Scholar
T. Anderson Keller
Google Scholar ID: Tb86kC0AAAAJ
Research Fellow, Kempner Institute at Harvard University
Citations & Impact
Citations
293
H-index
10
i10-index
10
Publications
20
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- Published several research papers covering topics from natural inductive biases for AI to using traveling waves to improve neural network performance.
- - Natural Inductive Biases for Artificial Intelligence (PhD Thesis)
- - Improving the Mamba architecture through variable velocity traveling waves
- - Proposed a method based on relative representations to optimize model-to-brain mappings
- - Explored how traveling waves can be used to encode temporal information, improving performance in sequence learning tasks
- - Developed the DEUT framework for generating 2D representations with approximate equivariance
Research Experience
- As a Research Fellow at Harvard University, involved in multiple projects related to neural networks and machine learning, including but not limited to:
- - Understanding the Convolutional Layer in State Space Models
- - A Spacetime Perspective on Dynamical Computation in Neural Information Processing Systems
- - Relative Representations for Model-to-Brain Mappings
- - Natural Inductive Biases for Artificial Intelligence (PhD Thesis topic)
- - Image Segmentation with Traveling Waves
- - Flow Factorized Representation Learning
- - Traveling Waves Encode the Recent Past and Enhance Sequence Learning
- - DEUT -- 2D Structured and Approximately Equivariant Representations
Background
- Research Fellow at Harvard University. Research interests include understanding the convolutional layer in state space models, a spacetime perspective on dynamical computation in neural information processing systems, and application of relative representations for model-to-brain mappings, among others.
Miscellany
- Active on social media platforms such as Twitter and GitHub, with academic publications available on Google Scholar.
Co-authors