Associate Research Scientist

Johns Hopkins Univeristy

Ben Haeffele

I am a research faculty member in the Johns Hopkins Mathematical Institute for Data Science (MINDS) and Center for Imaging Science (CIS). My current research is broadly on developing theory and algorithms for processing high-dimensional data at the intersection of machine learning, optimization, and computer vision. In addition to basic research in data science I also work on a variety of applications in medicine, microscopy, and computational imaging.


  • PhD in Biomedical Engineering

    Johns Hopkins University

  • BS in Electrical Engineering

    Georgia Institute of Technology

Research Interests

Machine Learning

Deep learning theory | Sparse and low-rank methods | Matrix/tensor factorizations | Subspace clustering | Generative models | Non-linear/manifold clustering | Physics constrained learning

Computer Vision

Image/Video segmentation | Image classification | Compressed sensing | Object detection | Multi-object tracking


Non-convex optimization | Low-rank problems | Convex relaxation/lifting methods


Lens-free imaging | Holography | Two-photon imaging

Biomedical Applications

Image analysis for microscopy | Hematology microscopy | Infection detection and monitoring | Digital pathology


A (very incomplete) selection of some of the topics I work on (past and present). More pages under construction.


Structured Matrix Factorization

Learning Matrix Factorizations with Structured Factors

Neural Network Loss Landscapes

Characterizing the loss surface of neural network training problems.

Understanding Dropout

Understanding the effects of Dropout regularization when training neural networks.

Recent Publications

A Critique of Self-Expressive Deep Subspace Clustering

Understanding the Dynamics of Gradient Flow in Overparameterized Linear models

A novel variational form of the Schatten-$ p $ quasi-norm

Doubly Stochastic Subspace Clustering

Generative optical modeling of whole blood for detecting platelets in lens-free images

On the Regularization Properties of Structured Dropout

Adaptive online $k$-subspaces with cooperative re-initialization