My current research is on Selectivity, Discriminability and Ablation in Deep neural network.

Deep learning approach in the field of artificial intelligence emphasizes high capacity, scalable models that have the capability to learn distributed representations from the input dataset. Although these models have outperformed humans as measured on the datasets in question, how they are able to do so remains unclear. The aim is to get better clarity on the selectivity, discriminability, and ablation of neurons of neural networks in order to understand deep learning models.

Thesis:

Deep Learning based frameworks for cyber security applications, M.Tech, 2020.

Artificial Intelligence-based smart mirror, B.E, 2018.

MOOC’s completed:

● Machine Learning on Coursera.

● Introduction to TensorFlow for Artificial Intelligence, Machine learning and deep learning on Coursera

● Deep learning Specialization on Coursera containing Neural Networks and Deep Learning, Improving Deep Neural Networks Hyperparameter tuning, Regularization and Optimization, Structuring Machine Learning Projects, Convolutional Neural Networks, and Sequence Models.

Acadamic Projects Undertaken:

● Different Word Embedding Algorithms For Part-of-Speech Tags (Sanskrit, Hindi, and Telugu): A comparative study using different word embeddings like word2vec, fastText, and BERT with three case studies namely, Sanskrit, Hindi, and Telugu. Used different machine learning algorithm for evaluation.

● Fingerprint Classification: Binary classification of fingerprint data set which classifies a finger into Live and Fake. A comparative analysis is done using various pre-trained architectures.

● Human identification by Ear Recognition: Identification of humans with their respective ear alone using deep learning architecture by applying data augmentation on mobileNetV2 architecture. Achieved an accuracy of 90% with 155 subjects.