You guys are really doing well I appreciate it and looking forward to work with you both of you ..I wish I could present in Bangalore but we will see........
When the interviewer is not right, there is no point in taking the feedback. Companies are interested to hire the candidate, the Right interviewer will not confuse you. He/she will always help you with hints. Interviewing is an art! Interviewers shouldn't be arrogant. If the interviewer is confusing you that means they are not the right colleague to work with. Every member is working toward building good products for orgs and that requires collaboration, discussion & not confusion!
For the machine question: 1 -> machine does not work 0 -> machine works precision: out of machines model predicts not working , how many are right, recall: how many not working machines model correctly identifies, So metric to be chosen is recall since it is more important to identify all non working machines (so they can be catered to) much more than correctly identifying that a machine is not working when we say it is not working, working machines classified as non working can be tolerated, the reverse cannot be
For threshold, if 1 -> working, then if prob >= 80 , then predict as 1, so threshold is 80 or much higher say 90, that is unless you are damn sure its working you classify it as not working, because identifying all non working machines is important
Edit: Sudhanshu sir answered it at the end of the video. Here's the thing. For the parking project, you don't even need to use a neural network. This is Ch1 of DIP. You take an image of an empty parking lot. Then take an image every 10 seconds or so and subtract it from the first. The patches where the pixel is 0 or near 0, the space is empty. If its not, then there is something occupying the space.
1. Machine is working fine or not? we can check with business what is the most valued thing, either false positive or false negatives. Since this use-case is health-care, then we focus on False Negatives and hence we go with Recall.
Since in this case false negative is what it says our machines health condition is good but prediction is saying that it's not good. So we can go and check that particular machine. In case of false positive , in actuality machine health is not good but prediction is saying that our machines health is good. It's a danger , since we will not care that machine since it says machine conditions is good. So it's a point where we need to think. So we need to reduce false positive and hence we can go for increasing precision.
This guy is new to ml and ds area.. The questions are flooding him. He doesn't want to stop answering but he clearly didn't understand the questions.. But he tried though..
After a time in the life people should not ask for interviews. Rather they should offer after knowing themselves how someone works in the current organization.
15.59 Recall says out of posive class how many correctly predicted , precision says out of posively tagged ( machine failed ) how many actually failed here Recall is more important as we dont want to take risk about machine going to fail as not failing can use F1 or F beta
If the Model is slow and huge in parameters my suggestion would be to first do mixed precision training using Apex and do inference in fp16 also we must choose the right hyperparams for the model as fp16 consumes less memory low cost gpu can be used and could be made more implemetable. Some people are saying deploying models aren't a part of data science but it is a part you should come up with algorithms to solve them instead of throwing money on high compute GPU. Also for parking problem shouldn't have gone to NN It is like using sword to cut potatoes
Hi Krish, I really like your videos....they are worth watching. Just a small request, could you please do the mock interview for an experienced candidate?
Answer ,1 : if we dont want to miss fault machine even when we need to check non fault machine for maintenance we will go for recall . And if dont want all fault machine are correctly classification as fault then precision
Since in this case false negative is what it says our machines health condition is good but prediction is saying that it's not good. So we can go and check that particular machine. In case of false positive , in actuality machine health is not good but prediction is saying that our machines health is good. It's a danger , since we will not care that machine since it says machine conditions is good. So it's a point where we need to think. So we need to reduce false positive and hence we can go for increasing precision.
Good feed back Krish. Even i have the good knowledge but i am not able explain clearly in interview . due to which im not able to clear interviews. I am working on the same.
Sir, I am from the marketing campaign management, campaign analysis sort of profiles, looking for the ml projects for customers segmentation, customer churn, lead score generation, basket analysis sort of projects, would highly appreciate if you could suggest me some links
Can i take the internship at iNeuron without attending the online courses? I just need assistance on deploying end to end model, i have the theoretical and mathematical knowledge.
Hello Sir, I have a commerce background (CFA level 2 cleared and PG Diploma in Finance is going on), but over the past few month i have developed a interest in Data Science and i am very keen to enter into the industry. I do wanna go abroad and enroll for a Data science. But i have doubt if i can enter into univiersity courses given my commerce background. Thus, i wonder if you help in that area and also which universities i can apply to. Thanks
You guys don't even give the answer that you want to hear. You are asking questions which doesn't have definitive answer and confusing the candidate. Let's suppose !
If you expect a single guy to build a model and deploy it with a business owner mindset. Why would he work for you? He would have done his own business and would not look for such interviews. 😅 Seriously this guy has put a lot of effort and asking such questions to him is gonna demotivate him. I feel pitty for him.
I can see two possibilities. One is where the company is immature in its data science and expects way too much of a data scientist, the other is that they want to see you are quick on your feet, know a bit about deployment and would be able to at least assist with productionization. It doesn't sound too difficult to get some deployment expertise, just playing around with AWS as an example.
@@krishnaik06 You are right but when an interviewer ask this. It doesn't seem to get basic understanding of deployment. Like Sudhanshu in this video asked for proper deployment plan. I understand the candidate may or may not provide appropriate reply but the dilemma of being rejected stays there post interview after this question. :)
I do agree .. but here in usa .. i had to implement a machine learning algo ... productionalize it .. n deploy it to cloud .. P.S this was an intern project
He doesn't know the ethics of interview. The way he does an in interview is like what he wants to hear . And due to this, this type of guy lacks to think in different angle of solving the problem. This is an overhype
Please don't think Data Science is simple it's not .Even i know more than the chayan in just my 3 months but after 4 years of working i can say it's tough .You need to go all in Note:I m not trying to make you sad or demotivate you .I m just giving my personal experience
@@sharan9993 start with Great learning statistics for data science course it's free on RUclips after that do Python for beginners then Data science course by Great learning .
Keeping machine learning knowledge aside this guy really need to improve his presenting and speaking skills. If I was the interviewer I would get a bit irritated from the start.
@@niveditaparab6772 I haven't said anything about selecting or rejecting. However, soft skills matter a lot in interview. Interviewer has to take decision based on how you have presented your skills in those 1-2 hours.
43:30 is the most important feedback of the interview. I hope u all see that part ...
You guys are really doing well I appreciate it and looking forward to work with you both of you ..I wish I could present in Bangalore but we will see........
Can I get to learn like he did ? Internship while working ?
When the interviewer is not right, there is no point in taking the feedback.
Companies are interested to hire the candidate, the Right interviewer will not confuse you. He/she will always help you with hints.
Interviewing is an art!
Interviewers shouldn't be arrogant. If the interviewer is confusing you that means they are not the right colleague to work with.
Every member is working toward building good products for orgs and that requires collaboration, discussion & not confusion!
i think you've got the wrong idea about arrogance buddy.
For the machine question:
1 -> machine does not work
0 -> machine works
precision: out of machines model predicts not working , how many are right,
recall: how many not working machines model correctly identifies,
So metric to be chosen is recall since it is more important to identify all non working machines (so they can be catered to) much more than correctly identifying that a machine is not working when we say it is not working, working machines classified as non working can be tolerated, the reverse cannot be
For threshold, if 1 -> working, then if prob >= 80 , then predict as 1, so threshold is 80 or much higher say 90, that is unless you are damn sure its working you classify it as not working, because identifying all non working machines is important
I am commenting as I watch the video, so for ratio, class weights would be 1:9, one class(100) * 9 = another class (900)
I would love to appear for a mock interview !
bro don't study. you won't be able to make it in the industry, trust me. you'll have to leave india to progress in this industry
Edit: Sudhanshu sir answered it at the end of the video.
Here's the thing. For the parking project, you don't even need to use a neural network. This is Ch1 of DIP. You take an image of an empty parking lot. Then take an image every 10 seconds or so and subtract it from the first. The patches where the pixel is 0 or near 0, the space is empty. If its not, then there is something occupying the space.
please bring more interview videos like this sir for 2023 data science preparing students
1. Machine is working fine or not?
we can check with business what is the most valued thing, either false positive or false negatives. Since this use-case is health-care, then we focus on False Negatives and hence we go with Recall.
Since in this case false negative is what it says our machines health condition is good but prediction is saying that it's not good.
So we can go and check that particular machine.
In case of false positive , in actuality machine health is not good but prediction is saying that our machines health is good.
It's a danger , since we will not care that machine since it says machine conditions is good.
So it's a point where we need to think.
So we need to reduce false positive and hence we can go for increasing precision.
This guy is new to ml and ds area..
The questions are flooding him.
He doesn't want to stop answering but he clearly didn't understand the questions..
But he tried though..
That is a typical scenario when you Are not is actual research...just watching and reading does not help.
After a time in the life people should not ask for interviews. Rather they should offer after knowing themselves how someone works in the current organization.
Precision or Recall: basically depends on whether FN is important or FP.
Recall intended to reducde false negative rate precision intended to reduce false positive rate
Roadies for Data Science !
Agreed
Really??? What is original here??? Like to know that
😂
😂😂
15.59 Recall says out of posive class how many correctly predicted , precision says out of posively tagged ( machine failed ) how many actually failed here Recall is more important as we dont want to take risk about machine going to fail as not failing can use F1 or F beta
had good exp.with the virtual Mock Interview.
wish to have more and more for our knowledge.
If the Model is slow and huge in parameters my suggestion would be to first do mixed precision training using Apex and do inference in fp16 also we must choose the right hyperparams for the model as fp16 consumes less memory low cost gpu can be used and could be made more implemetable.
Some people are saying deploying models aren't a part of data science but it is a part you should come up with algorithms to solve them instead of throwing money on high compute GPU.
Also for parking problem shouldn't have gone to NN
It is like using sword to cut potatoes
Hi Krish, I really like your videos....they are worth watching.
Just a small request, could you please do the mock interview for an experienced candidate?
Answer ,1 : if we dont want to miss fault machine even when we need to check non fault machine for maintenance we will go for recall . And if dont want all fault machine are correctly classification as fault then precision
Since in this case false negative is what it says our machines health condition is good but prediction is saying that it's not good.
So we can go and check that particular machine.
In case of false positive , in actuality machine health is not good but prediction is saying that our machines health is good.
It's a danger , since we will not care that machine since it says machine conditions is good.
So it's a point where we need to think.
So we need to reduce false positive and hence we can go for increasing precision.
Good feed back Krish. Even i have the good knowledge but i am not able explain clearly in interview . due to which im not able to clear interviews. I am working on the same.
Sir, I am from the marketing campaign management, campaign analysis sort of profiles, looking for the ml projects for customers segmentation, customer churn, lead score generation, basket analysis sort of projects, would highly appreciate if you could suggest me some links
Felt like my first interview 😮😊
Bro you are great...
Hero
I thought the interview is the view between two people. You may rename it Viva-Voce.
Great work sir....
If a fresher goes on an interview for Data science or Machine learning are there any technical rounds or only Basic theoretical stuff would be asked?
Can i take the internship at iNeuron without attending the online courses? I just need assistance on deploying end to end model, i have the theoretical and mathematical knowledge.
Hello Sir,
I have a commerce background (CFA level 2 cleared and PG Diploma in Finance is going on), but over the past few month i have developed a interest in Data Science and i am very keen to enter into the industry. I do wanna go abroad and enroll for a Data science. But i have doubt if i can enter into univiersity courses given my commerce background. Thus, i wonder if you help in that area and also which universities i can apply to.
Thanks
Is there any website or apps for giving mock interview? Please answer
You guys don't even give the answer that you want to hear. You are asking questions which doesn't have definitive answer and confusing the candidate. Let's suppose !
Thanks Krish
How can I give this interview?
If you expect a single guy to build a model and deploy it with a business owner mindset. Why would he work for you? He would have done his own business and would not look for such interviews. 😅 Seriously this guy has put a lot of effort and asking such questions to him is gonna demotivate him. I feel pitty for him.
I can see two possibilities. One is where the company is immature in its data science and expects way too much of a data scientist, the other is that they want to see you are quick on your feet, know a bit about deployment and would be able to at least assist with productionization. It doesn't sound too difficult to get some deployment expertise, just playing around with AWS as an example.
Sometimes its not about implementation, its about how you can pitch ur idea :)
@@krishnaik06 You are right but when an interviewer ask this. It doesn't seem to get basic understanding of deployment. Like Sudhanshu in this video asked for proper deployment plan. I understand the candidate may or may not provide appropriate reply but the dilemma of being rejected stays there post interview after this question. :)
I do agree .. but here in usa .. i had to implement a machine learning algo ... productionalize it .. n deploy it to cloud .. P.S this was an intern project
hyy krish sir and sudhansu sir
Can someone please tell me whether all interns are going to have a virtual interview like this one?
yes u just need to ask
No, not required , but if you want you can drop a mail to krish , he will arrange.
@Krish and @Jagdish thanks for replying!
Aisa thumbnail pe ranvijay laga interview lega roadies ka
I am not an Ineuron Student but I would like to participate in these kinda interviews, i wanna evaluate myself!!, is it possible!?
@krish Naik please coment about if approach of candidate is correct for questions you ask . we want litsen rite answers
Over Powered !! = OP
Can a BSc statistics graduate can apply for a job after complition of data science course?
Too much atitude. You can even see that on red shirt guys face.
He doesn't know the ethics of interview. The way he does an in interview is like what he wants to hear . And due to this, this type of guy lacks to think in different angle of solving the problem. This is an overhype
@@subho2859Truly said bro. To be honest, he is the only reason why I unsubscribed Krish Naik.
I have wasted my 3 year in preparing for govt job. Is it possible for me to do career in ML/AI and will company accept me as their employee?
Yes u can
Govt job is a trap in india
Please don't think Data Science is simple it's not .Even i know more than the chayan in just my 3 months but after 4 years of working i can say it's tough .You need to go all in
Note:I m not trying to make you sad or demotivate you .I m just giving my personal experience
@@sachinanbhule8955 can u guide me I am a fresher. I hav learnt few things I am more confused than ever.
@@sharan9993 start with Great learning statistics for data science course it's free on RUclips after that do
Python for beginners then Data science course by Great learning .
Candidate telling wrong about Precision and recal
Go slow .. take time .. explain in proper way
I hope i dont get interviewed by indian people haha, their brilliance is intimidating
Hello Sir
Keeping machine learning knowledge aside this guy really need to improve his presenting and speaking skills. If I was the interviewer I would get a bit irritated from the start.
@@niveditaparab6772 I haven't said anything about selecting or rejecting. However, soft skills matter a lot in interview. Interviewer has to take decision based on how you have presented your skills in those 1-2 hours.
Interviewer is toxic.