Thank you Professor. It was super helpful !

Score of document 3 should be 13 instead of 11.

Great vid, thank you very much!

Excellent, clear introduction, thank you!

how do i write this in python??

Clara Parkways

Rozella Ways

Amazing!

Thank you sir 🙏

Such a nice explanation!!!

THANK YOUUUUU

I wish those who make these videos would pay attention to the sound level. I can't complain too much because, other than the time invested, I didn't pay anything to watch it

Now this is how you teach. Teaching should be an adventure filled with teeny tiny realizations to produce an outcome.

i am algortithm

Great explanation, thanks!

excellent, thank you

2:32 but that point (7,4) is actually closest to (5,4) so this k-dimensional tree to find nearest neighbor actually gives you the wrong answer?

Very smooth explanation. Loved it!

Good video. You should have used the arrow / laser tool. Which points are you talking about ?

Very nice explanation! Thank you very much!

I love data.

K means this video.

So helpful. Really lucky to have found this goldmine!!

What happens if one of the centeroids does not get any datapoint assigned to it in the first clusterin round?

You´re amazing

# SUMMARY A discussion on web search algorithms, focusing on the impact of data quantity and link analysis techniques like PageRank. # IDEAS: - Web search engines handle staggering amounts of information, making architecture maintenance a significant challenge. - Google’s architecture processed 20 petabytes of data per day five years ago. - Large data volumes make computational tasks harder but simplify algorithmic processes. - A random subset of web pages is used to build search engine indexes. - Precision at rank 10 measures the accuracy of the top 10 search results. - Competitors with larger data sets can achieve higher precision in search results. - Distribution of scores for relevant and non-relevant documents remains unchanged with more data. - Precision at a fixed rank improves with increased data volume. - Search engines can improve rankings by increasing the amount of crawled data. - Larger data sets can outperform better algorithms if the latter have less data. - The density of relevant documents at the top of rankings affects precision improvements. - Historical example: Quill had an index size four times larger than Google’s. - Larger indexes lead to better search results if algorithms are comparable. - Precision as a function of rank generally decreases, with more relevant documents at the top. - More data in the index leads to better performance for free. - Link analysis techniques like PageRank are crucial for ranking web pages. - PageRank evaluates the importance of web pages based on link structure. - HITS algorithm identifies hubs and authorities in web content. - Combining large data sets with effective link analysis improves search engine performance. - Search engines must balance computational challenges with algorithmic efficiency. # INSIGHTS: - Large data volumes simplify algorithmic processes despite increasing computational challenges. - Precision at a fixed rank improves significantly with increased data volume. - Larger data sets can outperform better algorithms with less data. - The density of relevant documents at the top of rankings is crucial for precision improvements. - Combining large data sets with effective link analysis enhances search engine performance. # QUOTES: - "Google's architecture was churning through about 20 petabytes of data per day." - "Having that much data actually makes some things a lot easier." - "You can never get the entire web; nobody has the entire web." - "Precision at rank 10 would be 40%." - "The overall distribution of scores shouldn't change because you're just getting four times the data." - "Precision at a fixed rank will actually go up." - "The accuracy of the top page of your results depends on how much data you've crawled." - "Quill's index size was four times as big as Google's." - "If you have the same algorithms but four times as much data, you'll do better." - "Precision as a function of rank generally decreases." # HABITS: - Regularly update and maintain large-scale data architectures to handle vast information volumes. - Continuously gather and analyze large random samples of web pages for indexing. - Focus on improving both algorithmic processes and data collection efforts. # FACTS: - Google processed 20 petabytes of data daily five years ago. - No search engine has access to the entire web. - Larger data sets lead to higher precision in search results. - Quill had an index size four times larger than Google’s. # REFERENCES: - PageRank - HITS algorithm - Quill search engine # ONE-SENTENCE TAKEAWAY Increasing the amount of crawled data significantly improves search engine precision and performance. # RECOMMENDATIONS: - Regularly update and maintain large-scale data architectures for handling vast information volumes. - Continuously gather and analyze large random samples of web pages for indexing. - Focus on improving both algorithmic processes and data collection efforts. - Invest in gathering more data to enhance search engine precision and performance. - Combine large data sets with effective link analysis techniques like PageRank.%

u speak like Jordan Belfort...lol

Great collection of videos, Thoroughly loved it..

You just cleared every doubts on this topic, it's 10 days before my exam watching your video and getting everything cleared

why does the covariance matrix rotates the vectors towards the greatest variance?

great explanation, simple and visualized. Thanks! =)

THANKS, you've answered a lot of questions in my mind with your amazing explanation!!!!

you sound like Gale Boetticher from breaking bad

This video has (by far) the highest knowledge/time of any other video on this topic on KZbin. Clear explanation of the math and the iterative method, along with analogy to the simpler algorithm (k-means). Thanks Victor!

Content is good, but please amplify audio.

When andrew tate explaining Math

sir we cant see your cursor omg

how to know the value of P(b) and P(a)

Excellent explanation

ty

Its amazing this works at all, because the first step is to take a 2d image that makes sense, into a 1d image that has lost ALL spatial information. A 1d stream of pixels is not an image.

Splendid. Example very well portrays the algorithm stepwise!

Thanks for this. Your video helped bring clarity to the problem statement.

but is not projection (y .e)e where y = x - mew

at 3:08 the variance estimator shoud be divided by (nb-1) as corrected estimation and not nb .. that's what we call Bessel's correction

Great tutorial! But why the slop of two eigenvectors are expected to be the same?!

Outstanding!

thanks brother

Great Explanation Sir. I don't know why it motivated me to appreciate and comment on the video.

Andrew Tate of machine learning