1. Forecasting Food Sales in a Multiplex using Dynamic Artificial Neural Networks.
This work comprises of online learning and feature engineering by data correlative analysis in conjecture with densely connected Neural Network to address the concept drifts and latent time correlations present in the data respectively, to make day-ahead prediction of food sales for one of the top multiplexes in India. It has been published at the Computer Vision Conference (CVC) 2019, held at Las-Vegas, Nevada, USA. Link
2. Location Based Species Recommender System.
Prediction of top k species in a location based on the environmental features given in the form of tiff images. Devised a whole new architecture to factor in the concept of hierarchy, using embeddings. Done as part of GeoCLEF part of CLEF 2018 at Avignon, France. The work has been presented as working notes at the Conference and Labs of the Evaluation Forum (CLEF) 2018. Link
3. DeepTrace : Generic Deep framework for Cross-Domain Univariate and Multivariate Time Series Forecast.
An architecture and it’s variants that eliminate the need for data preprocessing and feature engineering, and are also capable of forecasting for any time series dataset across domains, by considering these time series as being auto-correlative. It had been submitted to “Assocation for the Advancement of Artificial Intelligence” (AAAI) Conference 2019, and we incorporated the reviews that we had received. It has been accepted to be presented at the International Work Conference on Artificial Neural Networks (IWANN) 2019, to be held at Gran Canaria, Spain, during June 12 - june 14, 2019.
4. Microsoft Summer Research Workshop 2018.
As part of the MSR Summer Workshop 2018 on ‘Machine Learning on Constrained Devices’, worked on extending two algorithms Bonsai and ProtoNN for regression. Worked on the Tensorflow version of the edgeML library of MSR, for extending these algorithms. I was also a part of the team that jointly won the MSR research grant. All the code and documentation can be found here. Link