Lecture 1: Algorithmic Thinking, Peak Finding
HTML-код
- Опубликовано: 6 авг 2024
- MIT 6.006 Introduction to Algorithms, Fall 2011
View the complete course: ocw.mit.edu/6-006F11
Instructor: Srini Devadas
License: Creative Commons BY-NC-SA
More information at ocw.mit.edu/terms
More courses at ocw.mit.edu
It's so cool that these top schools release courses like this one online free of charge. I may not get a chance to go to MIT or Stanford, but I'm sure as heck going to learn as much as I can from them. Thanks!
sellout
MIT gets billions of government funding. Billions.
I paid for my MIT education in cash!
Its like real life.. you can go in most big university classes, sit down and learn for free... (as long as there's an open seat)
ATXpert sellout because hes sharing knowledge?
MIT Tuition is $49,892/yr $24,946 /per semester which is about $4,989.20 per course. Semester runs 15 weeks. So that equals to about $332.61 per week.. I'd say this course runs twice a week, so this day of class cost $166.30, and you got it for free. Now that's what it cost.. what it's worth is much more.
Moral of the story. Be thankful to have such information at our fingertips. FREE MIT courses on Algorithms?!? How awesome is that!?
True, but just to point out, that $166 also includes the ability to talk with peers, professors, teaching assistants etc. It might also include the usage a cluster or supercomputer that MIT runs.
And such free education is only good for some courses like Computer Science or theory knowledge in most of the other disciplines. For example, if you are learning electronics, you can get free theory knowledge from such videos, but you wont get any practical knowledge about soldering or circuit design etc coz you would need real hardware. Same applies with biology etc.
However, we should indeed be grateful for such videos.
@@blasttrash It's worth noting that students are also paying this money for a qualification and proof for potential employers that they possess all of this knowledge.
Goodwill hunting
A library card goes a long way
@@blasttrash
EE student here,
bought components, soldering iron, and multimeter on Amazon
Saw free lectures on RUclips and instructables
Knowing what the knobs on a osciliscope do before touching it is great, most in my class had no clue(Myself included)
But ultimatly expirience of any kind is king, especially self taught
You may not have a supercomputer, but you defenitly got the knoledge and will to learn
Some in those classes don't even have that
@@slackerengi2401 True its all about will. I am from ECE too, although I graduated long back and I dont even know how to solder stuff. I can barely remember the difference between series and parallel connections on a bread board.
I work in CS field now, so all is good I guess. :)
"Inefficient but correct is definitely better than being efficient but incorrect" - Srini Devdas
Pretty obvious.
"Move fast and break things." - The Zuck
the mantra of cryptograph and information security in general
"but it works" -literally every programmer
Really grateful to MIT and Srini for making this course accessible through OCW... Really enjoyed this course
Alongside this course what did you follow to master ds, algo?
CLRS
Was it worth??i mean whole course
@@KrutikaPatil0109 you read all of it?
@@juanmoscoso0 I doubt that. She did Andrew Ng's deeplearning specialization, was familiar was supervised, unsupervised and semi-supervised learning from work.
The lecture starts at 0:22.
Course overview : 2:51
Course content : 8:37
First problem (Peak finder) : 15:31
"Straight forward" algorithm : 18:53
"Divide & Conquer" (a recursive) algorithm : 27:42
Second problem (2D Peak finder) : 36:16
Greedy Ascent algorithm : 37:30
"Divide & Conquer 2D" algorithm (defunct) : 42:40
"A working 2D recursive" algorithm : 47:35
bro could plz suggest me the best data structure video channel...i'm a begineer
You're awesome
doing god's work
Wow I love you bro.
Where's my fucking son?
I failed math in high school and over 20 years later actually enjoyed watching and learning from this. I’ve been interested in algorithms and appreciate MIT sharing this class session.
girl you are not alone. but i am thinking its because its now that you can understand what you can do with mathematics. in our days we were taught math and physics like robots. We had to shove everything inside that head or get whippings.
@@PHlophe ngl it's still taught like that
@@mastershooter64 this is a straightforward Algorithm
endless supply ...Stanford, Berkeley, CMU, Cal Tech and Harvard ...machine learning, AI....
Math is something one can learn at any age , yes it can be difficult but with propper material this challenge we call education can become simple when all the steps are clear and at any given time the next step is obvious . Join a community online , that is much better than self learning by its self
It feels like this Professor is very passionate and knowledgeable about the course material! The hour flew by
set playback at higher than 1x... flies by even faster ;P
Try Harvard CS 50
@@jeffschlarb4965 that's a foundational course, this one's lil more advanced than that
@@harv609 I was speaking in regards to a passionate teaching style...
@@jeffschlarb4965 oh I see, yeah. I agree
Learning a course after a lecture feels like I'm into college, Thank you so much MIT for all your open source curriculum 😊🙏🙏🙏
This is still the best on-line introductory course for algorithms I am aware of. Coupled with programming exercises, it would make legendary content for MITx and edX.
No bullshit, no cheap jokes. Just pure joy of knowledge. How college courses should be taught!
Thanks MIT for the course. I can now understand things very easily as a correspondence student, it's very hard to understand things, but MIT open course-ware has made life and learning very easy.
Honestly, the best lesson I've ever watched!
MIT is the best for a reason! I'd never hesitate to dive right into MIT OCW by this time.
Keep it up MIT ....it's such a nice job to give lectures from excellent professors in online for completely free.....THANKS
I just wish i had find this earlier in my career, but back then I was not ready to understand how blessed I am to access this. MIT is dream but I can access such courses at my fingertips now!!
Complete course: 47 Videos
ruclips.net/p/PLUl4u3cNGP61Oq3tWYp6V_F-5jb5L2iHb
I really feel like I am at MIT and having a live Lecture. Thank You, MIT
What if the cameraman started learning from this lecture and finds himself becoming a programmer :)
He’s already programmed a robot to do the camera work.
MIT president is the cameraman.
Yep
Or the janitor.
Nothing wrong with being a cameraman. If you imagine a world with only programmers, it would be awful
Great initiative by MIT. I never imagined i will be able to sit and learn from an MIT professor for free!! 👍
എല്ലാ മലയാളികൾക്കും എന്റെ ഹായ്
from kannur......
@@amalfrancis4846 njanum kannur
@@amalfrancis4846 evideya
Kottayam
puthenvelikara
I remember taking an intro to Algos class as an undergraduate way back when. And I remember not having the slightest idea how to get O(logN) for the peak finding problem. And then just being in total shock when some student from the back answered it the first day, like at @24:52, thinking that kid was the next Dijkstra or Primm lol. I'm now in my 30's grinding away at leetcode and reading CLRS for interviews. And now that I think about it, that kid at @24:52, was already doing what I'm doing now well before he even took the class. Start while you're really young!
do you think in long term software engineer role is good considering the dynamic nature of job scenario
@@utkarsh_108 software engineering as in dumb code, probably not. But there will always be a need for people that can solve problems and put that into code as a tool or final product
The array there is not necessary sorted, so binary search doesn’t seem like a valid approach. Did I miss something?
@@ViktorKishankov All you need to do is find ANY peak in the array. The search starts in the middle, if the value is rising to the left, there must a peak somewhere on the left, same for the right. Otherwise, you must already be on a peak. Imagine you were on a mountain looking for a peak, you would head in the direction where you have to walk uphill. Binary search works because it is any peak, not the greatest peak.
Watching this video from Argentina and I think it's amazing and great you guys share this w/ the community. Thanks!!
Don't worry about peak finding algo. If u r confused about binary vs random for 1d peak finding or global peak vs relative one, just learn that this video is for introduction and give students with non programming background a feel of algorithms and a little of divide & conquer. Hope u watch the whole playlist and i can assure you there are no such confusions in later videos!
Thanks MIT for the amazing playlist
I feel really dumb about this problem.
I don't understand how it's log(n). I figure you need to check around you on both sides to know if you're a peak. But also you can't eliminate more elements than the ones directly around you as potential peaks because the array is unsorted. How are we cutting potential solutions in half on every check? I've understood some DSA concepts, but even this first class just left me confused lol.
Edit: Ah I misunderstood the problem. We're looking for any peak not the highest number. We know that if a number is higher we can jump into the middle of that side of the array and if the number to our left is higher than there must be a peak between the one we started with and the one we just chose. If the number on the right is higher instead we just keep cutting through. Worst case scenario the numbers just increase constantly in one direction and we eventually find the only peak on the extreme right of or left of the array.
@@das6109 That was exactly my concern lol, from first go it looks like we may skip a peak on the side we're skipping, but seems like if A >= B, then there is a guaranteed peak on side which includes A :)
Thank you MIT for uploading this content. Its truly a blessing and you are doing a good service to our planet Earth.
this is wonderful. Not having to depend on incomplete resources across the internet and then coming across the complete course taught by the BEST is wonderful. Please do put more such courses and help those students who can't afford to have a quality education for one or the other reason, thank you
Is it a complete DSA course??
I actually passed the equivalent course to my university by watching these videos. I also scored an almost perfect score! These videos are a goldmine! Thanks a lot for providing these courses to the general public.
se poio panepisthmeio
@@annapatata6258 Μετσόβιο
@@giwrgosnikolaidis1096 etsii kai egw ekei eimai
@@annapatata6258 odws? Ara h hmmy h semfe eisai. Opws kai na exei bohthane full. Egw me auta ta video perasa algorithmous.
@@giwrgosnikolaidis1096 hmmy eimai 1o etos
Prerequisites
A firm grasp of Python and a solid background in discrete mathematics are necessary prerequisites to this course. You are expected to have mastered the material presented in 6.01 Introduction to EECS I and 6.042J Mathematics for Computer Science.
If you have not taken and been successful in each of these subjects, please speak with a TA or professor before enrolling. We do allow students who have equivalent, other experience with the material described above to enroll, but with the firm understanding that mastery of this material is assumed and that course staff will not feel obligated to cover it or to help students who are struggling with it.
6.006 is a 12-unit (4-0-8) subject and serves as a Foundational Computer Science subject under the new curriculum. It is a direct prerequisite for 6.046 Design and Analysis of Algorithms, the theory header.
Wait so I won't learn anything if I only know C?
Bored Chinese Even though this class uses Python you could still learn something. Really depends on your ability to grasp and apply the lessons to the C language.
Hi Could you please provide the link or direct to proper links for the prerequisite courses. I searched in ocw.mit site but there are 4 courses(Spring 2005/2010 and Fall 2005/2010) with 6.042J as title.
Peter Reginald thanks
Hello, I wanted to ask If I focus on Java programming and I would like to have a better grasp of algorithms and data structures, in order to help me in Programming, is this course appropriate for that or should I look for something different ? Thank you.
This channel is a gold mine. I'm so happy about finding this :D Thank you very much.
This channel is a goldmine. Great information
I have absolutely no clue what he's talking about but yet I'm here listening like I'm really trying to learn it .
Spend some time on a Basic Data Structure course. It will make sense then.
Awesome! Please go try to get attention somewhere else then please!
@@yojodingy6334 Can you please explain what an algorithm is?
@@Thisisnotmyrealname8 no thanks, watch the video
@@yojodingy6334 hey!! stop acting Smart and try to get a life please!!
One of those best initiatives I have ever come across. It would create monopoly in the Indian market over all ed-tech platforms and offline coaching tuitions. Only a matter of awareness. Never heard a negative review about this.
I am so glad that I found this. Been interested Algorithms for a long while.
Great lecture. I'm looking forward to the rest of the series!
Wow what a course !! the professor makes it so easy to understand algorithms thank you MIT !
What a time to be alive for having such high quality course for free, free education is the best thing that can happen to humans, learn from each other for free, i understand it's difficult to have free education or anything really, because those people need to be payed to live also, but if we can find a way to have as much free education as possible it's just incredible and highly valuable for humanity, and MIT are doing a great job by contributing to this project/idea, thank you
I seriously envy the professors at MIT. The profs at my university can't hold a candle to their organization and ability to simplify difficult concepts.
My uni professors make simple concepts more difficult, what's worse is that they do it with confidence.
So much knowledge in just an introduction course, love it.
Those of you interested in algorithmic puzzles, check out my new book "Programming for the Puzzled" mitpress.mit.edu/books/programming-puzzled
www.amazon.com/dp/0262534304/
It is very awesome that we can follow a great courses from great institutions like MIT for free .
I cried. Just visit the link in description and see how wonderfully everything is arranged for us to learn. Thank you team MIT. I cannot explain what this means to a student like me from small indian village. It's a dream come true. Best gift for life.
hey bro i have just learned all the basic and important things on java..can you suggest me from where i can study and concepts of dsa..plz help me i am about to sit on my college placements
@@anarghanayak8586 ruclips.net/p/PLUl4u3cNGP6317WaSNfmCvGym2ucw3oGp
xd
Just finished 6.0001 and 6.042J, now starting 6.006, just as a head start before I join university next year. Really grateful to MIT to provide these lectures for free.
you will have a great understanding already going into uni. Good luck on your studies. I wish I had dived into these courses when I started my programming journey.
I'll do those 2 first you did, they seem pretty good. Thanks for commenting man !
@@aishwaryadharmadhikari7165 yup they are very helpful and infact they are the prerequisites for this course.
@krit grover could you please share the link of 6.0001 and6.042J,and 6.006 ,pizzazz
It would be a great help for others, plzzz share
@Aishwariya dharmadhikari plzzz provide all the link
I LOVE how the teacher handles the syllabus!
Buy the book CLRS. Will help you with this course. Get it here: amzn.to/2EGg3kG
I felt smarter after watching just the course overview, excited to watch the entire series!
Hey
@Deonex replying to a 7 y/o comment xdd
Hello, how are you after 7 years?
@@danid1464 I'm a full on software engineer! Did an internship at Microsoft and have been working at startups which interest me since then
@@JohnnyDoeDoeDoe man, this gives me hope for the future. I'm wondering, did you go to a university?
I like how the chalk is soft ^_^
svellene Lol Lol ASMR*
Jill Talk literally listening to this for asmr but multitasking to get some of that brain power going too.
MIT
bitch I didnt ask
MIT
Thanks for featuring this class. I need to know algorithms for my current computer programming learning in Python, Artificial Intelligence etc. I listen to these classes to reinforce my current knowledge.
best time to be alive, you can study from best teacher across the world for free at home.
8 modules in this course:
1. Algorithmic thinking: Peak finder
2. Sorting and trees: Event simulation
3. Hashing: Genome comparison
4. Numerics: RSA encryption
5. Graphs: Rubik cube
6. Shortest path
7. Dynamic programming: Image compression
8. Advanced topic
I feel like the content of MIT and lower ranked institutes is the same. The difference is in the students and the delivery of the explanations.
looking at this course after taking design and analysis of algorithms course as a senior this fall 2021 semester with my university. hopefully watching this video helps solidify the concepts learnt.
15:30 specific problem -peak finder
28:10 divide conquer algorithm
34:30 how usefulness depends on size of input data
36:15 peak for 2d
I always dreamed on going to MIT... and now I can do it from home. Thank you :)
I think the professor have not explained the implication of the 1D and also 2D algorithm.
The key is that for any 1D or 2D array there is a largest element(>= any other) and thus must have a peak. (Theory 1)
However, in an array the sub-array's peak may not always its parent's peak if the largest element is on the edge and has a bigger element next to it. (Case 1)
To find a peak in a sub and at the same time avoid Case 1, we must choose a larger side to ensure that even if the largest value is on edge, it is still larger than its boundary in the parent array.
This is also why in 2D search we want a global maximum. We want to ensure that even if the max is at the boundary, it is still larger than the any element in the boundary column containing the global max we found. And thus the global max in the larger side must contain a peak.
thanks for uploading the recitations, I watched this series long time ago and now I get to review it with recitations!
once you get it, that moment will make your day.
Thank you MIT for sharing videos! They are incredibly useful.
What a comforting teacher...
Very clear and straight remarks
The questions of life are very import
We will find how defaults defaults are l8fe
I feel super grateful to the free content that MIT has put up
Same here. It's incredible. I've been watching lectures, revising the lecture slides, making notes, etc. All for free.
All these students are the ones who say they did not study for exam and still top the class.
Well, to be fair, you don't have to study FOR the exam to ace it. You just have to study period and that's what they do
Doubt it, MIT courses are hard as hell
@@joeyGalileoHotto MIT courses are quite easy....But our UN****** education system quite hard(Deep fact).
@@alphaco3868
Can I rely on this course to learn data structures and algorithms
I mean we are in 2020 and this videos are before 7 years
@@mohammedjassim9250 Yes you can...I recently watched this lecture __ruclips.net/video/JPyuH4qXLZ0/видео.html This lecture is about algorithm analysis. Our professors also used this technique. You understand what i'm taking about. If you complete algorithm and data structure from here & understand concept fully. I'm sure you'll find it very helpful no doubt. They also provided textbook that is available for free(.pdf)
Excellent camera work. Wish this camera guy could film all lectures -- he handles the lecturer's walking really well.
I wish my University lecturers were so well spoken and were more interested in getting me interested in conveying excitement about the topic instead of just getting through the lecture...
there is a boundary condition.if you n/2 is the left most element then there is no a[n/2-1].
If n/2 is the leftmost element then we can just check with simple greater than algo...need to apply any methodology
When finding the middle column, the equation should be j=(m+1)/2. If j is a floating-point number, then it needs to be rounded up or down.
i swear youtube recommended me this. i was only looking for how to spell mathematics.
I needed this. My current algorithms class is a joke.
For the 1st 6 weeks professor, on average, cancelled class once a week (we have classes 2x a week). Half of the 1 class we meet up, is him telling stories/making jokes. Then the whole corona virus thing hit, and we are now doing lectures online, where the first week he cancelled a class again. His current online lectures all choppy because I think he is using a laptop webcam to record him writing on a whiteboard. The quality of the videos are 240p at best, and can drop even worse at times. Looking at the syllabus, we are 4 weeks behind schedule.
hey isn't this a little outdated? do you still find it relevant to study from this series? It was created in 2011.
now this is called real work
Complet guidance from cyclomatics complexity , just root work , great sir 🔥👌
If the peak that particular peak dataset is always a pyramid shape then its constant time.
Check 2 points on either end of the array and then find the intercept of the two resultant lines, that is x intercept of the two lines is the index that the peak is in the array.
The limits of this approach are the precision of the numerical calculations.
Great lecturer. Thank you, professor.
computer science undergrad major core courses: calculus, linear algebra, discrete mathematics,
data structure, assembly language, operating systems, and algorithm,...etc.
i liked the whole playlist because every video has subtitles that didnt created by Google but someone listened to the lectures himself. so i can follow the lecture more efficently with the subtitles thanks for that. love the lesson, hate the math part of it.
Thank you for this learningful class, you can explain the problems very clear in the algorithms
I really appreciate this free of cost content, because people who were not much concerning about their academics during their college days also get a chance to revisit(of course I am not speaking about MIT alumni)
This Answer is for all those people who are confused by the 'peak finding problem'.
1. Here we are not finding THE PEAK, We are finding A PEAK.
2. For this problem the definition of A PEAK element is that its adjacent LEFT and RIGHT elements are smaller than the element.
so for ele[ i ] to be a peak ele [ i ] >= ele [ i - 1 ] & ele [ i + 1]
3. Also we are assuming that the value of elements is increasing from LEFT to N/2 and Again decreasing from N/2 to RIGHT.
So, This algorithm makes sense.
Also note that this is not a version of binary search algorithm. It is divide and conquer strategy.
For binary search we require a sorted array.
Assuming the array is sorted in ascending order, The last element is definitely going to be A PEAK element so no need of this dividing and finding the peak.
This was a TOY PROBLEM.
Thanks.
assuming the array in sorted order then why professor draw peek in middle ?
we are assuming that the value of elements in array is increasing from LEFT to N/2 and Again decreasing from N/2 to RIGHT.
Fuck off
Vikas Choudhary do you mean ele[n/2] >=ele[1] & ele[n/2] >=ele[n]?
wait a sec, if value increaes till n/2 and then descreases from n/2..then won't e[n/2] is one of the peak for every input?
Great to see free course from MIT!
FYI.... For One Dimensional Array the Binary Search algorithm works only on a Sorted Array.
Thank you MIT for sharing this.
Wow, there's 47 of these. I love the internet :3
It's so cool that these top schools release courses like this one online free of charge. I may not get a chance to go to MIT
Someone clarify this for me: In the 2d peak finding algo, the complexity is O(NlogN)? After choosing j, we find the global max, which takes O(N) time, traversing the whole col. After that we choose the left or right half depending on which is greater on the same row than the global max of current col. In best case, the global max is greater than both of them, here we exit the loop as peak is found. The second step takes O(logN) time as we are eliminating the search space by 1/2.
Complexity is O(nlogm), where n = height of rectangle, m = width. Values of N and M could be wildly different, even consider where N < logM for some inputs N and M, so it’s important to differentiate. As for the recurrence relation, finding the global max in column j takes O(n) time. There isn’t any operation that takes logM times, it is the O(n) operation of finding the global max that is repeated logM times. The O(n) operations are executed in the exact same way a binary search is. Hope that helps
this is my great fortune to have a class of MIT, its amazing to particpate in it☺☺
I'm not understanding the complexity of the 2D peak algorithm (@52:41). Shouldn't I have: log2(m)+1 terms "O(n)" ?
For example with m=2: T(n,2) = T (n,1) + O(n) = O(n) + O(n). Here I have 2 O(n) that is log2(2)+1.
Thanks in advance if you can help me to figure this out.
Классный канал, от подачи не устаёшь
Took some time to realize why he found the global max. Try for A = [ [12, 15, 16, 27], [10, 8, 3, 28], [9, 7, 25, 26], [8, 6, 2, 3] ]. In first iteration j will be 2nd column of A, in next iteration j will be 3rd column of A, and in final iteration j will be 4th column of A. Answer = 28. Realize, as we move from 2 --> 3 column, the condition checked was 15(max of 2nd)
In the 2nd attempt of the 2D version, this algorithm works to find *a* peak, not the maximum element in the matrix, correct?
When I read the comments I saw there were some people with a confusion about peak finding.I think It's better to know that he was not finding the "Highest Peak". He was just finding "A Peak". correct me if I am wrong.
i think so
When you go through confusion and feeling that, it can't be log n and prof is wrong, then, watch again, edge cases 17:37 of video. It will clarify the problem statement.
heck yeah, saw this guy and took the nearest clean nb and started making notes
Can rightmost or leftmost element may be a peak in 1D array ?
If so , then I think all the arrays have minimum one peak element .
Help me to take decision
Can we find the peak element in column just like we had done in 1d array through binary search instead of finding global maximum and then do the same procedure of binary search between the columns as shown..
No. We need to ensure that the element found in column j+1, (find_row(j+1), j+1), is greater than the adjacent element in the previous column, (find_row(j+1), j)
[otherwise we could have (find_row(j), j+1) > (find_row(j), j) > (find_row(j), j-1) and (find_row(j+1), j) > (find_row(j+1), j+1) > (find_row(j+1), j+2), so that we get caught in a loop moving between columns j and j+1].
For the max, this is guaranteed by the fact that
(max_row(j+1), j+1) >= (max_row(j), j+1) [since (max_row(j+1), j+1) >= (i, j+1) for all i]
> (max_row(j), j) [since this is how we choose to switch from column j to column j+1]
>= (max_row(j+1), j) [since (max_row(j), j) >= (i, j) for all i]
This is a wonderful course the only minor problem is some of the recitation videos and notes are missing. Thanks, MIT!!
This is a gem of a course
This lecture made me reconsider going to Computer Science. Hard af
this guy seems like a really experienced teacher...
I wish the colleges and universities here had teachers like this man. They all have thick accents and do not explain in detail like this instructor. No wonder MIT has the rep it has. wow.
0:25 thats my dream sentence to say in the future
One good thing about watching this at home is you don't have to answer the instructor's questions for couch cushions. =)
Agree
47:55 How do we find the global maximum. 48:27 What allows us split the columns in half?
My master said : " GREAT TEACHER WILL ALWAYS MAKE GREAT STUDENT " So , i'm not great student , and i blame to my teacher for this .. but this is outstanding knowledge for free , wow .... Peace from Surabaya , Indonesian ..
Thank you MIT. Great lecture
Hey! It's been 5 years since no one had like your comment I'm the first to like 😂😂
This channel is a gold mine!
every student needs this kinda cool prof.
everynight I'm listening to this as sleep music.
Thanks MIT.
"Peak finding" is just a fancy name for finding the max value of an ANALOG continuous function. It won't work for a DIGITAL or discrete function. Only useful for processing some things.
Its literally the same concept for a DT function? Look into Signals / Signals and Systems it will teach you more about these subjects.