Pranav Mody

Observed following network properties using pySpark, GraphFrames and networkX :

1) Implemented a function degreedist, which takes a GraphFrame object as input and computes the degree distribution of the graph. The function returns a DataFrame with two columns: degree and count.

2) Implemented a function closeness, which takes a GraphFrame object as input and computes the closeness centrality of every node in the graph. Determined the nodes that are important using this… Show more

Observed following network properties using pySpark, GraphFrames and networkX :

1) Implemented a function degreedist, which takes a GraphFrame object as input and computes the degree distribution of the graph. The function returns a DataFrame with two columns: degree and count.

2) Implemented a function closeness, which takes a GraphFrame object as input and computes the closeness centrality of every node in the graph. Determined the nodes that are important using this output.

3) Implemented a function articulations, which takes a GraphFrame object as input and finds all the articulation points of a graph. Articulation points, or cut vertices, are vertices in the graph that, when removed, create more components than there were originally Show less

Expected to develop and compare three different deep neural network (DNN) architectures for the task of online video classification. Expected to propose different architectures to explore for the task.
Examples include but not limited to:
• optical flow/CNN fusion
• one frame at a time ConvNet
• ConvNet features passed to LSTM RNN
• ConvNet features passed to MLP
• 3D CNN

o May choose TFLearn, Keras, or PyTorch for implementation
o Implementation will be Python-based

Reference paper : http://arxiv.org/pdf/1410.1231.pdf

In this project, predicted the price variations of bitcoin, a virtual cryptographic currency.
Used machine learning technique, Bayesian Regression, and implemented this technique in Python.
The raw data used can be found here : http://api.bitcoincharts.com/v1/csv/. Performed some cleaning on this raw data.
Then implemented the algorithm provided in the above mentioned reference paper

A simple game coded in C which expects you memorize content displayed on screen and difficulty increases after every round. Also added features like : extra life on 5 correct answers, high scores and statistics.

Given a set of advertisers each of whom has a daily budget 𝐵𝑖. When a user performs a query, an ad request is placed online and a group of advertisers can then bid for that advertisement slot. Maximized the amount of money received from the advertisers in this Project. Implemented Greedy, MSVV and Balance algorithm. Selected the best algorithm based on the competitive ratio.

Divided a broad target market into subsets of consumers or businesses that have or are perceived to have common needs, interests, and priorities. Implemented a community detection algorithm for attributed graphs, found the relevant market segments, and evaluated the obtained segments via influence propagation.
Algorithm used was SAC-1 which is described in this research paper : Community Detection based on Structural and Attribute… Show more

Divided a broad target market into subsets of consumers or businesses that have or are perceived to have common needs, interests, and priorities. Implemented a community detection algorithm for attributed graphs, found the relevant market segments, and evaluated the obtained segments via influence propagation.
Algorithm used was SAC-1 which is described in this research paper : Community Detection based on Structural and Attribute Similarities
(https://www.thinkmind.org/download.php?articleid=icds_2012_1_20_10025) Show less

Apache Spark, Python
• Used publicly available song data from audioscrobbler, applied collaborative filtering techniques and splitted the data into training, validation and testing sets. Trained the model with implicit feedback and then by performing a parameter sweep, chose the model that performs best on the validation set. Used this selected model to recommend music.

Spark Streaming APIs, Kafka, Python
• Performed basic sentiment analysis on live data stream of realtime tweets using Spark Streaming APIs. Used Apache Kafka to buffer the tweets before processing. Generated a plot showing count of positive and negative tweets at every time step.

(Python, Tesseract OCR)
• Technologies used were Image recognition and kNN algorithm for classification. Achieved an accurarcy of around 40%
• Integrated Google’s tesseract OCR engine with our application.

• Designed a multi-client system for taking feedback and used a centralized database for storing and processing it.
• Designed a database with logical relations between subject, teacher, class division and students thus allowing us to automatically load the concerned teachers and subjects for a particular student based on Roll number.
• Evaluated the feedback results by drawing comparison graphs and creating a summarized report.