Anmol Rajpurohit

Paper Abstract:
During the last few decades the field of neural networks and massive parallel computing has evolved rapidly. Artificial Neural Networks have been studied since 1943, when Warren McCulloch and Walter Pitts wrote their first paper on how neurons might work. Since then there has been a tremendous research work on neural networks to simulate brain processing. Neural networks are computation intensive and therefore, parallelization helps to reduce the training time significantly.… Show more

Paper Abstract:
During the last few decades the field of neural networks and massive parallel computing has evolved rapidly. Artificial Neural Networks have been studied since 1943, when Warren McCulloch and Walter Pitts wrote their first paper on how neurons might work. Since then there has been a tremendous research work on neural networks to simulate brain processing. Neural networks are computation intensive and therefore, parallelization helps to reduce the training time significantly. In this paper we would cover how neural networks evolved over time in distributed environments. We would start with the early implementations of neural networks, where researchers started to leverage parallelism and cover the recent advances in this direction that has made neural networks such a big success. Neural networks are widely used in the field of natural language processing, image recognition, speech recognition, etc. Since neural networks generally contain a large number of layers it has been considered as one of the prominent Deep Learning architectures. Show less

Social media has become an important part of lives of everyone today. Nearly two-thirds of American adults (65%) use social networking sites. There has been a rapid development in this field during the past decade. Earlier, it was treated just a medium to interact online with friends and family. However, it has evolved over the years and now it is an integral part of not only individuals but also teams sitting in large corporate organizations. On the same line, there has been a constant… Show more

Social media has become an important part of lives of everyone today. Nearly two-thirds of American adults (65%) use social networking sites. There has been a rapid development in this field during the past decade. Earlier, it was treated just a medium to interact online with friends and family. However, it has evolved over the years and now it is an integral part of not only individuals but also teams sitting in large corporate organizations. On the same line, there has been a constant development of tools and techniques to leverage social media at its best for marketing, campaigning and other purposes. There are umpteen tools available in the market, which perform text analysis to get hidden insights from mass chunk of social media data. Text analysis lays foundation for sentiment analysis, an art to understand the sentiment from text. However, with the evolution of social media, now it involves a large mass of media content such as pictures, videos, etc. Although everyone would agree with the expression by Tess Flanders that “Use a picture. It's worth a thousand words.” there has not been any significant development of analysis of media content such as images/videos on social networks. In this paper This project aimed towards filling that missing piece by implementing an approach to analyze images on a particular social network. Show less

This embedded systems project involved:
Use of an android cell phone’s microphone to record sound that is played when a string is plucked.
Android application processes the signal from microphone to determine if the string is in tune or if the pitch is too low/high.
Phone communicates to an Arduino board via bluetooth.
Arduino controls a servo module to rotate the guitar peg accordingly.
The above process is repeated until the correct pitch is achieved.

Report Summary(can be shared on request):
Adopted open-source implementation of Connect Four Game.
Worked on enabling computer as a strong opponent against human player.
Applied various mechanisms and analyzed their impact on the analysis of game tree and moves.
Performed a comparative analysis of various approaches starting from brute force to smart heuristics.

Built a relational database system from scratch in C++ (The source code for the project is private as required by the course policy. If you are interested, I'll provide access based on your request):

Layer 0: Implemented a Paged File (PF) system and few operations of a Record-Based File (RBF) manager.
Layer 1: Finished all the operations of the RBF manager and implemented a Relation Manager (RM) on top of the basic record-based file system.
Layer 2: Implemented an Indexing (IX)… Show more

Built a relational database system from scratch in C++ (The source code for the project is private as required by the course policy. If you are interested, I'll provide access based on your request):

Layer 0: Implemented a Paged File (PF) system and few operations of a Record-Based File (RBF) manager.
Layer 1: Finished all the operations of the RBF manager and implemented a Relation Manager (RM) on top of the basic record-based file system.
Layer 2: Implemented an Indexing (IX) component which provides classes and methods for managing persistent indexes over unordered data records stored in files. Built indexes using Top-down B+ Tree approach.
Layer 3: Implemented a query engine which provided functions such as Filter, Project, Join (Grace Hash Join) and Aggregate. Show less

In-class competition held on Kaggle (http://inclass.kaggle.com/c/how-s-the-weather)

Presented my experiments with multiple approaches, ranging from Ensemble of Learning to Deep Learning Networks on the weather data to predict the rainfall.

Results for weighted ensemble of learners landed me in top 10% of class, with the testing root-mean-squared error being 0.5741.

Researched the current popular approaches to enhance MANet (Mobile Ad hoc NETworks) performance through location services. A comprehensive qualitative comparison was performed and documented in the research paper.
Key concerns with the existing protocols were identified and listed as opportunities for future research.
Simulation tool: NS2

Designed, implemented and tested a real-time distributed database by leveraging tools and applying techniques. Used Oracle Database 10g Express Edition

Designed an efficient algorithm to assist a web crawler through prioritization and optimization of the input URL set. The algorithm was tested against URL Reputation Data Set (a set of 16,000 URLs with over 3 million attributes). The design included data selection, preprocessing, entropy calculations and finally, information gain comparisons to deliver the list of selected URLs, sorted in order of importance.
Development platform: C/C++, Matlab

Developed a document management system that personalizes the document search results to the user’s activities, making the search results much more relevant. The software reads the PDF files, extracts keywords, prepares an index, records user activity and delivers search results through a smart integration of all the information.
Development platform: Java (Using DynamicPDF class library)

Built the E-R model, data schema and database application for tracking goods at shipping ports and during transportation. The application enabled real-time monitoring and management of goods across various stages of transportation – right from pick-up to the ultimate delivery. Different user roles were created with varying access to data and management tools.
Development platform: MySQL, PHP, HTML, CSS

Designed and developed a simulator program that runs different CPU scheduling algorithms and produces utilization matrices including CPU utilization, waiting time of each process and average waiting time, response time of each process and average response time, turn-around time of each process and turn-around waiting time.
Development platform: C++