Nikan Doosti (Experimental Page)

Through my research, I like to produce practical and accessible tools to be used by scientists, content creators, engineers, and end-users as a connecting media between different domains. I plan to do this by devising self-supervised physics-aware deep learning method. I am interested in computer graphics and applications of deep learning in computer graphics.

I am a MSc Computer Engineering graduate from Iran University of Science and Technology (IUST) and I am working as a machine learning engineer. Previously, I did a research internship in Artificial Intelligence aided Design and Manufacturing (AIDAM) group at Max Planck Institut für Informatik (MPI-INF). I worked on Structural Topology Optimization (TO) via Deep Learning (DL) under supervision of Vahid Babaei (AIDAM group leader) and in collaboration with Julian Panetta (Assistant Professor at University of California, Davis). We published this work in ACM Symposium on Computational Fabrication as my first-ever peer-reviewed publication. That was when I resolved my doubts about being a research scientist! Hence, I am looking for an unique PhD opportunity.

In my spare time (and quiet a few times before the deadlines :D), I play video games, read about history, watch standup comedy, and spam KEKW and OMEGALUL on Twitch for no obvious reason. On top of that, I like open source and I use Github as my main social media app.

news

Oct 23, 2022	I defended my master’s thesis with full grade at Iran University of Science and Technology.
Apr 24, 2022	I joined Nahal Gasht as a full-time machine learning engineer.
Mar 4, 2022	I gave a talk at Toronto Geometry Colloquium about my SCF21 publication.
Oct 1, 2021	We published and presented a peer-reviewed paper in SCF21 conference!

selected publications

Topology Optimization via Frequency Tuning of Neural Design Representations

Nikan Doosti, Julian Panetta, and Vahid Babaei

In Proceedings of the 6th Annual ACM Symposium on Computational Fabrication 2021

Abs Bib HTML PDF Code Poster Slides

Structural topology optimization seeks to distribute material throughout a design domain in a way that maximizes a certain performance goal. In this work, we solve the topology optimization problem by parameterizing the designs via recently introduced coordinate-based neural networks. Specifically, we show that networks with Fourier feature mapping can achieve state-of-the-art performance. Our method enables the realization of a range of designs using a single mesh via tuning the frequency content of the solutions independently of the finite element discretization grid. This frequency control offers attractive properties, such as mesh-independent results and sub-pixel filtering that leads to appropriate designs for upsampling. We demonstrate our method on the compliance minimization problem, optimizing for the stiffest possible structure within a weight budget for a prescribed set of loads.
@inproceedings{10.1145/3485114.3485124, author = {Doosti, Nikan and Panetta, Julian and Babaei, Vahid}, title = {Topology Optimization via Frequency Tuning of Neural Design Representations}, year = {2021}, isbn = {9781450390903}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3485114.3485124}, doi = {10.1145/3485114.3485124}, booktitle = {Proceedings of the 6th Annual ACM Symposium on Computational Fabrication}, articleno = {1}, numpages = {9}, keywords = {neural networks, generative design, Topology optimization}, location = {Virtual Event, USA}, series = {SCF '21} }