The Data Intelligence and Computational Engineering Lab at Iowa State University is an inter-disciplinary group of researchers who develop novel machine learning techniques from a foundational perspective, with focus areas spanning optimization algorithms, information theory, and statistics.

## News

• March, 2019
Released a new dataset:Binary 2D morphologies of polymer phase separation along with a python toolbox providing API and utilities to aid the usage of our dataset. Documentation is available here.
• December, 2018
Received a DARPA AIE grant for a project on physics-aware machine learning (along with Soumik Sarkar and Baskar Ganapathysubramaniam).
• December, 2018
Paper on autoencoder training accepted to AISTATS 2019.
• November, 2018
Received an ERP grant for a project on adversarial machine learning (along with Soumik Sarkar and Anuj Sharma).
• November, 2018
New paper on physics-aware deep generative models posted on ArXiv.

## Research

#### Robustifying Machine Learning under Semantic Constraints

We explore the space of adversarial examples in terms of semantically valid images. Our approach relies on the use of generative models to simulate the semantic transformations of images.

#### Physics-aware Deep Generative Models for Creating Synthetic Microstructures

One of the key goals in engineering is to design systems (in this case materials) that achieve a desired property. In this project, we integrate classical engineering approaches (i.e., physics models) with machine learning models such as generative models to transform and accelerate such design exploration process.

#### Reconstruction from Periodic Nonlinearities, with applications in HDR imaging

Our aim is a reliable estimation of a signal or image from its periodic nonlinearities, with a focus on a periodic nonlinear observation model named modulo sensor encountered in high-dynamic range (HDR) imaging.

#### Solving inverse problems using GAN priors

We leveraged the ability of Generative Adversarial Networks (GANs) of learning the real data distribution by using the Generator function as a prior on natural images.

#### Provable Algorithms for training ReLU networks

We prove linear convergence of both gradient descent and a new scheme called alternating minimization for training ReLU based 2-layer networks.

#### Sparse image super resolution

In our previous work on structured phase retrieval, we noted the advantages of incorporating a sparsity constraint in the phase retrieval algorithm (we call our algorithm CoPRAM). The advantages of this are two fold: fewer number of samples are required for efficient recovery and the recovery procedure is also computationally much faster.

#### Phase retrieval of structured signals

We consider the problem of recovering a signal from magnitude-only measurements. This is a stylized version of the classical phase retrieval problem, and is a fundamental challenge in nano- and bio-imaging systems, astronomical imaging, and speech processing.

#### Fast Low-Rank Estimation for Ill-Conditioned Matrices

In this paper, we study the general problem of optimizing a convex function $F(L)$ over the set of $p\times p$ matrices, subject to rank constraints on $L$.

#### Fast and Provable Algorithms for Learning Two-Layer Polynomial Neural Networks

We study the problem of (provably) learning the weights of a two-layer neural network with quadratic activations.

#### Demixing Structured Signal form their Nonlinear Superposition

We study the problem of \emph{demixing} a pair of sparse signals from noisy, nonlinear observations of their superposition.

#### Towards Deeper Theoretical Understanding for Unsupervised Learning

Unsupervised learning

## People

### Faculty

##### Chinmay Hegde
Assistant Professor, ECPE