Good day, Dr. Thirunavukkarasu M. I'm Arjun. I've had a look through your background and your professional record. I'm interested in hearing more about your work and how you see yourself fitting in here with our faculty. Ready to start?

So myself is Doctor M Thirunakarasu. So regarding my academic situation, I have completed my PhD from Belur instead of technology where my PhD is related to optimization of hybrid enabling system. Whereas here I am using various optimizations and techniques which is securely based upon the AI as well as ML methods. To to integrate the various energy energy resources and the using various optimization models, whether it can be regathered to.

From basic statistical from ARIMA to as is now deep deep any methods very famous deep learning methods such as annualization, neural network and Transformers. And also I recently I completed my.

Uh, applied in artificial Indium tech, applied in artificial intelligence, uh, from National Institute of Technology in Nagpur. Uh, regarding my experience, uh, uh, I am having around 9 years of academic teaching experience and more than four years, uh, uh, having an.

Uh, depicted dedicated research to research work and then, umm, uh, I, I have published uh, several papers in.

Various highly impacted reporter International journals.

Yes, that's all about myself.

Thank you for sharing that, Professor Thirunakarasu!

Yes.

Public research systematically Yes Sir. Basically this Artificial intelligence machine learning is a recent advanced technique where we need to understand some basic things. It comes from the statistical and probability. These are the very basic mathematics concept we have to learn before entering artificial indices as machine learning process In generally these concepts are. Based on developed various algorithms.

Dr. Thirunavukkarasu, let's evaluate your suitability for the role of Artificial Intelligence & Machine Learning Professor. I’ll assess your ability to teach, mentor, and conduct research systematically. To

There is algorithm whereas where is set up procedure uh that that have very.

Have a very various intellects between the uh, what can I say? It can be termed as a relationship between uh, the various times it is data. For example, if you want to analyze the time series data means uh, for example, if you want to predict the sales, uh, whether in terms of any product, let us consider in sales increase sales support the ice cream product means. Uh, you need to various data. Data is nothing but a numeric data which is related to past.

Uh, sales value. So based upon the numerical data, we, we need to analyze, uh, uh, the future prediction for this, uh, for this prediction, basically we need an, uh, practical, uh, valid data set that is a continuous learning data set as well as a model machine learning model. Uh, what is the process of predating with use of the machine learning model message? First of all, you need to select. A best model and uh, before that we need a numerous amount of data set whereas.

And then you have to select a model and then you have to train that's, uh, test, uh, split the data for training as well as test for testing purpose. And then you have to training the model with our data set. And finally, you need to evaluate the performance of our model, which is using various statistical performance. That is, that means route means square root by a square coefficient determination. These are the parameters which I have to use to determine the performance of the model. So once the model has been.

Innovative technology and a recent technology while applying the students we have to now first you have to defend the basic concepts of what are all the basic concepts related to EA and machine learning algorithm. First you have to relate so many equations. Simply it is definition as a probability and statistical methods. You have that to recall our students, these are the basic concepts and then there are several passes, but how to choose the data, how to make the model, how to train the model?

Kind means, uh, you have to, you can you can implement this model to real time applications, uh, for the students. While I am this is a very uh.

Uh, prepare your documentary model for each and rise on a step by step by step process. We have to teach the students in order to get their, uh, basic understanding concept related to misleading and artificial intelligence algorithms.

Actually, we want to.

So regarding the lab session, the very basic model, the students have to learn about the Python software which can able to design a machine learning models present by Python. Apart from the Python, there is several models like tensorflow. These are the models which has been. Readily available where you can be able to.

Analyze the, for example, if you consider the time series determinants, uh, this model can be able to analyze, analyze the data. That means the data preparation. There is a lot of data preparation that is a normalization, scaling and then data splitting process. Recently we are using a gain and method gain and method, which is used to extract the most information futures or patents from the time series data as well as from the.

Dr. Thirunavukkarasu, you’ve described the concepts behind artificial intelligence and machine learning with attention to the foundational need for statistics, probability, data preparation, model building, training, and performance evaluation. You also placed emphasis on teaching these processes clearly to students. Let’s now explore your ability to design laboratory or theory courses. How would you create a laboratory session for students to build and validate a basic machine learning model? What would the structure

Numeric image data, that is. And image data, so there are several methods, for example, the most recent now I told this is the most advanced techniques which enable to extract the patents as well as the feature. And also with the help of this algorithm, we can able to analyze the data Sir whether there is having a trends seasonality with the help of the from various decomposition methods, for example, so wavelength decomposition method. So the data preparation is a very.

Considering important concept uh while defending miscellaneous application and also the most thing is uh, uh.

Tomorrow at 5:00 tomorrow, how we are going to model the time, So how we are choosing the optimization parameters, for example, the learning rate and, uh, uh, optimal rules that that is called as A and the Adam optimizer. And also you need to, if you consider the CNN architecture means you need to consider the most important thing, the activation functions and then the input layers, dense layer and the output layers and then parameters. How many number of multiplications and addition you have to?

Uh, analysis. These are the things we have to analyze well in practicing in the case of practical cases and also.

Uh, analyze the parameter. For example, the visualization is the most important thing in the lab session. You have to visualize in the data. Uh, by with the help of the visualization, you can able to analyze the data and you can able to analyze what type of methods you have to implementing and then. Uh, finally after visualization uh, you can able to.

Uh, predict the model, predict the model evaluation with the help of the various, uh, uh, correlation. You need to find out the correlation factor auto relation factor. So with the kelp of PAC Flot as well as the, uh, confusion matrix. So these are the methods you have to, uh, implementing in the in the lab session apart from the theoretical courses.

Play.

Yeah, regarding my research project is uh, my research project is based upon in my, during my PhD, I have done how to integrate the various, uh, renewable energy sources and how to optimize the parameters with the help of the artificial intelligent technique techniques and how to analyze the model performances and how to measure the statistical performance. Uh, in my PhD courses, I got used in various artificial algorithms. For example, in the case of sine Cos and algorithm high risk algorithm.

Professor, let’s now test your ability to guide student projects and research initiatives. Can you provide an example of a research topic you would propose to your

Uh, so these are the different types of algorithm I have been used to optimize the parameter and which can I can able to build the renewable energy resource efficient renewable energy resources, which is effectively to meet the demand. Uh, after that, then during my Mtech and applied an artificial intelligence, I am used various algorithm, for example, so LSTM algorithm seen in algorithm transformer algorithm. So temporal component algorithm. So with the help of these.

The help of that, these technique, uh, I have used to predict the value of the renewable energy sources, for example, wind speed, solar speed and temperature radiation. So before we are going to build a model, uh, build a model, uh, the input data is very, very necessary. So in the case of the research work, I have used some real data works from from the websites in order to predict renewable energy sources, for example, wind and solar energy.

So far I have used different types of models, models not only a convection models and I have also used the enciple models. For example the combination of CNN and transformer gated record. In case of the model several model high avenues and I have analyzed these models how these models are performed with the help of for my my data set so and also I am applied various techniques how to improve the model accuracy predictions. These are the technique I was improved.

So regarding my student project, it is necessary to guide the student project apart from the theoretical, the theoretical concept. So the real time problem, several main time problems has been associated.

Uh, in practical application, so the real type problem has to, uh, analyze the student, uh, they, they have to overcome the real time projects with the, uh, several different models techniques, and most most recent one is the transformer, uh, which is the basic one, which is used to, umm, machine translation, for example, from in the case of, uh, language translation, this one is the recent advanced technique transformer. So the students have to.

Place this various components and they should engage the real time projects. And also the main thing is the projects based on the consent work and the grant proposals. And the very important, most important thing the student have they can't can't put the research to the industry relative. So not only for the theoretical if the students have done the research based upon the indexed collaboration means. It will be more efficient and it will be more successful.

For the feature directions.

What's 5 days to research methodologies?

Certainly, Professor. You’ve detailed your guiding philosophy regarding research, where the emphasis appears to be on integrating practical problems with theoretical foundations. Additionally, you highlighted advanced techniques like transformers, ensemble models, and optimizing model performance using real-world datasets, particularly in renewable energy prediction. To assess further, how do you ensure students learn to critically evaluate the ethical implications of artificial intelligence and machine learning in their research projects?

Sorry, I'm not understanding.

Artificial intelligence.

Uh, the very first basic things in things is, uh, you need to test the fundamentals. What are the fundamentals, uh, knowledge referred to machine learning as well as, uh, uh, a methods that is not as a fundamental, It is purely based upon the statistical and probabilities. So if the students are familiar with the probabilities, statistical probabilities model methods. They can able to understand the basics about the machine learning manner, so initially for step-by-step process.

Uh, we have to identify the.

The topic is under the various models step by step process, uh, have to with the help of the mathematical modeling equations. So apart from the only uh, teaching at theory based, it is very efficient, uh, if the students satisfy mathematical concepts behind each and every models. Methods were used in the artificial intelligence and miscellaneous models, and also apart from this mathematical modeling. Uh, it is necessary to practice a hands on training.

The store that knowledge and student gap. So where the students as a weakens weakness, whether they in a conceptual gap and mathematical modeling or mathematical equation or model system and you have to prepare a.

We are giving a very small projects and small codes, uh, and they're asking the student to solve the course. Uh, these are the very basic things in order to train the students. Uh, but with the help of these students can able to learn the code.

Also I think the students can able to understand the knowledge apart from this only theoretical concepts and also outcome based research is very, very important the outcome based research. And basically you can attend with the help of the project based learning as well as the industry collaborations, for example, mini projects and also some contacting some assessment apart from their assessments and practical quizzes and core testing. Real, real kind of implementation techniques. These are the techniques we have used to understand the students.

And and implementing the small small projects related to AML.

Regarding the concepts of the AML.

Thank you, Professor.

What else?

You're welcome, Professor. Would you like me to proceed with further questions regarding this role, or clarify anything else?

Yeah, apart from that academic research, apart from the teaching, teaching I have to maintain a proper schedule timeline, uh for research process and student mentoring and then can able to publish. Publications in very high impact at international Jonas. So the standardization is very, very important to prepare the guidelines. For example, if I have a very complex task, I have to distribute that complex task into small summer time series, time series.

And I have to allocate the particular time slot line and then apart from that I have to uh, apply a primary attribute for the each and each objective. Primary objective for.

In each and the real time applications. So what I'm coming to say means the standard situation and procedure and time scheduling is a very important from very important role in the academic health. So apart from the academic side, the RPT level, the revised blossom level is very, very important now to in order to to gain much more knowledge with the help of the some concept that is remembering, understanding, apply, analysis, creating and evaluation.

That was a great deleted and evaluated the model that is validated the model. Uh, this other 10 circular to, to teaching and then research research concert is also very, umm, important role for an in terms of academic. So you have to prepare several, uh, research material, research methodologies and rebuilding methodologies. So the and collaborative work with other industries, other interdisciplinary work. It is very, very important in the research work. And then finally.

Then aggregation work related to department and also every academic year has an I should be a relationship role in all the activities in the in the base of institutional level. So that's why it it will become the researchers academic care has also and also the institutions will also grow the student.

Will gain more experience as well as some more knowledge.

Management spell.

Dr. Thirunavukkarasu, you’ve emphasized the importance of time management, task delegation, research

Management.

That is correct, "management" is spelled as M-A-N-A-G-E-M-E-N-T.

Sorry, I'm here only.

You don't have the curries not help.

No problem at all, Professor. Let’s continue. How do you ensure that your research remains

No, due to due to unstable of Internet connection I can't able to recognize you.

And professor, no worries.

Thank you, Professor. I don't know if you'd like to give.

Thank you, Professor. Let me know if you'd like to

Thank you, Professor.

Welcome, Professor.

You're welcome, Professor. Let me know if there's

Thank you, Professor. Let me know if you’d like to continue or conclude the discussion.

Yes.

Alright, Professor. Let’s proceed further. How do you align your teaching and research goals with the current developments in Artificial Intelligence and Machine Learning to ensure relevance and innovation in your work?

Sorry to interrupt, we're down to the last 5 minutes.

Yeah, sure.

As we know, for example, there are the very.

In our Internet, for example, let us see, let us consider one small problem statement everyone has using a Wi-Fi Wi-Fi concept.

Practical problems has been available.

So if I, if I ever suppose, let us consider a space which having a dimension and X&Y and in the, in the spacing we are having our Wi-Fi. So the Wi-Fi signal has been spread uh, throughout the entire room. Suppose in case if you consider the only one person has been utilizing the Wi-Fi at any point of the corner means the signal has been sprung overall throughout the entire run, so. Uh, So what will happen? Means, uh, that particular person has utilized only a small signal because.

Uh, there's, uh, there are the most of the signals are unbased. So as a result, the signals, it may be some decomposition and, uh, ultimately the signal gets lost at distort disorder and also the user has able to attain the very big signal. So in order to avoid these practical problem, it is necessary to install the various syns as the sensors has, has been. Analyzed the current time, current real time data patterns. And with the help of the data which has been received from the census, we can able to build the air and machinery model. That model can able to predict where the person has been exactly located and given instruction.

The model given instruction to the sensors that particularly the sensors, has been.

Thing, uh, directed to the only the particular location, so where the users has been available. So this is one of the most recent example about the machine learning model. So by doing this we can able to, uh, get more signal, uh, with last discussion. So these are the real time practical applications and also coming to the medical side and coming to the healthcare center with the help of the.

Model which can able to detect or predict the disease which which has not been in there later, Sir, but initially but very at the very first stage only we can we can able to analyze the what can you tell? We can able to analyze the DCS in healthcare center and also in several cases we can able to identify. For example, in the case of medical properties, medical documents there are.

Several documents which is maybe a string value or numerical value. So with the grip of the a machine algorithm, if you want to able to and be able to predict only the for example the contact number of a patient or contact number of a doctors who can able to schedule.

Schedule a investigation time according to our need. So these are the most practical applications related to AM model.

Yes, that's all I said.

Share those insights.

Thank you for calling safety.

To welcome professor, to keep strongly.

You're welcome, Professor. Just to confirm, would

Yes, I would like for dinner.

To confirm, Professor, are you saying you would like to conclude the interview now?

Yes, No. Could you repeat please?

What's the professor are you requesting?

Sorry.

Good afternoon.

Yes, yes, I would like to continue.

Proceed the approach back.

Understood faster.

Hi Professor, could you explain how you say?

Professor, could you explain how you stay updated

Sorry you are the my Internet is unstable I can't able to understand you.

Professor, can you explain how you can?

I understand.

Yes. Could you repeat please?

Can you explain how you say update at all?

Yeah, Jason, so we got.

Regarding updation, uh.

Adam uh, so from the various standard methods, if you consider the standard method means the standard method has used only can able to capture the linear relationship between between the data uh, whereas it is the standard AML method not able to.

Uh, handle the nonlinear data, uh, uh, so as the temporal data and the spatial data. So it is very difficult if the data is data set is very large. So proportionally like you're moving from standard arithmetic model to the most recent model is the deep learning model, which is a convert CNN, which is a convolution neural network, uh, while coming to the architecture of the conversation neural network.

I need to consist of the uh, the convolution layer, filter, pooling layer at feed for a network and the input dense layer, output layer and activation functions. And these are the various architects related to the CNN. So these are the most recent technique nowadays were available, which has been.

Uh, very familiar in the AML techniques.

How do you create these attachments?

Add a clarify, Professor.

No, I yes. Could you repeat please?

Of course, Professor. How do you incorporate advancements like CNNs into your teaching or research methodologies to ensure relevance and effectiveness?

For an images that means to analyze the images. For example, let us consider we have a large set of data that means we have plenty of datas around millions of datas related to images, which one which is related to docs as well as cat. These are the input images we are going actually we are going to fit into the CNN. But the output, it may be maybe classification or regression. Uh, here I am considering the classification. Uh, we are going to, uh, we are, uh, we are already having a millions of data which is related to the as well as cat. We are going to model the CNN model, uh, with this input data. So what actually, what actually happens in the CNN data? The CNN has extract the futures has extracted the futures, uh, related to the input data which I am feeding, for example.

Yeah, CNN method. If you consider the CN method, it is actually CNN method is basically designed.

Uh, single dimension underwear having that, we're going to applying a filter, uh, as a three cross 3 dimension. So with the help of this filter, we can able to, uh, predict the relevant pattern or relevant feature from the images. And then again, it is going to Max pooling pooling layer. It may be Max pooling. So Max pooling misses extract, uh, uh, most specific patterns from the images. And then it is forward to input layer and output layer.

In simple, the input image has been converted to.

And then Slayer. So finally, uh, if it is a, uh, classification project means classification means, uh, when we're applying a new data already, the model has to be trained with these data. So with the help of the future features, when you apply a new data means this is the CNN has able to extract the features, whether the features related to cat or dog, for example, The most important feature is if you consider the, uh, dark and tight in terms of seismic, the catheter small size. I said, Sir, if you compare the US, Mr. the cat is small here, so the cat very big ears and similarly with the help of the color, you can add to identify the breeze, whether it is a cat or cat. So these are the few, these are this is the basic things in the conversion layer layer conversion neural network, it is nothing but the advanced techniques of neural network. Whereas in the case of neural network, we have used only a simple. Only a small layer, that is one input layer, one output layer 12.

Whereas in the case of convolutional layer so between the input and output layer we are using multiple hidden layer. So with the help of this the CNN can able to extract the feature. It is nothing but it is simple mathematical calculation.

Umm, I simple my it is purely based recommend statistical and probability message. So with result of this we can able to identify the uh classification.