Lecture 01 : Part I

[Music] hello and welcome today we are going to start the first lecture on this very important course called artificial intelligence and machine learning in materials engineering that's what I said you are welcome to the course on on artificial intelligence and machine learning so I welcome you again and this is at nptl course that means it will be offered on swam portal in nptl the full name of the course is artificial intelligence and machine learning in materials Engineering in short form we can call a IML in math engineering so we will refer this name very often the short name very often in this course first let us discuss about myself about the course about the things which are going to learn in the next 10 minutes then we'll talk about it what is AI and how it is going to change the whole world Al as well as how we can integrate machine learning as a part of this course so my name is Krishan vishwas I am working as a professor at the department of Material Science and engineering at IIT khur that is Indian of technology kpur and important thing is my email address that you should note down that is K biswas at IIT k. c.in if you have anything doubtful anything you need to clarify or if you want to have some discussions please use this email address as a mode of communication for this first otherwise also you can contact me in using uh you know phone numbers so normally this is the internal phone number you have to add 0512 as a city code for kpur and then you have to add 679 and then 6184 so again I repeat if you want to contact me by phone the city code for kpur is first 0512 then 679 6184 you can call on this number if you want to contact me but my main uh communication with you will be happening through this email address so please note it down if you want to know details about myself please uh browse my website which is available uh we do work on artificial intelligence machine learning for last Almost 78 years so this lectures will be available online on nptl forums well before you begin about any course first and foremost thing to know what are the what are the objectives of these kind of courses because as you know AIML is a bajart so everybody talks about artificial intelligence nowadays irrespective of branches of science engineering and even in literature so for this particular course we must know what are the objectives I list down few of them first to know basics of AIML this is very important that mean this is a introductory level course this is not an advanced course this is you here you know about some basic aspect of IML and how to correlate with the design and discovery of new materials what kind of new materials new materials with betterment of properties mechanical functional or other properties so we'll discuss about some examples later on in future lectures where you will see how AIML can be effectively utilized to correlate design and discovery of these kind of new materials secondly it is important to know technological aspects of utility of machine learning in design of materials because we are engineers and as a part of engineering we must have good knowledge about technological aspects of using machine learning and design of materials because design of material is something which is universal whether you're designing material for catalytic reaction designing material for let's say something like a brass bar you need to know how to do that using technological aspects whether we can model it or find the best material using machine learning that's what you'll be taught thirdly a IML is a tool as you know this is a very important tool nowadays so how we can use this tool for discovery of new materials yes I'm stressing the word Discovery because Discovery is something which is very fancy Discovery means thinking of something out of the box new which has never been done never been thought so therefore whether this AIML can be used or the tools ofl can be used for new material Discovery this is something which will be knowing these aspects I hope you have understood the objectives so the end of this course we have to see whether these objectives are satisfied or not and if not what are left behind and if it's so then whether you can improve further or not well before we start even we should know also whether do you need some prior requirements or prerequisites for this so we know B you need to have some basic idea about Material Science that is structure characterization materials mechanical functional properties a basic idea I'm not talking about a deep understanding you need to have know that stru and me crystal structure structure mean micro structure structure means some kind of a you know Atomic ements in the materials also you need to have some idea about how to characterize this material so that we can use ex defraction or we can use some other microscopic Techni like ACM or Optical microscope to characterize similarly mechanical property like tensile deformations or spere deformations kind of things functional properties like electrical properties electronic properties also many other like you know functional properties you should have some basic idea I'm not talking about deep understanding some basic idea will help you to make your learning much better that's something which is required from you well how this course is structured let's talk about a little bit on these aspects first six lectures will be based on the introductions introduction means introduction of a IML as well as some basics of Material Science that's what I term as basics of materials I mean to say basics of Material Science we'll talk about these aspects in six lectures and as you understand these lectures are divided into two parts every lecture will have two parts 25 minutes duration approximately about 25 minute duration and so therefore there will be total 12 parts for this section of the lecture this will followed by how we can do material design at different length scales what I mean to say is that if I look a component let's say I have a I have a watch this is a component right and this component has different parts and how do I design each of these parts to make this is an useful Gadget so design starts with the length scale which you see here that is the meter level or millimeter or centimeter level every component will be having certain size it can be very small millimeter level it can be very big like meter levels right so therefore we start with that and can we go down to the atomic level and we can understand the design principles at different lens scale that will be covered in this five lectures obviously I'll not be able to do justice because this design of material a different lens scale requires one full course but we'll try to give you important aspects of message how we can actually go through these aspect things this will be followed by almost 15 lectures on two topics which is the bulk of these course first the machine learning approaches different machine learning approaches for material design here we'll talk about how machine learning app have come through and how they are making used and then how these machine learning approaches can be used for both material design as well as to accelerate the material development and deployment see finally if I design a new material and make it I must deploy it upon going through all the stages of material development then only it is an success story so whether machine learning can help us in design development and deployment and most importantly whether we can accelerate that machine design and development and deployment aspects using the machine learning we'll be focusing these aspects in this this 15 lectures and as you know everybody talks about data science and materials is one of the old field on the Old Fields of research or development it's almost 5,000 years old the start of material happens when human were you know trying to develop uh the civilizations right that's what so initial materials what what what in the mother nature okay let's say far far is easily available and it was used instead of this coat or these kind of shirs human use far to protect themselves from the atmospheric temperature or whatever it is similarly human app used also the available Metals which are freely available in the Earth or the crust of the earth so from that time to till today lot of materials have been discovered so that means we have huge knowledge on this subject it's compared to other branches which are coming recently so that means we have generated a lot of data so how to manage those data how to use those data so there's something called a material data science or rather I will say material informatics that's a better term to use and can you use this material informatics in a way so that we can use this as a mode of making new materials oring new materials so we will not talk about only informatics of materials but also experiments how we can accelerate the experiments how we can develop the new experiments using this kind of inform matics so all together there'll be 40 lectures that means if I consider each part there will be 20 80 parts for these lectures and you will be given a complete overview of the whole subject how we can use machine learning artificial intelligence machine learning combin so that you can get an idea about materials design in this course there a detailed aspect of the course first we are doing introducing the subject then we'll talk about Basics so Material Science that is processing structure and property correlations that's the basic thing in our material straon where we talk about how to make it and depending on your processing route structure will be governed by that and the structure will then govern the properties properties will govern the performance of any materials that's called material tetr and there are four pillars of that processing structure property and performance and that will be taught in this then we'll talk about as I told material design a different lens scale we'll talk about different kinds of techniques introduced how we can do that for metal level scales to about Atomic level Armstrong level skilles then we talk about machine learning approaches like compositional property space how we can design new compositions new materials how we can predict the properties or how we can Target the properties actually that's better way to tell you what I mean to say is that if I want to make a material with some Target properties can I make it and whether we can use machine learning for that and accelerating the material development and deployment cycle as well as informatics is going to there but you know I'll be talking about some case studies from my own research as well as from the research from others which will acknowledge as a part of this uh you know course and to give you the better perspectives of the course structure so this is in a nutshell is the whole a detail outline of this course so what are the books you will be following well three books will follow they are very good books first one is by lukman at all and you know this book is known as material design Discovery and design by means of data science and optimal learning this is a spinger book and I have a copy of this book I like to show you this is the book it's not a very uh you know thick book you can see that so a lot of good things are discussed in this book please uh if you want to have a copy you can download it's a available on the internet you can buy it or you can use it or you can you can have a hard copy also the next book is by Chang at all which is on artificial intelligence for Material Science I also have a copy of this book you can see that uh this this is the uh Artificial Intelligence on the machine learning and this is the material Discovery and design I'm showing with these two books so which will be using this these two books primarily for our purpose in addition I have also listed another book I don't have a copy of this book but I'll show you uh this book letter when I give lecture this is on this is by field de luna accelerated material Discovery how we can accelerate the material Discovery uh this is published by ders very recent books all of these are actually recent the last one is even hardly two years old so subject is also new books are new and you will be learning a lot of new things as a part of this course well let's look at what are the special empasis we give so we are going to talk about design of materials with improved properties and and we'll be using AIML tools for that that's our special emphasis on this course we are going to talk always on these kind of things how we can design materials with this improved properties so we'll talk about course TS later because this is something which will come later and we'll provide you on the internet so how we can evaluate what are the ways of evaluating you know first thing is that at the end of this course you'll have online or offline exam depending on the number of candidates taking the exam each candidate is very large then we have to go for online exams and if it's number of candidates is less then we can go for offline exams mostly I find online is more preferred there'll weekly assignments which will be uploaded after lectures are uploaded after about a weeks to at the end of the week rather the assignments will be uploaded you we'll get times about one week to solve the problems and to submit online and there will be minor online quiz also this is something I'm going to introduce where we can actually do online quiz by very small minor level online quiz to understand whether you are able to understand or graas the subject matter or not well that's about the course and details about that so now let's move a little bit on details of this course so that's I said you this is a course on nptl so everything will be available on swam portal that is swamp. gov.in uh slnc details by nptl and you have to search the this course name IML liit engineering you'll find out that and you'll get this course so at the end I I like to tell you that this is something I get it from from the internet you are going to take a journey on the unknown territory as usual and you know the running away from the problems only increase the distance from the solutions we are here to find Solutions of our Global problems each one of us and that's what is the important thing so therefore it is always easiest to escape from the problem problem than to solve it that's what is the human Tendencies but you uh focus on your subject you will learn it and when you embark on unknown teritory a lot of things which you will learn by yourself that's something will do yourself so therefore don't run away or lost in thoughts this is very important for you okay so first uh you know why why you need to use materials engineering and why this data informatics in important in material design why it is so let's talk about that that's what is the introduction but before that let me first tell you what do you think about AI everybody talks about artificial intelligence right very simple example of here is you know I can give you my own experience suppose I'm sitting in my office doing some work and uh you know I have a friend and we are thinking of going to my home and and you know and once we talk about it we're going to have some food or drinks maybe TI tea or coffee so how do I make sure everything is ready at home no I'm not talking about I call somebody at home and ask somebody to some human to prepare no I'm talking about whether can I use machines to that well yes we can do that using your mobile phones actually you can tune your microen in in in your home to make the food hot and ready before you reach the home that's basically an artificial intelligence okay you are you are using a tool which is allowing you to make your food ready this is one way of defining artificial intelligence you know AI is basically ability of machines and computers to mimic human mind decision- making process that's what is important so human minds are very powerful we have the abilities to decision making the decisions and as well also problem solving capabilities we can solve many problems that's why we are here the society civilization is here because we have solved many problems so here we talking about it ability of the machines and computers whether they can mimic this quality of human human has a natural intelligence so can machines use this as aspect to learn and you know this lability is OB Limited at this moment there are many reasons for it but they are limited with AI machine can understand your whatever Target task you giving and achieve specific goals very simple if you are giving machine do something whether the machine can do that job properly and achieve the specific goals or not that's something which is very important so therefore it is built on two important aspects AI one is machine learning other is deep learning so if you talk about AI as one block machine learning and deep learning are parts of that so machine learning uses structure data data which is structured which is nicely maintained obviously there are machine learning techniques where one can use unst structure but mostly they use structured data deep Ling use unstructured data some form of a text or images even you can use images also that's the difference so in a nutshell machine learning is more widely Ed because you have structured data and that's why many of the materials discoveries or material development has provided us all kinds of structure data and we can use in this but you know in this course I'm going to talk about lot of aspects about deep learning at the L part of this course so therefore there I'll teach you how we can use unstructured data like you can can you use a text or images an email is basically process by which machine automatically learn from existing data I hope we understand in this process we'll be giving the machine data and machine will learn from the data and then machine will give us specific or specific goals and achieve specific goals that's what it is how human intervention is still requ in this process it is not that machine can do it themselves but you know the way the scientific discoveries or technolog discoveries are going on one day machine will be able to do with everything without intervention of human so uh so therefore let's talk about a little bit more what is AI if I say system that can think like human is it AI yes it is true correct human can also do that but systems can also do that they can think like that see system can act like human do you think so no system cannot act like human always but system can think system that can think rationally what about that or system that can act IR rationally they are called irrational process so for the machine for AI first part is called thought the system which can think like a human or they can think rationally also that's called a thought second part which is actually call a system which can act like a human or which can act rationally is called behavioral aspects what I mean to say is that you know sometime you think about something before you act right you're thinking of going someplace that's your thought process that's not actually done at the time beginning you're thinking only whether I I'll watch this game or not I'll go this movie or not you're thinking about it but then whether your thinking process is rational or not that's also important let's say uh if I say I'll go to Jungle or I'll tell one more example I go to Jew and I'm in the Jew is sitting and thinking that whether shall I go inside the cage of a tiger or not that's irrational thinking why because I may be getting killed by this process so that's what is human can always think rationally but there are a lot of people in the world who are think irrationally also forget about that but mostly human can think very nicely rally and they can think also that's the called thought process now once it is T once it is done let's say machine can think like human although that is not still not achieved to the level to which we can think about it and whether they can act like that also like it can act like humans you can actally not so these are the features AI normally uses we features which human has the the features a use say okay let's talk about the tuning test then we'll stop system that can act like human can I do a curing test art of creating machine that perform functions that require intelligence when perform by people that's basically taken from Kel okay as you know this is an invited Commerce and now second is the study of how to make computers think at which at this moment people are better computers do things at this moment people are better okay that's from reach and night so the these are the two text which are going to do Turing test so let's do that so you enter a room which is a computer terminal you have a fixed period of time to type what you you want and then stud replies at the other end of the line there's either a human being or a computer system let's do this experiment and you know it is a computer system at the other end of the period you cannot R determine whether it's a system or human human then system is deemed to be intelligent what I mean to say you have typed something on the computer and somebody else is there on the other side of the system who responding to you now if at the end of your typing you cannot detect the reply is coming for a system or human then system must be intelligent and it doesn't matter the system or computer or machine or a human so it clearly tells you that it if we can send you replies that's intelligence system that's what is a a test so system which can act like human are natural language they can do this kind of task actually uh if they are act like human first is natural language processing NLP you call it to communicate with human second is a knowledge representation that mean to store information effectively and efficiently that's called knowledge representation third one is automated designing to ret an answer using the store information that's what this experiment was that and last one is the learning to adapt new circumstances I hope you understand that these the four pillars of understanding what you require you know the cognitive task will be natural language processing knowledge representation automa designing and machine learning these are the four pillars of any AI system okay let us stop here and we'll carry over this on this subject in the next lecture also [Music] [Music] [Music]