The Center for Applied Algorithms

At the Center for Applied Algorithms, our mission is to provide foundations for the future of computing through innovative, analytically-grounded techniques that are scalable, accurate, and efficient. Our goal is to stimulate multidisciplinary collaboration and make a critical impact in a multitude of applications such as computational medicine and biology, large-scale machine learning, computer systems, and energy. At the heart of our approach is the design of scalable algorithmic methods that provide tradeoffs between the competing notions of accuracy and performance, ultimately providing principled approaches to decision making in complex environments and achieving significant improvements in scalability, cost, and responsiveness.

Computational Medicine

We aim to demonstrate the power of applied algorithms for healthcare applications. Our current foci include algorithmic approaches for targeted cancer treatments, where we aim to combine multiple quantitative image-derived parameters to determine the best metrics for evaluating a treatment response in therapy, radiomic approaches for faster COVID-19 diagnostics, scalable tSNE and UMAP for cancer cell analysis, continual learning for landmark classification in medical imaging where the model can constantly learn and integrate novel features characterizing a specific disease, arrythmia classification from wearable data and algorithms for stroke prediction in heart failure patients with mechanical circulatory support.

Computer Systems

We aim to design algorithms and efficient data structures for improving large-scale networked systems that efficiently utilize limited data plane resources while performing multiple telemetric tasks. CAA will advance the state-of-the-art at the intersection of networking, machine learning, and algorithms for automated decision making. Current foci include intelligent network analytics, anomaly detection, traffic engineering, and security.

Scalable ML

Machine learning achieved phenomenal empirical successes. However, these results are based on a tremendous computation power and require hundreds of thousands of GPU hours or thousands of years of human-equivalent time to train. Such an enormous need of computational resources severely limits the applicability of ML. Algorithmic methods are a promising direction for scalable machine learning to reduce communication costs in distributed and federated learning while maintaining training accuracy, compressing deep neural networks by pruning “insignificant” neurons, improving selective plasticity methods in continual learning. Our recent theoretical and empirical results showing high accuracy with pre-training compression for specific classes of activation functions, and we intend to investigate the viability of the extension of these techniques. We will explore approximate methods for federated learning and analytics that balance accuracy with other critical constraints, such as inference costs, privacy, or fairness delivering principled methods for heterogeneous distributed networks that can flexibly trade off between a number of competing goals. We will apply algorithmic techniques to reduce model size of reinforcement learning and hence reduce the need for samples and training time.