# 14 2026-01-18 Doubt Clearing Session

course: Module 2 — Machine Learning Algorithms
module: Module-2-Machine-Learning-Algorithms
date: 2026-01-18
type: transcript
video_url: https://personal-learn.armco.dev/files/_Recordings/Module-2-Machine-Learning-Algorithms/14_2026-01-18_Doubt_Clearing_Session.mp4

---
[00:11:35] Update. So, just allow me. In the meantime, we can… Uh, I had also given all of you the assignment, as you all had requested. Hope you are trying that assignment.
[00:11:49] So, if you have any questions, any problems, we can begin.
[00:12:13] that assignment is, uh, attached, madam.
[00:12:16] Uh, so I think that for that, uh, Cyn will be able to help.
[00:12:23] Okay.
[00:12:24] Because I had shared the assignment with her. I think she has given the… Uh… Simran, can you please share the link for the Sanyada? Link is there on the chat, you can see. I gave it a couple of days back.
[00:12:38] When all of you had requested for it. And, uh, the assignment is primarily on supervised learning.
[00:12:46] I had also provided a data set along with the assignment.
[00:12:50] So that you all can, uh, try something. Uh, you know, try on a different data set, and then it was based on the.
[00:13:00] Uh, supervised learning techniques. Uh, Sindran, can you please open the assignment here, share your screen, the…
[00:13:21] It's there on your LMS under Module 2.
[00:13:23] Week 8?
[00:13:28] Under Week 8, you can see an ungraded coding assignment on supervised learning.
[00:13:34] Over there, you should be able to download the PDF.
[00:14:34] Well, one of you found it. If not, I'll just show it to you again. Under Gen AI Batch 2 Module 2.
[00:14:39] machine learning algorithms. Under that, there is week 8.
[00:14:54] Yeah, got it. Thank you.
[00:14:59] Ma'am, where is the data set?
[00:15:13] You could find it in, I guess, a couple of
[00:15:14] This is defined in Kegel as well.
[00:15:19] Okay, so we have to download from there, okay?
[00:15:20] Under week 8, the last thing is an ungraded coding assignment on supervised learning. You can just click on it, you can click on this PDF, and it'll download.
[00:15:32] Similar, I have another question, uh, on that LMS, I'm not able to find live sessions for.
[00:15:38] First to third week. Are you planning to upload those sessions as well?
[00:15:42] For what?
[00:15:48] Uh, the live sessions recording, uh, it's not available for first, second, and third week.
[00:15:50] So that will be available under Batch 1, Module 1.
[00:15:54] On 11.
[00:15:56] foundations of AIML, if you go here, get started.
[00:15:59] Okay.
[00:16:00] Uh, week 1, week 2, week 3, you will find over here. That's a part of the first module. Up to week four,
[00:16:06] sorry, week 3?
[00:16:09] So you'll have to go to Batch 1, Module 1.
[00:16:12] So, first 3 weeks was Module 1, which was Foundations of AIML, then we moved to Machine Learning Algorithms, which is Module 2.
[00:16:31] Okay, uh, now if I may request you to please once again, uh… Share the assignment.
[00:16:38] So, I'll just discuss it, what is the problem, and how it is to be done.
[00:16:53] Okay.
[00:17:17] So, I may not have provided India. Diabetes data set.
[00:17:22] Uh, and uh… What we are supposed to do is that in this, I wanted you to apply supervised learning techniques like logistic regression.
[00:17:33] Okay, nearest neighbor. I hope I'm audible to everyone. I'm actually… just a second, allow me.
[00:17:42] MM, you're audible.
[00:17:43] I'm getting a message, uh…
[00:18:03] Which we have covered, like, you can look at random forest and so on and so forth.
[00:18:08] You can use linear regression. Anything that you fancy.
[00:18:13] Uh, the attributes, uh, are primarily numeric in nature.
[00:18:19] And there are missing values, so you need to actually, first of all, identify.
[00:18:25] Those values in the data that are containing null values, which means that the values are missing.
[00:18:32] And then, uh, you need to treat the data. So, the data set, which is the PEMA India diabetes dataset.
[00:18:40] The target variable is the class label, which is having a binary class label, 0 and 1, which is, uh.
[00:18:49] No disease and disease, and I'll request you to please scroll erectile, scroll down a little.
[00:18:55] Then the different attributes that have been provided are something like.
[00:19:01] Uh, the glucose level, blood pressure, thickness of the skin, insulin, number of pregnancies, etc. All these are numeric in nature.
[00:19:11] So, first of all, you have to identify the missing values, as I said, and then treat them.
[00:19:15] Now, to treat the missing values, I haven't provided any method.
[00:19:19] Because I gave it to you as an assignment, you all try different missing value techniques.
[00:19:25] Imputation techniques like. Filling missing values with, uh… Uh, mean, median, mode, or whatever you want to try. So, I haven't specified any, so you please try that.
[00:19:39] Then, uh, some ETA, the standard thing is that whenever we have a data set, we first do pre-processing.
[00:19:47] Like, resolving missing values and all, and then we do some ETA, standard EDA.
[00:19:52] To understand what is the data like. So, uh… So, you first do the shape of the data, look at the different… and the description of the data.
[00:20:05] Then, um, uh, you can actually do some visualizations. For example, you can do histograms.
[00:20:13] As we had discussed yesterday. There is 2 grams could be done on, uh, the different attribute value ranges.
[00:20:21] There could be strip plots also, as we had discussed, so you can actually get the range of values per attribute.
[00:20:28] per class, in this case, there are two classes, so you'll actually have in each.
[00:20:33] Plot of the attribute, two strips, signifying the range of values per class.
[00:20:38] Then you could do box plots, violin plots. Uh, to understand, uh, what is the median, uh, mean value, and what are… what is the variance in the data, what are the outliers, like, which are signified by the whiskers of the data.
[00:20:54] Then you can also do correlation, or the matrix to understand how the.
[00:21:01] Attributes are correlated, whether they are correlated with respect to each other, and which attribute or which set of attributes are.
[00:21:08] Positively or negatively correlated with the class labels. And then you can take the two class levels separately, that is 0 and 1, so you'll understand.
[00:21:18] Which attribute is correlated to plus 1 or 2Z?
[00:21:22] My, uh, two zero value. And you can do other things also in the EDA, like pair plots and all that stuff. I've just written some things.
[00:21:32] Uh, if you could just scroll down a little bit, uh… Uh, then the data preprocessing I already explained to you, you might be required to do some feature scaling.
[00:21:43] Uh, using standard scaler or some other scalar that is of interest to you.
[00:21:48] And, uh, then to perform, uh… Um, building of the classification model, you'll actually need to.
[00:21:58] Split the data into train and test. To split the data, you could do it in 3 ways.
[00:22:04] One, you could do the train test split by trying the different ratios.
[00:22:09] Starting with 60, 40, 70, 30, 80, 20, 90, 10, like that.
[00:22:14] And then choose that particular split. Where the performance of the classifier is coming out to be the best.
[00:22:21] That is one thing. Second is easier than that, decide on some particular value. For example.
[00:22:27] 80-20 and use that only? Third is that you do carefold cross-validation.
[00:22:33] And to do care for cross-validation. What you need to do is that, uh… Uh… you actually need to divide, uh, decide the value of K, and then use that.
[00:22:49] So, these are some of the things that, uh, you… Actually, uh, can do.
[00:22:55] Then, uh… Regarding building of the model, so you can build the models that you wish, like those which are pointed out, three of them, decision tree, KNN, logistic regression.
[00:23:09] And then evaluate the model. When you evaluate the model, you can evaluate using different metrics that you have studied, like accuracy, precision, recall.
[00:23:20] F1 score, ROCA, you see. Uh, and then finally, what you can do is prepare a comparative performance table to understand which.
[00:23:31] Particular classifier is the best one on this data.
[00:23:36] And, uh… then, uh, there can be some analysis that.
[00:23:43] Would be relevant to this problem. That which I have already said, one which, uh, is the classifier that performs best.
[00:23:50] And, uh, uh, then whether recall or precision, what would be important, and uh… I've directly written recall is more important, but why is it more important when we are talking about disease patient?
[00:24:04] And, uh, which model would be most interpretable, and it may be sensitive to feature scaling and all that. So these are some questions.
[00:24:14] That you can think of, but you can think of others also. So this is about the assignment that I had given you.
[00:24:20] Uh, it seems that most of you haven't seen the assignment, actually.
[00:24:25] So I thought that you will… because many of you were asking about the assignment.
[00:24:30] So, I was under the impression that you would have.
[00:24:33] Actually, seen the assignment. So, any questions on the assignment?
[00:24:40] Anyone?
[00:24:45] Um, do we have any deadline for the assignment?
[00:24:49] No, no, no deadline. It was a self-study kind of assignment.
[00:24:54] So, there's no deadline, nothing. It is up to you to do it or not to do it.
[00:24:55] Okay.
[00:24:57] But many of you had actually. Yeah?
[00:24:58] And, uh, when we… when we are submitting, uh, will you be reviewing it?
[00:25:03] No, I will not be doing it, because that's what I had pointed out, that it will be difficult for me.
[00:25:10] Mm-hmm.
[00:25:11] Uh, but, uh, it was, uh… Uh, indicated by many of you that you all will be discussing amongst yourself.
[00:25:18] But what I can do is that in the next doubt resolution.
[00:25:19] Mm-hmm.
[00:25:22] session, if you have any doubts, or you wish to.
[00:25:26] Uh, discuss, uh… Uh, you know, the… Uh, the answers that I can do.
[00:25:33] Maybe. But I will not grade it, or I will not on a… Regular session, I will not discuss it because all of them are already.
[00:25:42] you know, slated out that what are we going to do in that?
[00:25:46] Okay.
[00:25:47] Yeah, madam.
[00:25:48] Okay, uh, one more question, ma'am. Uh, I downloaded the data set from Kaggle, uh, for this.
[00:25:55] Pima, so in this one, like, data set is not raw data, actually, it's, uh.
[00:26:00] It doesn't have missing values and alls. So, is it… is it targeted one, or should I… download the raw data.
[00:26:09] Which is having missing value, then clean up and all those.
[00:26:12] So, it is up to you. You can take any form. Suppose you don't… you want to start with the.
[00:26:19] uh, building of the model and EDA part directly, then you can choose the pre-process data set that you have downloaded and work on it.
[00:26:27] Mm-hmm.
[00:26:28] If you wish to try your hands on pre-processing also, then you can search for the raw data, which also contains the missing values.
[00:26:36] Mm-hmm. Okay.
[00:26:37] And then you can choose that. So, it's purely up to you how much time you have.
[00:26:40] How much you want to do, and all that.
[00:26:45] Okay. Yes, okay, thank you.
[00:26:46] So you will find both the versions.
[00:26:50] Any other questions, anyone?
[00:26:53] The outcome is, uh, we need to analyze our, uh, whatever the model performance overall, our, uh.
[00:27:02] Whether it is comparing performance table, that's all, correct?
[00:27:06] Or is there any other thing we are… to predict.
[00:27:07] Yeah. So, I gave you some exemplary analysis questions also.
[00:27:14] Comparative table will help us to compare which particular… out of which I have given.
[00:27:20] Uh, which particular supervised learning method will work best? You can try more methods.
[00:27:26] Then do the comparison. And then do the analysis, like, why, uh… If any specific classifier is working well, why it is working well.
[00:27:38] on that data. And there are other analysis questions also that I've given. You can also think about others.
[00:27:44] So, what are the exercises that I have given you is a kind of a minimal problem.
[00:27:50] But it just gives you a suggestion to do some self-reflection and think.
[00:27:55] What all you can do if you are given our data.
[00:27:57] Our learning, our understanding, we can able to predict, that's all.
[00:28:04] Right.
[00:28:05] Yeah, yes, yes, exactly, that's the idea. That's why I haven't, uh, written too many questions, but these are some of the things that you can start thinking.
[00:28:11] Yes. Thank you.
[00:28:12] Yeah. Yep.
[00:28:38] Okay, any other questions on the theory or any part that you… Wish to ask, because, uh.
[00:28:45] We keep doubt resolution sessions at some frequency, maybe after a module or.
[00:28:53] After one and a half module and something like that.
[00:28:57] So, uh, anything else?
[00:29:27] Ma'am, so, with all the sessions, we are logically concluding Module 2 today, right?
[00:29:34] Uh, yeah, so, uh, I think I'll request, uh… Simran to pull up the, uh, curriculum. Children, uh, can you please pull up the curriculum?
[00:30:03] Hello?
[00:30:10] Yeah, I think it's coming up.
[00:30:46] Uh, kindly enlarge slightly, please.
[00:30:52] Yeah, yeah. Okay. So, we, uh, please scroll down.
[00:31:06] Uh, no, not so much. I wanted to show where we are.
[00:31:10] Yeah, so we are at this clustering k-means hands-on, uh, is also done.
[00:31:16] Uh, that is at this particular week. Okay. And, uh, then this lasso enrich regularization, though I had covered, uh.
[00:31:26] In the last place, but what I feel is that, uh… Actually, there can be many more interesting things to do in the deep learning and text part.
[00:31:37] Uh, therefore, I'm not going to cover it in my class. We can have a master class on it.
[00:31:43] Maybe the next master class, and we'll start with deep learning.
[00:31:46] Uh, from the next portion. If you are all okay, if you wish that I should cover, because.
[00:31:52] In Lasso and rich regularization. Regularization, I will tell you in the next class, I'll also talk about the.
[00:31:59] Cluster validity indices, which, uh… I haven't yet covered, so I'll keep it in the next theory class.
[00:32:06] I can discuss what is regularization. But it takes up two forms, L1, L2, or lasso enrich regularization.
[00:32:14] And, uh, there is some math and theory behind it.
[00:32:18] So, if you all wish that I should cover that, I will… but what I was planning is that.
[00:32:24] If, uh, you all feel, then we can start with deep learning.
[00:32:30] And, uh, then, um… Uh… add some more portions in the text NLP part.
[00:32:39] So, is it okay, uh… Or you want me to also talk about lasso and ridge in detail, because it will take.
[00:32:47] Uh, if we talk about lasso enrich, then it will take 1 year.
[00:32:51] One theory and one hands-on session.
[00:32:55] And we can move with that deep blood.
[00:32:57] Ma'am, can you have a hands-on session on Lasso and Rich?
[00:33:02] Sorry, I couldn't hear, uh, it. I think there were two people saying something. Can you please come back again?
[00:33:10] Uh, so, Lakshmi, this side, can we have one hands-on session on lasso and race?
[00:33:12] Mm-hmm. You can have, but if you wouldn't have done the theory, then, uh… then probably you may not be able to appreciate what is happening in Lasso in Rich. I can take theory and hands-on both for you if you wish.
[00:33:28] Um… so, I can take both for you. Okay.
[00:33:30] It may be recommending.
[00:33:37] Um… So I can see a couple of you saying, um… You would want to go with lasso-enrich theory and hands-on. Uh, Simran, I request you, can we have a poll?
[00:33:49] Uh, in which the question could be, should we have lasso-enrich theory and hands-on?
[00:33:55] Uh, second option is that, uh… Uh, we have more portions in the deep learning and NLP part.
[00:34:03] In its place, something like that, two questions you can please put on the poll.
[00:34:08] And then, accordingly, uh… We can actually do it.
[00:34:21] In the meantime, Jitu, do you have a question? You have raised a hand.
[00:34:24] Yeah, I have a general doubt regarding this L1 interregularization.
[00:34:31] Mm-hmm.
[00:34:32] Like, uh, I don't know, it's right or not, uh, like, uh, is it, uh, used… is this a regularization is used for optimizing the model?
[00:34:37] Like, making the model, like,
[00:34:42] Something like that.
[00:34:43] Yeah, so actually, regularization prevents the model from becoming, uh… overfitted or underfitted, so it's kind of… so, regularization is used in that way.
[00:34:56] So, uh, this is a technique to reduce the impact.
[00:35:01] Of overfitting, primarily. Yeah, I can tell you briefly what is regularization, but if you all wish.
[00:35:11] We can do L1, L2, that is lass origin detail, and then have a hands-on on it.
[00:35:16] Or we can cover extra portions in NLP part.
[00:35:20] Uh, that's up to all of you. Okay, uh, I think you can, uh… I think… You can do a stop share on the curriculum, Sindran, please, and please post the poll.
[00:35:35] Uh, so that people can indicate what they wish.
[00:35:40] More portions of the deep learning slash NLP or LASO and Witch.
[00:35:48] I think the poll is open now.
[00:36:07] Kindly, uh… Kindly see the poll and respond on it, so that we can conclude on it.
[00:36:44] Uh, so, Simran, can we see the responses as they… grow dynamically the responses of the poll, as people are responding, we can see the options.
[00:36:53] I'll have to end the poll to show the results.
[00:37:01] Uh, okay, what I was saying was that we can dynamically also see how people are responding. That's what I was asking.
[00:37:05] a kind of… Uh, you can see it because you're co-host.
[00:37:08] Not just cartoon, but also want them to…
[00:37:18] Yeah, okay, thank you.
[00:38:04] who have… has everyone who wanted to answer has answered, because I can see not going… people not…
[00:38:11] I mean, we are at 63, it's not growing.
[00:38:21] So, as for this, uh…
[00:38:23] What I can see is that, uh…
[00:38:24] Most of the people, around 73% out of 63 who have responded,
[00:38:31] want to cover extra in deep learning and NLP part.
[00:38:42] So I think mostly everyone has responded, whosoever wanted to respond.
[00:38:46] Though we do have 82 people out there.
[00:38:49] Anyone else who wants to respond, please do it a little quick.
[00:38:59] So, what I will suggest is, uh…
[00:39:01] I think we can end the poll, and you can keep the results of the poll with you, uh, Singra.
[00:39:08] So…
[00:39:09] I'm sorry, I didn't receive the poll.
[00:39:13] Uh, now it has ended, unfortunately.
[00:39:19] Okay.
[00:39:20] I think something. You should be able to see it, though.
[00:39:32] You want me to share the results?
[00:39:37] Yeah, I think… yeah, yeah, you can just share the results, I think so.
[00:39:41] So, I don't know the link of the poll was here, so…
[00:39:46] why you would not receive it, I…
[00:39:49] Anyhow, I think quite a large number of people have actually gone for the option of
[00:39:57] Having extra DL and NLP session, rather than anything else.
[00:40:05] So, what I suggest is that we will not, uh…
[00:40:08] I'm not saying that we'll not do lasso and ridge at all. What we can do is that…
[00:40:13] Lasso in which can be covered in a master class.
[00:40:17] Uh, I'll tell you briefly what is regularization.
[00:40:20] Uh, and the details of these two can be covered in another session in the master session, and I will take some
[00:40:29] more portions in deep learning and NLP, which, uh…
[00:40:33] By saving these two terms.
[00:40:34] Okay.
[00:40:36] So… I'm doing a stop share on this.
[00:40:48] Any other questions? Anyone? Anything else that you want to know?
[00:40:54] Or, um, any… any portion that is not clear.
[00:41:12] So, because, uh, now with these foundations that we have, uh, done, the foundational study,
[00:41:20] Uh, these anyway will be required, uh, with anything that we'll be doing.
[00:41:24] And in most of the NLP and the deep learning and NLP problems, we will be using supervised learning.
[00:41:31] Uh, that will be required. The knowledge of supervised learning will be required, because most of them utilize the concept of supervised learning.
[00:41:41] And, um, obviously other things that we have done.
[00:41:44] Uh, we'll start with, uh, deep learning.
[00:41:47] And, uh,
[00:41:49] Uh, then having covered some…
[00:41:51] things about deep learning, then we can start with the NLP part.
[00:41:55] Uh, and, uh, with the NLP part, what we will do is that we will do some basics of NLP, and then we'll start with
[00:42:03] with different specific feature extraction techniques and all that.
[00:42:14] So I think Lokesh wants a recap of what we did in the last 15 minutes. Location in the last 10 minutes, we didn't do much. We only did a…
[00:42:24] uh… poll.
[00:42:32] So, we just did a poll on whether people would want to have, uh, lasso in reach regularization theory and hands-on.
[00:42:38] or you would, I mean, you all would want more portions, something extra to be covered on deep learning and NLP. This is what we had done.
[00:42:49] And based on the poll, the results are also there on the chart.
[00:42:52] Uh, I think most of the learners wanted to have more portions on…
[00:42:57] NLP and deep learning. This is what we decide.
[00:43:04] Uh, so Lokesh, um, I believe, uh, I discussed a lot of things about the assignment. They are ungraded.
[00:43:12] And you can do it on your own. This is a summary of what we discussed.
[00:43:17] The rest, you can actually have a look at the recording. I just went through
[00:43:20] Broadly over what was required in the assignment, what you're supposed to do.
[00:43:24] Uh, all that, that is what. And it is an ungraded assignment. Uh, it is only for…
[00:43:30] Uh, you know, it is only to be used as a self-test.
[00:43:35] Uh, to see whether you have understood the concepts that have been taught. And that's it broadly, that we discussed.
[00:43:42] Details, you can hear the recording end.
[00:43:44] Find out.
[00:43:46] Any other questions on any of the sessions, anything that we have done?
[00:43:51] So far.
[00:43:56] By the way, I just want to know, I hope all of you would have now started interacting with each other and would have thought about some groups that you would want to form.
[00:44:07] So, are you interacting with each other, and do you have…
[00:44:10] some information about the groups that
[00:44:13] You may be for me?
[00:44:24] Not really, ma'am.
[00:44:31] We are interacting somewhat level on… with each other in the group, but uh… Not the farming group for some assignment or.
[00:44:41] project like that.
[00:44:48] I was saying we have a somewhat level of interaction with each other, but not at the level of forming groups or for, like, assignment or project purpose or something like that.
[00:44:49] Sorry, can you please go over again, uh, what, uh, what you are asking?
[00:45:00] So, I think, uh, uh…
[00:45:04] The way you could interact is via the LMS, uh…
[00:45:09] Because I think, uh, that, uh, uh…
[00:45:12] I also see a comment on the chat that, uh…
[00:45:15] about the WhatsApp group. Actually, as a policy, it was decided right in the beginning.
[00:45:20] that the WhatsApp group will not be formed.
[00:45:23] Uh, so in that, uh, it's a policy, and you can discuss with Futurents in more details. I think there was a…
[00:45:30] long discussion about this in the first, uh…
[00:45:34] first one or two, like, uh… I think in the orientation part.
[00:45:40] Uh, so that is a thing.
[00:45:41] Perfect.
[00:45:42] Uh, just to add on, there's a discussion forum on the LMS.
[00:45:44] So you guys can discuss about forming groups over there.
[00:45:59] So I think Simran has, uh, responded to some of the queries.
[00:46:03] So, pergroup strength, what is there? I will confirm to you, I will just have a look at the final enrollment.
[00:46:12] And then I'll be able to confirm the number of people that may be there in the group.
[00:46:17] Uh, that I will confirm, but in the meantime, I strongly suggest…
[00:46:23] Uh, that, uh, you please try to interact and, uh, try to…
[00:46:28] see how best you can, uh, you know, form the groups based on your expertise, your geographical location.
[00:46:36] or anything else regarding that.
[00:46:38] Okay, uh, so I think there's a question by V that can we get a timeline of any graded assignment?
[00:46:45] So, right now, I'm not given you any graded assignment.
[00:46:49] And I've also not given you any graded test.
[00:46:52] So, any question that has been… questions or quizzes that have been given, or the assignments are all ungraded.
[00:46:58] But if you all wish, we can do that also. We can do a graded assignment, if you wish.
[00:47:03] In that case, you all have… will have to attempt it at the same time.
[00:47:07] And then we can have a quiz or a…
[00:47:11] assignment of that kind, and that… the grade or the marks can be counted in your final marks.
[00:47:17] So, do you all want to have a graded quiz or an assignment?
[00:47:30] Yes.
[00:47:31] Maybe some other people also can respond. Would you like to have a graded quiz or assignment?
[00:47:32] Okay, thank you, ma'am.
[00:47:33] Yes.
[00:47:35] Perhaps just have a poll again.
[00:47:36] Yes, ma'am. Maybe assignment would work.
[00:47:39] Sorry?
[00:47:44] Okay, so I think, uh…
[00:47:45] Uh, yeah, graded assignment would work.
[00:47:46] Because I can see almost equal number of yeses and nos.
[00:47:51] So, once again, Simran, I'd like to request you to please put a poll.
[00:47:55] Uh, with two questions, graded…
[00:47:58] assign, uh, graded quiz.
[00:48:01] slash assignments.
[00:48:03] Uh… so you want to go with, uh, graded quiz and assignments, and the other option you don't want to go.
[00:48:12] With graded quizzes and assignments. Maybe you can put up that poll?
[00:48:15] And then all of you, please respond to that.
[00:48:18] I think there was, in the meantime, a question about the cluster validity in DICES. Yes, I will be covering them in my next theory class.
[00:48:27] I will be covering different extrinsic, intrinsic, all different kinds of…
[00:48:31] common and popular cluster validity indices, I will be covering.
[00:48:38] Ma'am, the tax, I missed out in emails, can you forward.
[00:48:45] Sorry, what are you saying? Which Excel sheet?
[00:48:46] it to me again.
[00:48:51] Uh, the course. Gene course syllabus.
[00:48:54] So, I…
[00:48:55] Weeks-wise listed as…
[00:48:56] Yeah, yeah, so I think Sindran already pointed out the location, so it must be there on the LMS. You can please pull it up.
[00:49:04] Okay, sure.
[00:49:15] Uh, Simdan, uh, I hope you are putting up the poll. We are waiting for that.
[00:50:20] Uh, ma'am, do we have an option where, you know, we can have a couple of assignment which is ungraded, and then.
[00:50:25] Eventually, when we are sort of familiar with the concept, then we can have the graded one.
[00:50:32] So, see, the two options are already there, that you have ungraded or you have graded.
[00:50:37] Uh, what you are saying is you have ungraded and later have graded, something like that?
[00:50:44] Yes, ma'am, yeah.
[00:50:45] So, at what point would you like to have a graded assignment? Can you please, uh, clarify a little bit?
[00:50:51] We can have few, uh, which is ungraded, like, I think one you have already given.
[00:50:54] Mm-hmm.
[00:50:56] So, uh, because… We have done more of a learning, but I think in one or.
[00:51:02] 2 or probably 3 assignments will understand how to apply this concept.
[00:51:07] Okay, so…
[00:51:08] So, you'll be the right… yeah. Sorry.
[00:51:09] Uh, ma'am… Can you suggest something?
[00:51:10] Maybe one person can speak, uh, just let the first person finish, then please come in.
[00:51:16] Yeah, okay, can you please finish, Deepa?
[00:51:19] Yeah, I was saying that I think you are the right to, uh, sort of take the decision, but… a few ungraded assignment, and then later on, graded.
[00:51:26] Okay, yeah. See, it's not my decision, I'm asking all of you only. That's up to your batch of learners.
[00:51:35] how you want to go ahead.
[00:51:36] Uh, we can go with any of these options.
[00:51:41] Okay, so Nali, did you want to say something, or who was saying the next person whosoever was speaking?
[00:51:49] Okay.
[00:51:50] Sorry. Uh, ma'am, actually, even I agree with the previous person.
[00:51:57] Like, we do want graded assignment, but prior to that, if you can give us some, uh, like, examples where we can practice before our graded assignment.
[00:52:07] And it doesn't have to be at a certain point.
[00:52:10] You can decide that, like, next week, you can give us the ungraded, and then the week after, you can give us the graded assignments.
[00:52:20] The only thing we need is for practicing purpose.
[00:52:23] Yeah, uh…
[00:52:25] Yeah, you all need for practicing that… that is what I understand.
[00:52:33] Okay, who else wants to say something? Lakshmi, is it you?
[00:52:40] Yeah, uh, so, ma'am, yeah, I'm also agreeing these two. Uh, maybe we can have twice a month one graded, and meanwhile, two to three assignments for ungraded, so maybe you can divide the year just periodically and.
[00:52:55] We will have 2 months of practice, maybe 3 assignments or 2 assignments ungraded, and then 1 graded assignment or quiz that will help us, you know, tracking the slabers along with practicing and.
[00:53:09] I will build up confidence eventually by the main written exam comes in.
[00:53:15] Okay, um, so it could be a quiz or an assignment that could be graded.
[00:53:23] Yeah. Periodically, maybe throughout a year, we can have 3 to 4, yeah, 5, whatever you decide, or, or… amount of slavers you decide.
[00:53:34] we cover, for example, till now, we have almost covered ML, or maybe by the next week, we will complete the ML, so when we can have 2 to 3 ML practice assignment, and graded one, and.
[00:53:47] Yeah, so it's no point having, uh…
[00:53:48] Then one quiz for entire ML.
[00:53:50] assignments on the same topic again and again.
[00:53:53] Say, for example, I gave you already unsupervised learning,
[00:53:57] Uh, probably what we could do is have some assignment on unsupervised, or whatever we do in future, we can have assignments on that.
[00:54:07] And then as we sequentially go ahead, the concepts become more and more complex, and you can have some graded assignment on some
[00:54:15] Subsequent topics. This is how I foresee, instead of
[00:54:17] Having assignments which are either graded or ungraded.
[00:54:20] I don't think we can… we should have multiple
[00:54:23] of them on the same topic, because then we are only leading to redundancy, and there's a long way to go.
[00:54:30] Uh, so you will have large number of topics, and probably
[00:54:34] Uh, those, uh, on those also, you can have graded assignments as well. So we can do that. I can give you some…
[00:54:41] For some subsequent topics, I can give you
[00:54:44] ungraded assignment, and then we can go with a graded quiz.
[00:54:48] And maybe with a graded assignment?
[00:54:51] When you do a graded quiz, then all of you have to be available online to take the quiz at the same time.
[00:54:58] Okay, so that could be during the…
[00:55:00] duration of a lecture.
[00:55:03] And maybe the quiz can…
[00:55:06] If you want the quiz, then it could be announced maybe one week, or…
[00:55:10] two weeks in advance, something like that.
[00:55:13] Okay, and then I…
[00:55:16] Yeah, makes sense for me.
[00:55:17] Yeah, and then assignments we can give, and again, we can give a timeline, say you have two weeks, or…
[00:55:21] You have 3 weeks to do the assignment, which may be graded.
[00:55:26] Uh, and then to submit.
[00:55:27] And uh… then we can do the grading.
[00:55:31] Based on that, though, of course,
[00:55:34] Uh, again, in that also, there are a lot of… there can be a lot of issues, like, for example,
[00:55:40] Uh, whether the code is done by the person, or is it done by AI itself?
[00:55:46] All that will be hard for us to assess.
[00:55:48] Our damn list, that is all.
[00:55:50] Uh, there is a…
[00:55:52] availability of a software that can help us do that.
[00:55:58] Anyway, we can give some try to it.
[00:56:00] So that is what we can do.
[00:56:03] Or, uh, better still, the assignments could be ungraded, but the…
[00:56:08] Because the quiz you do in real time.
[00:56:11] Right then, those could be graded.
[00:56:15] And, uh, like that, we could do.
[00:56:20] Okay, so, uh, so what…
[00:56:23] Uh, what we can do, uh, in the future is that I'll give you some assignments that are ungraded,
[00:56:29] And eventually, after some point in time, I'll declare a…
[00:56:36] graded quiz.
[00:56:38] Which you can take during the…
[00:56:40] lecture time itself, and then…
[00:56:43] Um, you know, that can be calculated.
[00:56:53] Yeah, so I think, uh, Simran, I request you to put up this…
[00:56:59] Uh, notification also.
[00:57:01] Uh… on the LMS, that there will be some ungraded assignments, and that will be followed by a quiz.
[00:57:08] Which will be graded, and which will be announced, like, 2 weeks in advance.
[00:57:09] Bye.
[00:57:13] And it will cover whatever has been done up to that point.
[00:57:21] Sorry, what did you say?
[00:57:22] I think it's appetizer or stuff.
[00:57:23] Sure, ma'am, I'll announce that. I'll share it with you before announcing, and then I'll put it on the events.
[00:57:32] Uh, actually, I still… sorry, I couldn't understand what is being spoken. There was some…
[00:57:33] there are some things. Now you are giving… Ma'am, I'm saying I'll upload it on the announcement section on the LMS.
[00:57:41] Yeah, I am. Yes, yes.
[00:57:46] Okay, I think Sundran already said she will do that.
[00:57:51] Any other questions? Anything that, uh, any comment, any inputs, or any questions that you have?
[00:57:57] Uh, for me, or for us.
[00:58:00] That, uh, we can take it.
[00:58:07] I do hope that you all are enjoying the
[00:58:12] sessions, the theory, as well as the…
[00:58:14] Hands-on session and the master classes, whatever.
[00:58:17] Uh, because, uh, I know that from ITR,
[00:58:21] Our team tries our level best to…
[00:58:24] Put the concepts to you in the most simple fashion so that you can understand.
[00:58:29] And then also give you hands-on.
[00:58:31] Uh, that can help you to understand what is being taught.
[00:58:38] Um, and use those concepts in the real world in your…
[00:58:42] whatever projects you are doing at office or otherwise also.
[00:58:47] If you still have some suggestions, you can please share.
[00:59:01] Yes, Adity, you have raised a hand.
[00:59:05] Yeah, ma'am, can you, uh, randomly create groups and assign, uh, us to that groups? Because if you ask us to do that,
[00:59:11] We will keep debating on how to do it, and we would never get to it.
[00:59:16] So, is it okay? Shall we just, uh…
[00:59:19] do, uh, just take the list of names and just…
[00:59:23] Uh, you know, cut the groups out of it.
[00:59:26] Uh, based on whatever the group size is and the sequence in which
[00:59:30] the groups are.
[00:59:35] Will it be better if. Okay. Okay.
[00:59:36] Yeah, man. That's better, yeah. Randomly, we can choose.
[00:59:37] That's… that's better, ma'am.
[00:59:39] You know, you want to… sorry?
[00:59:40] Yep.
[00:59:41] Okay, would it be a better? Okay, I'm so sorry. Would you be better if you can… Ask an option if people want to group themselves.
[00:59:50] Like, we are, like, 3 guys already. Um, but in the near to each other, during this course.
[00:59:58] Yeah, I think, madam, you have to collect that, uh, information, because when we are also doing already a group.
[01:00:04] So, after that, whoever is left, right, so maybe with that, they can.
[01:00:10] Uh, distributing a group.
[01:00:11] So what you're saying is that you have formed a group of some people, and if there are more to be included, that can be included.
[01:00:12] A smile.
[01:00:19] So, actually, the idea that was floated initially, which I also…
[01:00:20] Yeah.
[01:00:25] I wanted you all to do is that you can form some groups, maybe groups, small-sized group, like 2-3 people on your own.
[01:00:32] And then, uh, I'll just have a look and, um…
[01:00:35] Decide the group size, and based on that, you can include more people.
[01:00:40] Or we can include more people, but if you form some groups, maybe…
[01:00:46] 3-4 persons, or two to three people, it will be good for you.
[01:00:49] Because it can happen that, let's say, I just cut the list into subsets,
[01:00:54] equal to the number of groups required, it is possible that some of you may be on geographical different timelines,
[01:01:01] Or you may be having very different, uh, you know, uh…
[01:01:05] Uh, work hours, or something like that.
[01:01:08] So, that may create problems for all of you. So, it's good that if you can interact
[01:01:14] Here, uh, I mean, you can interact and just try to form some small-sized groups, and those can be augmented with more people.
[01:01:23] Or they can be augmented with more people by us or by you. That may be a good idea.
[01:01:29] And we have done that for the previous patch also. We allowed them to
[01:01:33] form the groups, and the groups that were deficit of people, and there were some unaccommodated learners, we tried to accommodate.
[01:01:41] In the groups. That's the way we did. And I think that works.
[01:01:44] best. So, you can try to interact, uh…
[01:01:48] Uh, and, uh, I think, uh, because here itself on the chat,
[01:01:53] Probably, if you wish, you can just, uh, after this post, this, uh, discussion, you can just post one or two lines about your…
[01:02:01] domain, or, you know, your expertise, or something like that.
[01:02:06] And, uh, those, uh, people who wish to
[01:02:10] collaborate can get together.
[01:02:12] Like that. So, the group size is not 2 or 3 officially, it will be more.
[01:02:18] Uh, I will actually have to look the final enrollment number, and based on that, I will be able to
[01:02:24] tell you the final group size. If you'll allow, I'll do it by the next week.
[01:02:28] In the meantime, I would suggest that
[01:02:31] Uh, you all can interact and try to form groups, at least of 2 to 3 people.
[01:02:37] Those which will be augmented.
[01:03:28] Yeah, sorry, I got muted, I didn't realize that.
[01:03:31] I got muted.
[01:03:33] So, I think, uh, the…
[01:03:36] uh… the idea of, uh, just maybe forming small groups and then augmenting.
[01:03:43] can be good. And you can just think about that idea. I'm sure it won't be difficult for at least
[01:03:50] two to three people to get together and, uh, you know, interact, and then collaborate, and then the group can be augmented.
[01:04:03] In the meantime, I'll work on the group size and let you know the next week.
[01:04:07] The exact group size.
[01:04:11] And, uh…
[01:04:13] We can also do this, I mean, earlier I've done it, uh, some fun activities, like, uh…
[01:04:19] You know, we did a…
[01:04:20] Group-wise, some… I used to give some assignment group-wise.
[01:04:25] And then it would be a small assignment, and the groups would be…
[01:04:29] Um, presenting.
[01:04:32] what analysis they have done,
[01:04:34] at that time. The group would be allowed, uh, so the classroom would be split into small, small…
[01:04:41] discussion groups, and each… all the people in the group would be put together.
[01:04:47] And they could discuss and come up with the…
[01:04:49] final solution, uh, of the small analysis problem, and that would be graded, and the marks would be
[01:04:56] added to the final mark. So these are some kind of fun activities, so…
[01:05:00] That have been done earlier.
[01:05:03] Now, obviously, it does require some time.
[01:05:06] to do that.
[01:05:11] Uh, I think there are some questions, like, can we form groups based on our location? Yes.
[01:05:17] You can form the groups anyway.
[01:05:20] As per whatever you feel is appropriate.
[01:05:23] If you wish, you can form the groups.
[01:05:25] On location, or you can form the groups on your expertise, domain expertise, on…
[01:05:32] Your number of years of experience.
[01:05:35] Maybe your proficiency in coding, whatever way you like, you can form the groups.
[01:05:41] It's up to you.
[01:06:10] Yeah, anything else? Anything, uh…
[01:06:14] Any other comments, questions, any other suggestions, anything that you would want?
[01:06:19] to discuss. Uh, that's why we have opened up, uh, this, uh, session.
[01:06:57] Anything else? Any other suggestion, comments, question on the theory that we have discussed?
[01:07:03] on, uh…
[01:07:06] on anything, pretty much, that you would want to know, or you…
[01:07:10] Uh, you know, you wish to do…
[01:07:17] Or you want to be covered, something, anything that you want to say.
[01:07:22] Because, uh, otherwise, we just dedicated this session for doubt resolution.
[01:07:34] Mm-hmm.
[01:07:35] Ma'am, just one point from my side. Um, so again, in my organization.
[01:07:37] Pretty much, I think they have been, um, including A in every place.
[01:07:41] Um, so we all got access to Copilot. Um, even in terms of IDE, as well as in Teams also, Copilot has got integrated.
[01:07:49] Mm-hmm.
[01:07:51] A lot of push from the leadership team is, how do we… measure the improvements, or how do we basically come up with a use case where the entire business can be benefited? So we have been bombarded with all these questions from leadership team at this point of time.
[01:08:05] Um, again, we are also trying to figure it out.
[01:08:08] What could be the best ROA in terms of a use case? But sometimes we… We do struggle to come up with one.
[01:08:16] Anything you can talk about it in general? How do we approach it, given that every tool are at hand at this point of time? How do we basically.
[01:08:25] Do the crafting properly, maybe that will… Health, uh, madam.
[01:08:28] So, actually, the thing is that, uh, you all may be having different domains, um,
[01:08:35] at your workplace. What I would suggest is that
[01:08:38] Uh, like, in the later part, your, uh, the portions on Agent TKI, I mean, generative AI and agentic AI,
[01:08:47] Primarily, we will be having text as well as vision models, both being covered.
[01:08:53] So, if your, uh, workplace, uh, does that,
[01:08:57] I mean, uses that, then?
[01:08:59] Actually, without understanding what is, uh…
[01:09:02] Generative AI and agentic AI, I don't think, uh…
[01:09:06] It will be very easy for you to…
[01:09:09] Actually, apprehend on what can be the use cases.
[01:09:13] Uh, but most of the business-relevant, I do not know what is… what you're doing at your workplace.
[01:09:19] But there can be a lot of questions, or a lot of things.
[01:09:23] That you can think about, that…
[01:09:25] actually can be solved with agents based on LLMs, or whether they are vision-based LLMs, or…
[01:09:31] like that. And accordingly, whatever the software is available, accordingly, that can be used.
[01:09:37] So, it will be difficult for me to generally say,
[01:09:40] Actually, I did provide some use cases of generative and agentic AI to the previous patch.
[01:09:47] If you all wish, I can share that sheet with you.
[01:09:50] That covers a variety of domains and some data sets and something something.
[01:09:55] If you wish, I can share it. You can go through.
[01:09:58] Some use cases where you can make use of generative AI, agentic AI, what may be the…
[01:10:04] open source data sets.
[01:10:06] Uh, that can be used.
[01:10:08] And what can be the objective?
[01:10:11] that can be solved with the… in those examples.
[01:10:15] So, that could be a precursor to you, obviously.
[01:10:20] You'll be able to understand more as you go along.
[01:10:29] In the meantime, I can see that I think, uh…
[01:10:32] Uh, Kosuri, uh…
[01:10:34] has said any topics on AI governance and security.
[01:10:38] So, security, we haven't covered at all here.
[01:10:41] Because security is an area in itself.
[01:10:44] And that requires a lot of lectures and other things, so that is not covered, but if you wish, we can have a master class
[01:10:53] To give you an introduction on AI governance and security, that can be done.
[01:10:57] But it cannot be taken as in-depth.
[01:11:00] Because that will be out of scope for this course.
[01:11:04] If that's okay, then we can do that.
[01:11:08] So, uh, Simran, I'll request you to, uh,
[01:11:11] Take a note of this topic, and maybe in one of the master classes.
[01:11:17] Uh, some, uh, topic, I mean, some concepts.
[01:11:21] Related to AI governance and security.
[01:11:24] That can be taken up, some introductory concepts, okay?
[01:11:35] So, Deepan, uh, to help you, uh, with whatever question you said, I'll share a sheet that will give you an idea of different kinds of
[01:11:42] data and what can be done with it, something like that.
[01:11:46] Probably that can help you to…
[01:11:49] Uh, you know, look at your own data and understand what else can be done.
[01:11:56] I think there is a…
[01:11:59] permit via location, that can we have industry expert session.
[01:12:04] on what type of projects, or what…
[01:12:08] company stuff, uh…
[01:12:12] uh, are expecting.
[01:12:14] It can be useful to build our portfolio.
[01:12:19] Yeah, so I think you all are having, uh…
[01:12:22] industry sessions, but I'll once again make a request to have somebody who's…
[01:12:29] We're working, I mean, in just…
[01:12:32] Not as a learner, but as an expert, he can give you some use cases.
[01:12:37] About generative and agentic AI.
[01:12:40] So that we can do.
[01:12:47] So, I think the question is about the agentic AI, generative AI session.
[01:12:54] We can have those. That's not a problem. I did give you some use cases in the beginning also.
[01:13:00] But we can still have an expert talk about it.
[01:13:07] Any other thing, anyone?
[01:13:13] Mm-hmm. Yes, you are.
[01:13:16] Uh, as you mentioned, uh, we will be having, um.
[01:13:19] text and vision models in upcoming lectures, right? Uh, I was also curious, like, um, will we cover, uh, speech, uh, speech models, like, uh, the input would be speech.
[01:13:33] And we would be training data based on speech.
[01:13:38] Uh…
[01:13:48] Yeah.
[01:13:49] Sorry, uh, the speech… sorry, I got muted. So, the speech models, uh, not part of the curriculum. We will focus primarily on the text and the vision part.
[01:13:55] So, speech is not there in that.
[01:13:59] So…
[01:14:00] Okay. Okay, but the data format would be, uh, similar, right? Uh, input vectors?
[01:14:03] So, now, speech, again, can be handled in a lot of ways. If you do a translation into text,
[01:14:09] And then handle it in a text way, yes, you can easily use the same
[01:14:14] models. I mean, tech space, LLMs, and agents.
[01:14:19] to do whatever we want to do.
[01:14:21] But, uh, other than that, uh…
[01:14:24] definitely the feature engineering will be different for, uh,
[01:14:28] direct, I mean, if you directly use the speech model, that feature engineering may be different.
[01:14:34] So, that will not be covered, but if you translate the…
[01:14:39] Uh, speech into text, then you can use whatever you learn on text, you can use it that way.
[01:14:52] Any other questions, anywhere?
[01:14:53] Okay, sounds good.
[01:15:08] Any other questions or comments?
[01:15:22] So, if you don't have any further questions or comments, then we'll just, uh…
[01:15:27] Wrap up the session here, because the whole idea was on doubts, queries, or any other comments, or anything.
[01:15:38] Oh, ma'am, I have one question. So, group of 4 would be fine, right? Or should we have the less or more, as per your experience?
[01:15:39] Yes.
[01:15:45] No, uh, I usually decide the group size based on the total number of learners we have.
[01:15:52] So I don't have that exact number right now.
[01:15:55] But it will be more than 4, definitely.
[01:15:56] Okay.
[01:16:01] Okay.
[01:16:02] But the idea is that if four of you can get together and start collaborating, you can always have more people add on to it.
[01:16:06] And that can help to build your group.
[01:16:08] It's always good to build it early.
[01:16:14] Yeah.
[01:16:15] Okay.
[01:16:16] In fact, if you wish to do the assignments group-wise, you can do it like that also, like, uh…
[01:16:21] People in the group can come together and try to…
[01:16:24] understand the problem, and…
[01:16:26] look at it, something like that also you can do.
[01:16:30] If you formed a group.
[01:16:32] Okay. Yeah, I'm interacting with a few people and.
[01:16:39] We have 4 come closer to, you know, make a group.
[01:16:42] Yeah, yeah, so that's really good. If you have 4 people, that's quite some good number.
[01:16:47] And I'll tell you the exact number in the meantime. Four of you can be together to do…
[01:16:52] to the ungraded assignments, or to any… do any kind of discussion and everything.
[01:17:02] Anyone else? Anything?
[01:17:19] Okay, if there are no further questions, comments, or queries, whatever points we have discussed, I'll request Simran to
[01:17:26] take a note, and I will follow up on those lines.
[01:17:31] I'll look at the exact numbers and indicate the group size in the next…
[01:17:37] We'll also work on graded, ungraded quizzes and assignments.
[01:17:43] As we discussed.
[01:17:45] Uh, right? So, I think those are the major takeaways for today.
[01:17:50] Oh, I mean, other than that, we also discussed that we'll be doing, uh…
[01:17:54] more on NLP slash deep learning part, and we place the…
[01:17:59] Ridge and lasso regularization with that. However, I'll be touching upon what is regularization and the cluster validity in DICE.
[01:18:07] Left oval cover, which I will cover.
[01:18:09] in the next, uh, 30+.
[01:18:12] So, with that, I think we can conclude the session for today, and…
[01:18:17] Uh, I think you can have some…
[01:18:19] Uh, you know…
[01:18:21] good family time with this.
[01:18:24] 45 minutes available.
[01:18:27] With you? 45 or 30 minutes, whatever is available with you from my side.
[01:18:35] So, we'll just wrap up the session. Thank you all. We'll meet on the next turn.
[01:18:40] Thank you. Bye-bye.
[01:18:46] Thank you