STUDENT'S PROJECT SUPERVISION
Completed Projects under my supervision
Analysis of Multiple Sequence Alignment algorithm on multi-core architecture
This work is about the speed up gained by parallelizing an iterative multiple sequence alignment (MSA) algorithms with the help OpenMP clauses and directives. We have used five types of implementation one is sequential and others are based on OpenMP thread scheduling clauses i.e. auto, static, guided and dynamic. We have noted down the execution time for each implementation mentioned above by using different size of input files. Results which we have obtained are based on the multi-core processor IntelICoreI i3-2350M CPU with clock speed 2.3GHz,which has two cores and support two threads per core.
Machine Learning Toobbox
Akshay Bhasin (1214331017), Aman Verma (1214310025), Rohit Keshwani (1214310129), Suryansh Kaushik (1214310175) Batch 2012-2016
Machine Learning Toolbox is a project that is basically meant to provide the user with a complete package where he/she can use it to know how different machine learning approaches and algorithms would work on a particular dataset. The toolbox would allow the user to provide a dataset either in structured or unstructured form and then can choose the features he/she wants to be used for training the machine. In this way the user will have a complete control in the comparison process. The user can choose from a set of algorithms and techniques that can be run on his/her dataset. After the machine has learnt from all the different algorithms the toolbox will show the results where how accurate which algorithm is will be displayed and the user will be able to compare among different algorithms and conclude on the basis of their accuracy.
REAL TIME OBJECT CLASSIFICATION AND IDENTIFICATION USING MACHINE LEARNING
KIRTIKA AGARWAL (1214310079), MAYANK JUNEJA (1214310086), PRAGYA KHANNA (1214321097) Batch 2012-16
This study examines a proposed method for object identificationand classification in real time. The proposed methodis motivated by the lack of algorithms for object identification and classification for real time input data set.The project reduces human effort for providing the input of image through defining the path and identifies the objects through an imaging device at runtime. The data required for identification process can be given statically or dynamically. Identification and classification is done using linear and non-linear machine learning based classifiers while working on both supervised learning methods.The overall appearance of object is acombination of its chromatic attributes (color) and its geometric attributes (shape, size, texture). The process starts from an initial set of data and derive values (features) which are informative, non-redundant, facilitating the subsequent learning and generalization steps, in many cases leading to better interpretations.
MULTIPLE SEQUENCE ALIGNMENT USING GENETIC ALGORITHM
TARU MAHESHWARI, PRINCY AGARWAL,SHUBHANJALI YADAV, VISHNU BALI Batch 2008-12
This study examines a proposed method for object identificationand classification in real time. The proposed methodis motivated by the lack of algorithms for object identification and classification for real time input data set.The project reduces human effort for providing the input of image through defining the path and identifies the objects through an imaging device at runtime. The data required for identification process can be given statically or dynamically. Identification and classification is done using linear and non-linear machine learning based classifiers while working on both supervised learning methods.The overall appearance of object is acombination of its chromatic attributes (color) and its geometric attributes (shape, size, texture). The process starts from an initial set of data and derive values (features) which are informative, non-redundant, facilitating the subsequent learning and generalization steps, in many cases leading to better interpretations.