An Introduction to Active Learning (Machine Learning)

Active learning is a technique used in machine learning to get good performance by labelling less data.
This video gives and introduction to active learning with an example.
Here are some additional resources for active learning:
Datacamp: www.datacamp.c...
Wikipedia: en.wikipedia.o...)
Active Learning Resource PDF -
burrsettles.com...
Image References:
www.statistics...
www.datacamp.c...
Iris Dataset: www.kaggle.com...
Music:
Acoustic Breeze - Bensound
Slow Motion - Bensound
#machinelearning #artificialintelligence #ml #ai #activelearning #deeplearning #datascience #computerscience #technology #dl #activemachinelearning #activelearningml

[@nasimsonboli3712]

Hi , do you now if there are any resources for active learning in recommendation algorithms or any library that can be useful for this purpose?

[@SambitSatapathy-e4b]

Great job. Nice explanation.

[@sandeepvivek6134]

4:00 Sir what if the algorithm labels that uncertain point incorrectly?? How can this issue be addressed?

[@thebrainycomputer393]

Hi @Sandeep V!
Here, the oracle labels that point/sample. After it is labelled by the oracle, it is used by the algorithm during the training phase.
(As mentioned earlier, the oracle is assumed to have complete knowledge and be able to label all points accurately. An example of an oracle would be an expert in that particular field).
Sorry for any confusion in the video.

[@SAINIVEDH]

So for the second iteration onward, will the algorithm does the labeling on its own for certain important points or will the oracle does the labeling ??

[@thebrainycomputer393]

Whenever you need to label an unlabelled sample, the oracle is required. The algorithm does not know anything about labelling. The algorithm only selects which samples to label. Then, the oracle will be used to label those. Thus, the oracle is required for all the iterations where there is labelling.

[@mahbub.bayzid]

Great work

[@aryansakhala3930]

Akhilesh can you help in a IEEE paper solving ?? It has main part for active learning... Shall i share you ML problem or link ?

[@thebrainycomputer393]

Sure!
Go to the about page of my channel and you will find my email ID (you will be able to view it on computer, not mobile phone). You can contact me through that

[@user-cj8mj3uk4q]

amazing explanation thank u so much!!

[@kingwsd]

Very good explanation, thanks for sharing😁

[@davidhuang7599]

Can you give a brief description of how an algorithm of active learning on a set of data would look in something like Python? I'm struggling with grasping how we can apply these concepts into a code.

[@davidhuang7599]

Wait, nevermind, you have a full article about it. Sorry.

[@thebrainycomputer393]

@@davidhuang7599 Hope you got what you are looking for

[@davidhuang7599]

@@thebrainycomputer393 Hi, it's me again. I have a project where I need to show on a graph that an active learning scenario is better than going through just by "random selection" in predicting something.
I'm given a set of student's grades in Quarter 1 and Quarter 2. The additional information I'll be using is a student's study time. So with these 3 features, I'm wondering what would be the best way to show that an active learning scenario is better than just going by random selection for predicting a student's final grade, the label.
I currently have a "linear regression" prediction coded in Python, which gives me a predicted accuracy using all the data. I'm thinking of using that to express the factors of random selection, but I believe I'll have to code something else for active learning. Overall, I'm just looking for what you think is a good pathway to solve this. Your videos have been very helpful lately, and I'd appreciate this help if you have time, thank you.

[@thebrainycomputer393]

​@@davidhuang7599 Hello! I would be happy to help. Send me an email and we can discuss in detail over email or we can fix a time.
To view my email address do this:
Go the My (The Brainy Computer) Channel Page > About > Scroll Down > Click "View Email Address" (Under the "Details" section and next to "For Business Queries Sub-heading")
You will NOT BE ABLE to view this on TH-cam app on mobile phone
You will be able to view this on a PC/laptop/computer. You will be able to view this on Mobile phone browser when you put "View Desktop site".

[@thebrainycomputer393]

@@davidhuang7599 comment here if you are not able to find the email address.

[@iqraarshad5389]

Good description thanks

[@buvaneswariramanan5010]

Concise and clear explanation! Thank you for the tutorial.

[@malikimoustapha4061]

Hello there, This video realy simplifyed more understanding after reading the Datacamp article about AL.
I have resently read an article on Deep Active learning and I m interrested in it. Please do know any practical (with code) that could teach me how to combine both techniques for future data labeling optimization ? Im new to AL

[@utkarshraj3681]

thanks bro, very helpful and simplified !

[@meghanaraokanneganti9675]

The video was clear, but could you revise it? It is very misleading and I think needs correction

[@thebrainycomputer393]

Thank you for this comment. If you can elaborate where it is misleading and where the correction is needed, I can have a look

[@sumit_incornate368]

Yh iit ki padai iit wake samje 😢