Machine Learning Methods – Computerphile




[ad_1]

We haven’t got time to label things, so can we let the computers work it out for themselves? Professor Uwe Aickelin explains supervised and un-supervised methods of machine learning.

Silicon Brain: 1,000,000 ARM Cores: https://youtu.be/2e06C-yUwlc
Brian Kerninghan on Bell Labs: https://youtu.be/QFK6RG47bww
Could We Ban Encryption?: https://youtu.be/ShUyfk4QB-8
Computer That Changed Everything – Altair 8800: https://youtu.be/6LYRgrqJgDc

http://www.facebook.com/computerphile
https://twitter.com/computer_phile

This video was filmed and edited by Sean Riley.

Computer Science at the University of Nottingham: http://bit.ly/nottscomputer

Computerphile is a sister project to Brady Haran’s Numberphile. More at http://www.bradyharan.com

Source


[ad_2]

Comment List

  • Computerphile
    December 1, 2020

    Nerd

  • Computerphile
    December 1, 2020

    this guy is my professor….

  • Computerphile
    December 1, 2020

    Thank you for explaining machine learning in an understandable way! I haven't seen any videos explain it this well.

  • Computerphile
    December 1, 2020

    Anyone have the full TKS playlist? I'll thank you in advance! T K S! T K S! KNOWLEDGE IS POWER. SOCIETY IS POWER. UNITE. THE KNOWLEDGE SOCIETY THE KNOWLEDGE SOCIETY THE KNOWLEDGE SOCIETY THE

  • Computerphile
    December 1, 2020

    Anyone else here from TKS?

  • Computerphile
    December 1, 2020

    This guy finally allowed me to understand the concept. it reminds me of that time when my lecturer talked about toilet seats. You want it to be hard, but if too hard, it becomes brittle and it breaks when you sit on it.

  • Computerphile
    December 1, 2020

    Great video, thanks for sharing!

  • Computerphile
    December 1, 2020

    The only video that helped me with understanding ML for my university assignments 😅👌🏾😍😍! Thanks a lot

  • Computerphile
    December 1, 2020

    So the "man in the loop" is 'considering' the human's input? Constrasted with supervised where the user overrides the program's decision, man in the loop treats the human input as a weighted factor that is part of the overall algorthim? The human factor influences, but doesnt effect a change?

  • Computerphile
    December 1, 2020

    Very good, but please, put some subtitles

  • Computerphile
    December 1, 2020

    Subtitles would be great on this video :p

  • Computerphile
    December 1, 2020

    K-means clustering ftw!

  • Computerphile
    December 1, 2020

    "…I wouldn't be researcher if we'd finished with it!" lol loved that phrase, the very nature of curiosity and the willing of discovering the world!

  • Computerphile
    December 1, 2020

    Is that comic sans?

  • Computerphile
    December 1, 2020

    please someone can contribute with english subtitles?

  • Computerphile
    December 1, 2020

    How does the machine come up with functions to try?

  • Computerphile
    December 1, 2020

    machine learning using c# for beginners by Latasha Morgan is the best ML book out there.Get it.It helped me loads.Lots of coded examples and explanations.

  • Computerphile
    December 1, 2020

    I love it!

  • Computerphile
    December 1, 2020

    all computer science introductory lectures should be like this!

  • Computerphile
    December 1, 2020

    My cardiac surgeon ( Dr Nick Hendel ) sat with me and entered data from tests and factors and ran the program. He said it informed him of what to do and avoid and whether I'd survive.
    Lying in the cardio ward watching the surgeon with an HP or TI programmable calc is pretty different.

  • Computerphile
    December 1, 2020

    The computer aspect is interesting.  However I am appalled with how easily you slip into use as a medical tool.  What if your child was a genius, but because he didn't fit a prescribed pattern got labeled as a dullard.  Please don't turn doctors into robots without interest in the human patient.

  • Computerphile
    December 1, 2020

    I think I just found what I want to study for my PhD.

  • Computerphile
    December 1, 2020

    Isn't this the way babies or toddlers learn and "label" objects themselves as well?

  • Computerphile
    December 1, 2020

    this field has scary applications

  • Computerphile
    December 1, 2020

    Why would you call clustering algorithms learning? It's a method of categorizing data, but I see no "learning".

  • Computerphile
    December 1, 2020

    Loved it!!

  • Computerphile
    December 1, 2020

    I find artificial evolution of neural networks to be a very interesting way of finding solutions to problems. I might actually make one for a current problem I'm having, which is sort of related to voice recognition.

  • Computerphile
    December 1, 2020

    I really like these videos.

  • Computerphile
    December 1, 2020

    Its impossible to be a polymath

  • Computerphile
    December 1, 2020

    Stick a Kohonen network against a neural network, job done.

  • Computerphile
    December 1, 2020

    Wow, i've been thinking about a for months. I am really glad to know that there are other people who think about the same thing and even researching it.

  • Computerphile
    December 1, 2020

    Great video

  • Computerphile
    December 1, 2020

    Great explanation. I can see that he knows much more, and he has a way to simplify enough so you can understand.
    I already have some questions about how some token examples to reevaluate the categorization in semi-supervised machine learning would work. Some confidence rating or sth…

  • Computerphile
    December 1, 2020

    I've been looking for a video like this for a long time. Extremely well explained and easily understood!

  • Computerphile
    December 1, 2020

    I find unsupervised learning much more powerful than semi/supervised learning. You are not restricted to the number of classes which is kind of the whole point of data mining. They are just being used for different purposes…

  • Computerphile
    December 1, 2020

    The blocks have Comic Sans on them.

  • Computerphile
    December 1, 2020

    Weirdly, in portuguese, "data" and "dices" are pronounced the same: "dados".

  • Computerphile
    December 1, 2020

    Humans learn everything in a semi-supervised way, am I right? We get some advice from "experts" (our parents, our teachers, etc) but we also experiment ourselves based on observation (in a sense, we build some "metrics" for the quality of everything ourselves, similar to the similarity metrics in unsupervised learning). Does this make any sense?
    BTW the video is extremely interesting! Amazing stuff! 🙂

  • Computerphile
    December 1, 2020

    Anyone bothered by the font on the blocks?

  • Computerphile
    December 1, 2020

    I find it very interesting that all this work has been done to automate things (take humans out of the loop), only to find that the 'probably' best solution is to actually have a human in the loop. 🙂

  • Computerphile
    December 1, 2020

    What, no mention of reinforcement learning? You're missing the most interesting part.

  • Computerphile
    December 1, 2020

    This sounds pretty close to some peoples definition of "consciousness". As in the ability to examine one's own thoughts.

  • Computerphile
    December 1, 2020

    I love these videos in general, but I found the audio very indistinct on this one, which made Prof. Uwe's presentation difficult to follow. I might try running the audio through an equaliser. Thanks for a fantastic resource.

  • Computerphile
    December 1, 2020

    I'm slightly disappointed at the immediate reach to neural networks. While NNs are proving to be very fruitful as modern learning methods, they aren't really a good example for introducing basic machine learning concepts unless you start mischaracterizing or really oversimplifying what nets do. Maybe start with something way simpler and build your way up? I mean you can even just go with perceptrons or some other linear classifier and that's way easier to understand.

  • Computerphile
    December 1, 2020

    Really good speaker.

  • Computerphile
    December 1, 2020

    I find it depressing how primitive these approaches are in terms of "learning". If you have an interactive method it shouldn't just add the new knowledge to the algorithm, but discover the reason the knowledge was missed in the first place and pro-actively test algorithms along the same "missing method" axes. I know that's asking a lot, but come on, it's no longer 1973

Write a comment