Uncategorized

Shedding Light on the Blackbox: Understanding Deep Neural Network Training Dynamics Authored by William Redman  Deep neural networks power everything from language models to image recognition, yet the process by which they learn remains one of modern science’s great mysteries. By viewing neural network training as a dynamical system, researchers are uncovering patterns and principles […]

AIMdyn Returns from 2nd Annual Koopman Operator Theory: Fundamentals, Approximations and Applications Workshop Official website of the workshop: https://uniri.hr/en/science-and-research/koopman-operator-theory-fundamentals-approximations-and-applications/ Workshop program: https://uniri.hr/wp-content/uploads/2025/09/10_9_Koopman_program_4_9_2025.pdf  From September 10 to 12, 2025, the Moise Palace on the island of Cres hosted the second international scientific workshop “Koopman Operator Theory: Fundamentals, Approximations and Applications.” The workshop is sponsored by the

Soft Robotics: From Nature-Inspired Flexibility to Real-World Applications  What if robots could wind around like growing vines, squeeze through rubble like octopuses, or gently assist patients with soft pads that detect pressure like a hand could? This is the promise of soft robotics, a field that challenges the rigid, metal-based foundation of robotics by studying

AIMdyn co-Founder, Dr. Igor Mezic, is in the 0.05 percent of the most influential scientists of all fields in 2023 and no. 54 in a subfield of engineering based on a list compiled annually by Stanford University using the data provided by Elsevier. The selection is based on the top 100,000 scientists (living or deceased)

New LAPACK version 3.12 has been released with AIMdyn’s high-performance implementation of Dynamic Mode Decomposition (DMD) with QR-factorization that provides considerable speedup in the case of high-dimensional data. The LAPACK library is a state-of-the-art library/software package for matrix computations. It is used as a computing engine in software tools such as MATLAB, Python, Octave, R.

NEW preprint out on Koopman Learning with Episodic Memory by the fantastic AIMdyn team (Dr. William Redman, Dean Huang, Dr. Maria Fonoberova, and Prof. Igor Mezic) and generously funded by the DARPA Small Business Innovation Research Program (SBIR) Program Office under Contract No. W31P4Q-21-C-0007 to AIMdyn, Inc. https://arxiv.org/abs/2311.12615 Koopman operator theory, a data-driven dynamical systems

Thank you to Dr. Allan Avila and the AIMdyn MINC team for their amazing work with BAE Systems on the DARPA MINC project!

The quote “’It’s tough to make predictions, especially about the future‘” is attributed to a baseball-playing philosopher, Yogi Berra. And yet, we always try, with more or less success.

In the latest contribution, the AI and Data Analytics company AIMdyn, Inc. compared one week ahead predictions of COVID-19 cases on the CDC COVID-19 challenge (https://covid19forecasthub.org/) made by a variety of prominent research groups.

Read More »

AIMdyn is so proud of all of our dedicated team members who were able to come to the first AIMdyn Summit and discuss the future of AI and the direction our company is taking. We are rapidly expanding thanks to our team’s hard work and unique set of skills and expertise. We thank our entire

Next week AIMdyn’s team of scientists will present their research at the SIAM Conference on Applications of Dynamical Systems (DS23) that will be held in Portland, OR. See below for brief descriptions of our team’s talks. Monday, May 15. 9:00-9:25 Koopman Reduced Order Modeling with Confidence BoundsRyan MohrAbstract. This talk introduces a reduced order modeling