LinkedInGoogle Scholar

Hello! I'm a machine learning researcher in NYC currently working as a software engineer at Google AI. Before graduating in 2018, I was a Ph.D. student at the Cornell Tech SE(3) Vision Group under Dr. Serge Belongie. I previously worked as a research assistant in the Vision and Security Technology lab and as a software engineer at Securics, Inc., both under Terrance E. Boult.

My hobbies include peer counseling, bouldering (still a beginner, VB-V1), lifting (beginner), hiking, and finding delicious quiet places in the city to drink tea and read books.

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/e28994c3-d5db-4346-9622-b28809ed483e/Untitled.png

All Publications

My full publication list is available on my CV or my **Google Scholar Profile Page.**

Download CV

CV kwilber Fall 2019.pdf

Selected Publications and Projects

Note that some work before 2018 is listed under a previous name.

Online Trust

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/f3510726-6e7d-4f78-8806-694b7bcb164e/Untitled.png

We created a system that can understand the aesthetic quality of user-created photos on online marketplaces like Letgo or eBay. This was a collaboration between Cornell Tech, eBay, and the Oath connected experiences lab.

Artistic Aesthetics

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/2cf03321-51ec-4810-bf38-3f2aea5fbc30/Untitled.png

We taught a computer about artwork! Our efforts led to the creation of "BAM," currently the largest semisupervised dataset of digital artwork on the Internet freely available for researchers.

Deep learning theory

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/259977ba-4d02-46e6-bc02-d38a4df03e24/Untitled.png

What happens when you perform brain surgery on a ResNet? Surprisingly, performance still stays the same when deleting several layers, even without fine-tuning. We investigate why in this paper.

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/8ce16b13-1f7c-4af6-871f-2e42ceda9ebc/Untitled.png

On flagship datasets like LFW and Caltech-256, the current top performing algorithms at time of writing do not satisfy the triangle inequality, symettricity, or even identity. Why are top-performing algorithms non-metric?