About

I am a PhD candidate at MIT's EECS department, working on machine learning with Professor Marin Soljačić. I obtained my BSc in math and physics from MIT. I am a student at IAIFI.

News

3/14/2022: I will be participating at MIT x Japan's Innovation Discovery Japan and will be a student leader for the visit of SONY & Sony Interactive Entertainment.

1/31/2022: Our SkyRadar solution won the best pitch at MEMSI by offering a new digital or "phygital" Airport City experience. Read more about our experience on MIT News. Let's chat about innovation!

Research Interests

• machine learning, artificial intelligence, natural language processing, computer vision
• self-supervised learning, representation learning
• AI for science, AI hardware/ software co-design

Selected Publications

Learning with less labels

	On the Importance of Calibration in Semi-supervised Learning Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, Seungwook Han, Ligong Han, Leonid Karlinsky, Marin Soljacic, Akash Srivastava Preprint. Under review., 2022 paper / bibtex / Demonstating that calibration is important for semi-supervised learning, and a new method improving the state-of-the-art.
	Equivariant Contrastive Learning Rumen Dangovski, Li Jing, Charlotte Loh, Seungwook Han, Akash Srivastava, Brian Cheung, Pulkit Agrawal, Marin Soljacic ICLR, 2022 paper / bibtex / code / blog / talk Method revealing the complementary nature of invariance and equivariance in contrastive learning.
	DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings Yung-Sung Chuang, Rumen Dangovski, Hongyin Luo, Yang Zhang, Shiyu Chang, Marin Soljacic, Shang-Wen Li, Wen-tau Yih, Yoon Kim, James Glass NAACL, 2022 paper / bibtex / code Equivariant Contrastive Learning contributes to state-of-the-art results among unsupervised sentence representation learning methods.
	Meta-Learning and Self-Supervised Pretraining for Storm Event Imagery Translation Ileana Rugina*, Rumen Dangovski*, Mark Veillette, Pooya Khorrami, Brian Cheung, Olga Simek, Marin Soljacic, ICLR AI for Earth and Space Science Workshop, 2022 paper / bibtex / code Novel few-shot multi-task learning benchmark for image-to-image translation with meta-learning and self-supervsied learning baselines.
	We Can Explain Your Research in Layman’s Terms: Towards Automating Science Journalism at Scale Rumen Dangovski, Michelle Shen, Dawson Byrd, Li Jing, Desislava Tsvetkova, Preslav Nakov, Marin Soljacic AAAI, 2021 paper / bibtex Application automating science journalism at scale as a neural abstractive summarization task. Application constrained by little labeled pairs (scientific paper, press release).

Inductive biases

	Fast Neural Models for Symbolic Regression at Scale Allan Costa*, Rumen Dangovski*, Owen Dugan, Samuel Kim, Pawan Goyal, Joseph Jacobson, Marin Soljacic, Preprint. Under review., 2020 paper / bibtex / code Novel neuro-symbolic method for fast symbolic regression at scale.
	Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications Rumen Dangovski*, Li Jing*, Preslav Nakov, Mico Tatalovic, Marin Soljacic TACL (presented at NAACL), 2019 paper / bibtex / code / blog Novel recurrent unit with improved long-term and associative memory.

AI for Science

	Surrogate-and invariance-boosted contrastive learning for data-scarce applications in science Charlotte Loh, Thomas Christensen, Rumen Dangovski, Samuel Kim, Marin Soljacic Nature Communications, 2022 paper / bibtex / code Making applications in science less "hungry" for data.
	AI-Assisted Discovery of Quantitative and Formal Models in Social Science Julia Balla, Sihao Huang, Owen Dugan, Rumen Dangovski, Marin Soljacic, Preprint. Under review., 2022 paper / bibtex / code Discovering laws in social science using OccamNet, our neuro-symbolic method.
	Koopman Operator learning for Accelerating Quantum Optimization and Machine Learning Di Luo, Jiayu Shen, Rumen Dangovski, Marin Soljacic, Preprint. Under review., 2022 paper / bibtex Accelerating quantum optimization and quantum machine learning with Koopman operator learning.

AI Hardware/ Software Co-Design

	Vector-Vector-Matrix Architecture: A Novel Hardware-Aware Framework for Low-Latency Inference in NLP Applications Matthew Khoury*, Rumen Dangovski*, Longwu Ou, Preslav Nakov, Yichen Shen, Li Jing EMNLP, 2020 paper / bibtex Novel architecture for vector-matrix multiplication.
	Data-Informed Global Sparseness in Attention Mechanisms for Deep Neural Networks Ileana Rugina*, Rumen Dangovski*, Li Jing, Preslav Nakov, Marin Soljacic, Preprint. Under review., 2020 paper / bibtex / code Data-informed attention pruning.