About Me

I am a Machine Learning Scientist at Harmonic Discovery. I completed my PhD at UC Berkeley under Michael Mahoney, where I worked on fundamental research in machine learning. Before Berkeley, I was a student at Arizona State University, where I studied mathematics and economics, and completed a master’s thesis under Sebastien Motsch in the mathematics department. I was also the machine learning lead in the Luminosity Lab. Outside of university, I spent two summers as a research intern at Salesforce Research, and have also worked for Amazon.com in the past. See my CV for more.

Research

During my PhD, I was primarily interested in theoretical aspects of deep learning, mainly focused on the question of generalization. Namely, I’m interested in when and how we can fit extremely complicated and expressive models to data, and expect these models to perform well on new, unseen data.

More recently, I have become interested in applications of machine learning to the field of drug discovery.

Below is a list of papers I’ve co-authored.

• Kim, H., Hodgkinson, L., Theisen, R., Mahoney, M.W. How many classifiers do we need? Conference on Neural Information Processing Systems, 2024.

• Theisen, R., Wang, T., Ravikumar, B., Rahman, R., Cichonska, A., Leveraging multiple data types for improved compound-kinase bioactivity prediction. Nature Communications, 2024.

• Park, R., Theisen, R., Sahni, N., Patek, M., Cichonska, A., Rahman, R. Preference optimization for molecular language models. NeurIPS Workshop on Generative AI and Biology, 2023.

• Theisen, R., PhD Thesis advised by Michael W. Mahoney. Beyond Worst-Case Generalization in Modern Machine Learning, 2023.

• Theisen, R., Kim, H., Yang, Y, Hodgkinson, L., Mahoney, M.W. When are ensembles really effective? Conference on Neural Information Processing Systems, 2023.

• Yang, Y., Theisen, R., Hodgkinson, L., Gonzalez J.E., Ramchandran, K., Martin, C.H., Mahoney, M.W. Test Accuracy vs. Generalization Gap: Model Selection in NLP without Accessing Training or Testing Data. ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023.

• Yang, Y., Hodgkinson, L., Theisen, R., Zou, J., Gonzalez J. E., Ramchandran, K., Mahoney, M.W. Taxonomizing Local Versus Global Structure in Neural Network Loss Landscapes. Conference on Neural Information Processing Systems, 2021.

• Theisen, R., Wang, H., Varshney L. R., Xiong, C., Socher, R. Evaluating State-of-the-Art Classification Models Against Bayes Optimality. Conference on Neural Information Processing Systems, 2021.

• Theisen, R., Klusowski, J. M., Mahoney, M. W. Good Classifiers are Abundant in the Interpolating Regime. International Conference on Artificial Intelligence and Statistics, 2021.

• Cao, F., Motsch, S., Reamy, A., Theisen, R. Asymptotic Flocking for the Three-Zone Model. Mathematical Biosciences and Engineering, 2020.

• Theisen, R., Klusowski, J. M., Wang, H., Keskar, N., Xiong, C., Socher, R. Global Capacity Measures for Deep ReLU Networks via Path Sampling. Technical Report, 2019.

• Weber, D., Theisen, R., Motsch, S. Deterministic Versus Stochastic Consensus Dynamics on Graphs. Journal of Statistical Physics, 2019.

• Theisen, R., Masters Thesis advised by Sebastien Motsch. Convergence Results for Two Models of Interaction, 2018.

Photography

I love to travel and spend as much time outdoors as possible. I always take my camera with me wherever I go - feel free to check out some of my photos.

Contact

Email: ryanctheisen [at] gmail.com