Download Vibhuti Mahajan - IITK - Indian Institute of Technology Kanpur

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Nonlinear dimensionality reduction wikipedia, lookup

Transcript
Vibhuti Mahajan
Contact
Information
Education
Senior Undergraduate
Dept. of Mathematics & Statistics
Indian Institute of Technology, Kanpur
e-mail: [email protected]
Mobile: + (91) 7897418844
Degree / Certificate
Institution
CPI / %
B.S. Mathematics & Statistics (2016)
IIT Kanpur, India
8.6/10.0*
Class 12 : CBSE Board (2012)
Mount Carmel School, Chandigarh
93.8 %
Class 10 : CBSE Board (2010)
Bhavan Vidyalaya, Chandigarh
9.8/10.0
*after completion of 6 semesters
Awards and
Achievements
• Recepient of Anita Borg Student Scholarship for the year 2015.
• Awarded Inspire Scholarship ’12-14 by the Govt. of India for meritorious academic performance.
• Selected for IISER Mohali Summer Research Programme 2014 among 5000 applicants.
• Ranked in Top 0.5% among over 500,000 students in IITJEE’12.
• Ranked in Top 0.5% among over 1 million students in AIEEE’12.
Publications
Conferences
S.G. Liao, C. Yin, V. Mahajan, Y.S. Park and G. Tseng, “SeqDesign: A Framework for Genome
Wide Power Calculation and Design Issues in RNA-SEQ Experiment”, to be submitted to The Annals
of Applied Statistics [pdf]
• Grace Hopper Celebration India (GHCI) 2015 for women in technology.
• Women Emerging in Finance 2015 organised by Goldman Sachs, India.
Research
Experience
SeqDesign: Framework for Genome Wide Power Calculation in RNA Sequencing Experiment
Mentored by Prof. George Tseng, Dept. of Biostatistics, University of Pittsburgh, USA (May’15 - Present)
Proposed a novel method, “SeqDesign”, for experiment design of RNA-Sequencing using Next Generation Sequencing (NGS) data. The goal is to predict power (p-value) of the sequencing experiment
as a function of number of people sequenced and read depth and design an optimal experiment given
desired power and budget constraints. Distribution of p-values is modelled as a mixture of negative
binomial distributions of differentially expressed (DE) and non-DE genes and is fit using maximum
likelihood techniques on pilot data. Wald test is used to test for DE status, Z statistics from the
test for pilot data are transformed to original experiment to find its p-value . In both simulation
study and real data analysis, our method provided genome-wide power prediction and optimal experiment design outperforming existing methods, reducing RMSE by a factor of 2 . Also developed a
user-friendly platform with customizable I/O format.
? R package on the method is currently in development. Research paper to be submitted to The
Annals of Applied Statistics.
Statistical Methods in Market Research [report]
Mentored by Prof. Amit Mitra, Dept. of Mathematics & Statistics, IIT Kanpur (May’14 - Aug’14)
Modelled and implemented SAS code to evaluate dynamics of customer relationships using typical
transaction data. Multinomial Logistic Regression and Markov Chain model are constructed and
estimated to model the customer transitions. Used Principal Component Analysis/Canonical Discriminant Analysis to reduce dataset dimensions and then clustered customers on monetary worth
using wards hierarchical clustering . Predicted revenue band of customer with 94% accuracy .
? Selected among top 30 (out of 100) projects at Poster Presentation Competition, organized by
Students Placement Office, IIT Kanpur and accepted at IITK Student Research Convention’14.
Selected
Projects
Low Rank Approximations for Kernel Matrices: Nystrom, ALS [report]
Course project, Learning with Kernels, Prof. Harish Karnick, IIT Kanpur (Aug’15- Nov’15)
Kernel approximation is an effective way to deal with high costs involved in computation and storage of
kernel matrix for the large scale data. Studied and compared SVD, nystrom approximations, Random
feature based approximation and Leverage Element Low Rank approximation (LELA). Empirically
showed that LELA algorithm accompanied with random sampling and alternate least square estimates
gave more accurate results as compared to other techniques. (Software used: R, Python)
Bike Sharing Demand (Kaggle) [report]
Course project, Artificial Intelligence & Data Mining, Prof. Amit Mitra, IIT Kanpur (Jan’15- Apr’15)
Formulated exponential gradient boosting method (GBM) for bike rental demand using weather data
and historical usage with 88% accuracy. Compared model with existing linear and non-linear techniques, GBM was found to perform better. (Software used: Python)
Recurrence of Breast Cancer [report]
Course project, Regression Analysis, Prof. Sharmishtha Mitra, IIT Kanpur (Jan’15- Apr’15)
Predicted recurrent cases of breast cancer based on tumour cell attributes using logistic regression with
accuracy of 72% Also predicted the number of days after which tumour onsets again in recurring
cases using a multiple linear regression model. (Software used: R)
Predicting Daily Exchange Rates using Time Series Analysis and Neural Networks
Course project, MTH517 (Time Series Analysis), Prof. Amit Mitra, IIT Kanpur (Aug’14- Dec’14)
Used linear deterministic time series (ARIMA) models to predict foreign exchange rates of INR.
Compared results with a non-linear neural network model to comment on the goodness of fit and
prediction accuracy of linear model. (Software used: R)
Relevant
Courses
• Statistics: Stochastic Processes & Applications*, Regression Analysis, Inference, Time Series
Analysis, Probability & Statistics
• Analytics & Artificial Intelligence: Machine Learning Techniques*, Bayesian Data Analysis*,
Convex Optimization*, Financial Engineering*, Learning with Kernels, Statistical & AI Techniques in Data Mining
• Algorithms & Computer Science Theory: Algorithms II (audited), Theory of Computation,
Mathematical Logic, Data Structures & Algorithms, Fundamentals of Computing
• Mathematics: Multivariate Analysis*, Partial Differential Equations, Ordinary Differential
Equations, Complex Analysis, Several Variable Calculus & Differential Geometry, Principles
of Numeric Computations, Abstract Algebra, Real Analysis, Linear Algebra
*presently doing
Technical
Skills
Programming Languages - C, Python, R, MATLAB, HTML
Software/ Platforms - SAS, 3ds Max , LATEX, Microsoft Excel, Linux, Windows
Positions of
Responsibility
Captain, Table Tennis Girls’ Team, IIT Kanpur - Currently leading the team, selected as Best Incoming Sportsperson, 2013 and consecutive second runners up in Inter-IIT Sports Competition.
Student Guide, Institute Counselling Service - Mentored 6 newly enrolled first year students to
acclimatize with university environment and academic system.
Senior Executive, Professional Affairs, Techkriti’14, IIT Kanpur Technical Festival - Worked in a
team of 30 and effectively planned and managed talks by eminent scientists.