FACULTY DEVELOPMENT PROGRAM (FDP)
FDP DAY ONE
21th June 2021 10:00am -11:30am
Topic: Introduction of Artificial Intelligence and Machine Learning
Speaker: Dr. Mukesh Kalla HOD Department of Computer Science & Engineering,
Dr. Mukesh Kalla explained about the following:
AI deals with stimulation of intelligence, it must be able to think rationally and act rationally. Physical symbol system is a hypothesis then it should be possible to program a computer to produce intelligent action. When evolution of artificial intelligence turning for the distinctive character, symbol has not performed well and it has not generated the intelligent actions. Rise of deep learning and various competition on visual recognition on image etc day by day AL evolution. AL had advantages of this era that influences advancement in large dataset, computational advancements, complex algorithms etc. Data can analytics by descriptive and predictive in AL. Estimation is used for predicting a continuous value and predicting an output variable gives input variable in regressions from algorithms like logistic, stepwise, support vectors etc.
Machine Learning (ML) is the subfield of AI. In this the task is measured by performance, performance is improved by experience and experience is processed by task.ML algorithms are: SL (classification / estimation), USL (clustering / prediction), RIL (decision making).
Deep Learning (DL) is the subfield of the ML. Supervised learning uses: prediction of future cases, knowledge extraction, compression and outer detection. Reinforcement Learning applications: self-driving cars, NLP, news recommendations, healthcare, robotic manipulation etc…. Regression is a part of SL used to predict continuous value; some Categories are: simple learning regression, multiple learning regression, polynomial learning regression.
21th June 2021 11:45am – 1:15pm
Topic: Supervised Machine Learning techniques Speaker: Dr.S.P Harsha (Mech Dept, IIT Roorkee)
Dr.S.P.Harsha explained about the following:
Improving a task or process with repeated learning of various scenarios. Classification of machine learning: Supervised, Unsupervised and Reinforcement. Process of Machine learning: Data Collection, Data Pre-processing, Feature Engineering, Algorithm Training, Marking Predictions. Supervised Learning consists of classification & Regression, Unsupervised learning, Reinforcement learning and their categories, respectively. Algorithm selection and training, Goal: making correct prediction like Incremental improvement, Use of metrics, Hypermeter Turing. Machine learning parameters: Modern parameters, Hyper parameters.
Classifiers: Classifier margin, maximum margin. Classifier Margin: Define margin of a linear classifier as the width. Largest Margin: The maximum margin linear classifier is the linear classifier with the maximum margin. Optimization Problem Solution: involves computing of inner products between all pairs of training points. Failure Response Analysis: Machine motion, Predictive feature. Detail of: Various aspects of SVM (Support Vector Machine), Kernel Trick, Classifiers, Failure response analysis, Performance & Comparison.
21th June 2021 2:30pm -04:30pm
Topic: practical implementation on Regression Model
Speaker: Mr. Saurabh M Sharma Research Scholar (MTech, CSE) at sir padampat Singhania university
Mr. Saurabh Sharma explained about the following:
Regression problem: regression analysis is a set of statistical processes for estimating the relationships between a dependent variable and one or more independent variables. Regression Challenges: data cleaning, finding dependent and independent variables, data pre-processing, Preventing one feature. Regression problem type: Univariant, Multivariant and polynomial linear regression. Hypothesis: simple linear regression, multiple linear regression/ multivariant regression, polynomial regression. Data pre-processing: check and fill the missing value, normalization (multivariant problem), categorical values and dummy values (multivariant problem). Problems from dataset on how to implement the accurate values to plot the graphs and to assign the values accurately. For loss/cost: square the error, sum squared error for all predictions and mean the sum of squared error.
Implementation: Raw implementation and Sklearn implementation with practical example.