Swaroop Venu

Participated in Kaggle Competition How much did it rain?
It aims at using the high quality polarimetric radars data to improve the rainfall predictions.

• Implemented preprocessing to handle large number of missing values.
• Implemented 3-fold cross validation for choosing hyper parameters.
• Implemented multiple model technique to handle missing values.
• Implemented techniques like linear regression SVM and Gradient descent Regression.

Most of us spends hours on googling, this extension helps opening the result pages faster and provides custom option. Pages open around 200ms minimum to sometimes 2s faster.

How?
Whenever you click on a link, Google opens a generated URL and then re-directs back to clicked page. Google uses your click to update their algorithm and history. This extension by-passes this updating and opens your search result.

This has over 100 active users

chrome web store link

ebaymashup.appspot.com/

Implemented the Principal Component Analysis (PCA) algorithm on MATLAB for data (image) representation compression and then use the compressed representation for classification.

Investigated the data from the Titanic disaster to see if passengers survived at random, or if characteristics like age, sex, ticket amount, cabin, etc. influenced whether they survived. For feature selection used Mutual Information and Sequential Feature Selection algorithm. Performed Logistic Regression using Batch Gradient Descent and Newton’s Method in Matlab.

• Crawled around 200,000 weapon images from a given seed list of 84 urls using Apache Nutch.
• Developed a Python script to retrieve MIME types of crawled data using NutchPy library.
• Developed a plugin for protocol-interactive-selenium to handle problems like Ajax interactions, protocol issue, retrieving data from behind a form and retrieve images present in php scripts.
• Extracted the text and metadata of the images using Apache Tika and Tesseract.
• Developed 2 deduplication algorithms to identify exact and near deduplicate images based on the extracted text and metadata retrieved from the weapons images using Tika. SHA-1 hashing algorithm was implemented to find exact duplicates while SimHash algorithm and Hamming Distance were implemented to find near duplicates.