Yeonhyang Kim (kim4y AT cmich DOT edu)
Leela Rakesh (leela.rakesh AT cmich DOT edu)
Xiaoming Zheng (zheng1x AT cmich DOT edu)
Date |
Speaker |
Title |
9/8/23 |
TBA |
|
9/15/23 |
TBA | TBA |
9/22/23 |
TBA |
TBA |
9/29/23 |
Hiruni Pallage (CMU) |
Recovery of Initial conditions through later time samples |
10/6/23 |
TBA |
TBA |
10/13/23 |
TBA |
TBA |
10/20/23 |
TBA |
TBA |
10/27/23 |
TBA |
TBA |
11/3/23 |
TBA |
TBA |
11/10/23 |
TBA |
TBA |
11/17/23 |
TBA |
TBA |
12/1/23 |
TBA |
TBA |
12/8/23 |
Hongsong Feng (MSU) |
Mathematics-AI for drug discovery and drug addiction studies |
Speaker: Hiruni Pallage
Title: Recovery of Initial conditions through later time samples
Abstract: Full knowledge of the initial conditions of an initial value problem (IVP) is necessary to solve said IVP but is often impossible in real-life applications due to the unavailability or inaccessibility of a sufficiently large sensor network. One way to overcome this impairing is to exploit the evolutionary nature of the sampling environment while working with a reduced number of sensors, i.e., to employ the concept of dynamical sampling. A typical dynamical sampling problem is to find sparse locations that allow one to recover an unknown function from various times samples at these locations. The classical problem of inverse heat conduction has been recently revisited by Devore and Zuazua (2014). We study conditions on an evolving system and spatial samples in a more general setup using bases. Specifically, we study when $u(x, t) = \sum_{n=1}^\infty a_nf_n(x) g_n(t)$, where $a_n \in \mathbb{R}$, $x \in [0, 1]$, $t \in [0, \infty)$ can be reasonably approximated through later-time samples at a single sampling location. The results of our research are relevant in applications, and we present examples of solving the Laplace equation and variable coefficient wave equation using our general method. Kim and Aceska (2021) retrieved the unknown initial condition function of the above systems via exponentially growing samples. However, in the approximation process, they observed exponential growth in error terms of the coefficients. Our recent research demonstrates that we can incorporate a linear growth pattern of errors in the recovered coefficients in these systems.
Speaker: Hongsong Feng
Title: Mathematics-AI for drug discovery and drug addiction studies
Abstract: Traditional drug discovery is a time-consuming and costly process, often taking over a decade to bring a new drug to market. AI has great potential to transform the entire landscape of pharmaceutical research and developments. It is promising to combine mathematics with AI to promote the drug discovery development. Due to the importance of accurate predictions of binding affinity between proteins and drug-like compounds, there is pressing need for reliable mathematical predictive models. Models based on persistent Laplacian (PL) theory were proposed in this aspect by our research group and can serve as a useful tool for binding affinity predictions. In my talk, I will present our recent machine learning studies with persistent Laplacian models and natural language processing (NLP) methods for drug discovery and drug addiction problems.
Yeonhyang Kim (kim4y AT cmich DOT edu)
Leela Rakesh (leela.rakesh AT cmich DOT edu)
Xiaoming Zheng (zheng1x AT cmich DOT edu)
If you would like to give a talk, please email any one of us.
Fridays, 2:00pm – 3:00pm, on Webex
Date |
Speaker |
Title |
1/20/23 |
TBA |
TBA |
1/27/23
|
Mohsen Zayernouri (MSU) |
Data-Driven Fractional Modeling, Analysis, and Simulation of Anomalous Transport & Materials |
2/3/23 |
TBA |
TBA |
2/10/23 |
TBA |
TBA |
2/17/23 |
TBA |
TBA |
2/24/23 |
TBA |
TBA |
3/3/23 |
TBA |
TBA |
3/17/23 |
TBA |
TBA |
3/24/23 |
TBA |
TBA |
3/31/23 | Andrea Liu (University of Pennsylvania) |
Machine Learning Glassy Dynamics |
4/7/23 |
TBA |
TBA |
4/14/23 |
Kyle Harshbarger |
Scheduling and Planning High Uncertainty Seasonal Products at Dow |
4/21/23 |
|
TBA |
4/28/23 |
TBA |
TBA |
Speaker: Mohsen Zayernouri
Title: Tittle: Data-Driven Fractional Modeling, Analysis, and Simulation of Anomalous Transport & Materials
Abstract: The classical calculus and integer-order differential and integral models, due to their inherently local characters in space-time, cannot fully describe/predict the realistic nonlocal and complex nature of the anomalous transport phenomena. Nature is abundant with such processes, in which for instance a cloud of particles spreads in a different manner than traditional diffusion. This emerging class of physical phenomena refers to fascinating processes that exhibit non-Markovian (long- range memory) effects, non-Fickian (nonlocal in space) interactions, non- ergodic statistics, and non-equilibrium dynamics. The phenomena of anomalous transport have been observed in a wide variety of complex, multi-scale, and multi-physics systems such as: sub-/super-diffusion in subsurface transport, kinetic plasma turbulence, aging polymers, glassy materials, in addition to amorphous semiconductors, biological cells, heterogeneous tissues, and fractal disordered media. In this talk, we present a series of recent spectral theories and global spectral methods for efficient numerical treatment of fractional ODEs/PDEs. Finally, a number of applications including fluid turbulence, power-law rheology, and material failure processes will be also presented, in which fractional modelling emerge as a natural language for high-fidelity modelling and prediction.