SciWorcs is a
Collaborative Scientific Workflow management System. SciWorcs follows a plugin-based architecture for the
scientific computational modules. It provides collaborative environment for scientific data analysis using
efficient attribute level locking scheme. In addition to the collaborative data analysis, SciWorcs also provides
real time monitoring of the computation steps.
iSeaborn is an interactive version of the popular statistical data visualization library - Seaborn. iSeaborn is
build on top of Seaborn (i.e., for advanced statistical computation) and Bokeh (i.e., for interactive plots). The
library
provides similar API as the popular Seaborn visualization library, so that it requires minimal effort to adapt and
get
interactive visualizations.
A code clone is a pair of code fragments, within or between software systems that are similar. This is a machine
learning based framework for automatic code clone validation - developed based on our recent research study.
The method learns to predict tasks or user-specific code clone validation patterns. The current machine learning
model has been build based on BigCloneBench - a collection of eight million validated clones
A code clone is a pair of code fragments, within or between software systems that are similar. This is a machine
learning based framework for automatic code clone validation - developed based on our recent research study.
The method learns to predict tasks or user-specific code clone validation patterns. The current machine learning
model has been build based on BigCloneBench - a collection of eight million validated clones
Given a character sequence
comprising of 10 characters: How are yo, the model is expected to predict the next possible character, such as u,
completing the sequence as How are you. The same process can be repeated with latest characters, such as last 10,
to generate further characters. pyPoet is a such a language to automatically generate relevant text based on a
sentence.
This is a sequence to sequence learning problem where the RNN model learns to add two numbers (as string). This is
a simple example where, the input and output length are always same, unlike other language models or sequence
problems. To always make sure that, the input is of a fixed length MAXLEN, the required amount of (space)
characters are padded.
Time Series Forecasting on the Shampoo Sale Dataset. The dataset contains information of monthy shampoo sales for
consecutive months. We train on the first few months using LSTM. We then evaluate the model for shampoo sales
forecasting on the remaining months of the dataset.
This is a Deep Learning based project for Machine Translation. In this project, I have been specifically working
on 'German to English' and 'Bengali to English' automatic language translation system.
Frequently used and Useful Visualizations for Data Science using Seaborn. This project is targeted to assist
the Exploratory Data Analysis quickly to find important insights.
CloneMapR -- is anear-miss code clone detection framework from massive amount of dataset. CloneMapR uses
approximation algorithm for estimating near-miss code clones while efficiently handling exhaustive pair-wise
comparisons in big data considerations. In addition to the efficient code clone approximation algorithm, we
exploit parallel computation using MapReduce programming model to accelerate the computation process.
Code clones are considered harmful, studies show that the existence of
some of them can inflate software maintenance costs as clones are one of the major
causes of the creation and propagation of software bugs throughout the system. uvClones
is a machine learning based code clone detection tool. uvClones uses Artificial Neural Network
model for automatically detecting and reporting the possible code clone pairs in a given
software repository.
This is a Finite State Machine Framework for game AI. This framework can be used for state machine based game
agents implementation. The framework allows group behavior of the agent using message passing abstraction among
the state machines of the agents. The framework is designed following Mat Buckland's book - 'Programming Game AI
By Example' (https://www.amazon.ca/Programming-Game-Example-Mat-Buckland/dp/1556220782)
Do people's commute mode necessarily changes with weather condition for a particular day? or What is the arrival
paradigm and regularity of arrival time of the students coming to the university? For answering them, we conducted
a survey for 29 days involving 105 participants. The obtained results were then tested via some statistical
testing for validity. By measuring the entropy rate of each arrival time to the campus which is a novel technique,
we have found bus commuters are more regular concerning the both entire day and the working hours.
BCS Flashlight is a software developed for Bangladesh Civil Service (BCS) exam preparation. It helps by providing
up to date reading contents for the exam preparation. Besides, the software helps the reader by monitoring their
progress and smartly recommends the reading contents based on previous responses.
coderBlock is a project for hosting online programming contests. Using coderBlock, it is possible to host
programming contests with a set of programming problems and participants. Based on the submitted solution codes it
calculates the real-time ranking of the participants or groups. The automatic judging program was written in C.
coderBlock uses Java for handling security attacks or vulnerable codes.
obstaclX is a 2D platformer game written in Java language. The animation class and physics of the game components
were written from scratch in Java. The components graphics rendering scripts uses the JavaFX.
I developed this game in 48 hours in a Game Jam - held at the University of Saskatchewan. This game was awarded as
the winnner in the audio-visual category in the Game Jam.
digitalDiary is a project for creating and managing individual's notes. The notes are encrypted for secure
storing. The project was written in C++ language in Qt Creator.