --> 00:00:58,286
questions.
Though the discussions we will certainly

10
00:00:58,286 --> 00:01:03,045
help all those having some difficulty with
the homeworks.

11
00:01:03,045 --> 00:01:09,725
Next we discussed enterprise search and
searching structured data, we found that

12
00:01:09,725 --> 00:01:15,745
it's not a solved problem and.
It's a little more difficult than

13
00:01:15,745 --> 00:01:21,060
searching.
Web data for a number of reasons.

14
00:01:22,100 --> 00:01:29,544
Finally, we turn to object search, and
covered a very important technique called

15
00:01:29,544 --> 00:01:35,952
locality sensitive hashing.
We also briefly touched upon associative

16
00:01:35,952 --> 00:01:43,020
memory, which is a related technique, but
doesn't actually store the objects.

17
00:01:43,520 --> 00:01:50,296
Now if you think about the problem of
searching for objects, and instead of

18
00:01:50,296 --> 00:01:57,977
actually wanting to find the objects which
are near your query, you only want to find

19
00:01:57,977 --> 00:02:02,405
which set of objects this is likely to
belong to.

20
00:02:02,405 --> 00:02:09,543
In some sense, you want to find the class
of objects, without worrying about the

21
00:02:09,543 --> 00:02:13,700
exact objects that are similar to one's
query.

22
00:02:13,700 --> 00:02:19,724
A bit like a associated memory.
Well, this is exactly the machine learning

23
00:02:19,724 --> 00:02:23,851
problem.
And that's we will going to get into next

24
00:02:23,851 --> 00:02:30,700
week, as you will find the approach that
we are gonna take for machine learning via

25
00:02:30,700 --> 00:02:37,055
information theory, is slightly unique and
not covered in many normal machine

26
00:02:37,055 --> 00:02:41,594
learning textbooks.
So, see you then and don't forget to

27
00:02:41,594 --> 00:02:44,153
submit your homework by Monday.