Machine Learning Engineer
Founder & Researcher
Laboratory for Space Research,
The University of Hong Kong
Email, Scholar, GitHub
I work on topics in continual learning, attention mechanisms, robustness and non-linear dynamics,
with a broad focus on theoretical deep learning, lifelong learning, representation learning and computer vision.
My previous work has mostly been at the intersection of dynamics and applied computer vision. I also work at the intersection of computer vision and astrophysical sciences.
Along with the applied deep learning aspects, I have a firm interest in abstract algebra and algebraic geometry.
Some recent topics:
As founder and researcher at my group Landskape, I collaborate and undertake research projects mostly at the intersection of analytical theory of deep learning and applied computer vision. Please visit our page for more details.
Prior to this, I completed my undergraduate in Electronics and Electrical Engineering (EEE) from Kalinga Institute of Industrial Technology (KIIT) under the guidance of Asst. Prof. Dr. Bhargav Appasani.
Collaboration Opportunities and Referrals:
At Landskape, we are open to new collaborations and research members for our selected ongoing projects. Please take a look at our projects to have an idea on the scope and domain of our current projects. To reach out to us, please fill in the contact form available on our page. Additionally, if you're interested in applying for any position at Weights & Biases, please do so via the official portal. Please do not contact me for referrals. You can take a look at the open positions at Weights & Biases here.
Recent projectsPublications listed below, in the present page.
Mish: A Self Regularized Non-Monotonic Neural Activation Function We propose a novel non-monotonic non-linear activation function known as Mish which outperforms conventional activation functions like ReLU and Swish. We further provide a highly optimized CUDA version of Mish along with linking the effect of non-monotonic smooth activation functions to the loss landscapes of deep neural networks.
Rotate to Attend: Convolutional Triplet Attention Module We propose a process that captures Cross-Dimenional Interaction (CDI). Using CDI as a foundation, we propose a novel attention mechanism for deep convolutional neural networks known as Triplet Attention.
Collaborators: Trikay Nalamada (Landskape), Ajay Uppili Arasanipalai (Landskape), Qibin Hou (NUS)
Echo Echo is an OSS deep learning package with support for TensorFlow, PyTorch and MegEngine, containing novel validated methods, components and building blocks used in deep learning.
Collaborators: Alexandra Deis (X the Moonshot Factory), Soumik Rakshit (IBM), Ajay Uppili Arasanipalai (Landskape), Sasikanth Kotti (TCS)
Robustness-Stability-Plasticity Trilemma We are investigating into the effects of Continual Learning methods on the adversarial and OOD robustness of neural networks. We aim to formulate a novel understanding of optimal lifelong learning with the preservation of generalization and robustness.
Collaborators: Himanshu Arora (Landskape, MILA), Vincenzo Lomonaco (Continual AI), Norman Di Palo (Imperial College London)
Factorized Super Resolution We propose a novel blind super resolution framework aimed at improving consistency and reducing artefacts generation.
Collaborators: Himanshu Arora (Landskape, MILA), Sanghun Jung (Landskape, KAIST) and Jaegul Choo (KAIST)
- January 2021: Our WACV paper's video is now out on YouTube. Watch it here.
- January 2021: I will be speaking at the W&B Deep Learning Salon on "From Smooth Activations to Robustness to Catastrophic Forgetting". I will be joined by Maithra Raghu from Google Brain. Watch it here.
- December 2020: I'm starting full time as a Machine Learning Engineer at Weights & Biases.
- October 2020: Our paper Rotate to Attend: Convolutional Triplet Attention Module is accepted to WACV 2021.
- September 2020: Gave a talk on my paper on Mish at the Robert Bosch Bangalore Research Office.
- August 2020: I completed my Undegraduate degree in Electronics and Electrical Engineering from Kalinga Institute of Industrial Technology (KIIT).
- August 2020: Gave a talk on Mish and Non-Linear Dynamics at Computer Vision Talks. Watch here.
- July 2020: My paper Mish: A Self Regularized Non-Monotonic Neural Activation Function is accepted at BMVC 2020.
- July 2020: CROWN: A comparison of morphology for Mish, Swish and ReLU produced in collaboration with Javier Ideami. Watch here.
- May 2020: Participated in an AMA for my paper on Mish at the Weights & Biases reading group.
- April 2020: Presented my views and discussed about Data Science on the The World is Ending Podcast. Listen to the episode here.
- February 2020: Talk on Mish and Non-Linear Dynamics at Sicara is out now. Watch here.
- February 2020: Podcast episode on Mish at Machine Learning Café is out now. Listen here.
- November 2019: Presented a talk on my paper on Mish at the University of Athens.
Mish: A Self Regularized Non-Monotonic Neural Activation Function
BMVC, 2020 [pdf][code][CV Talk Episode][ML Cafe Episode][Sicara Talk][Weights & Biases Salon Episode]
Rotate to Attend: Convolutional Triplet Attention Module
Diganta Misra, Trikay Nalamada, Ajay Uppili Arasanipalai, Qibin Hou
WACV, 2021 [pdf][supp][video][code]