I am from Thane, a city that’s next to Mumbai (Like New Jersey and New York). I am currently pursuing an integrated bachelor’s and master’s in physics at the Center for Excellence in Basic Sciences, Mumbai. I am interested in taking up a research career in biophysics, but I also want to expose myself to different schools of research and gain as much experience and knowledge as I can. I like to travel, especially in remote areas, and trekking is my favorite hobby. I also like learning about different cultures and languages.
Project: Visual Understanding of Artificial Neural Networks
I have taken the built-in neural network ImageIdentify in Mathematica and extracted all of its weights from its convolution layers. I have also visualized the activation of weights by passing an image through modified versions of the neural network. To visualize which kinds of images maximally activate the layers of a neural network, I have passed images through a neural network designed by Google and observed which features of images are enhanced as they pass through it, at each layer. These distortions have been coined as inceptionism effects by Google.
Convolving an image with weight from neural network, an example of inceptionism, and the visualization of activation of neurons upon passing an image through a neural network cut short: