Joint Seminar, Division of Engineering, Microfluidics Laboratory and
The Center for Fluid Mechanics
Abstract: Over the past decade, the demand for liquid handling of minute sample volumes has dramatically increased. This trend towards smaller volumes has on the one hand been incited by the rapidly increasing achievements in micro- and nanotechnology. It has on the other hand often been driven by the need to improve existing applications or to create new ones in the fields of health care, clinical diagnostics and life sciences. One of the important liquid-handling aplications in the environment of the in-vitro diagnostics regards the pipetting of small sample volumes. Thereby, a defined volume is first aspirated from a sample reservoir, then transported by the means of a robotic system and finally dispensed into a different liquid, e.g. a reagent.
The talk presents the new micropipetting system with integrated micromachined sensors, which can in one device both aspirate and dispense liquid volumes in the micro- and submicroliter range and which can furthermore displace the required volume within one stroke and high precision. Moreover, the micropipetting system is designed to work with the long and liquid-filled needles, which are routinely used in the laboratory environment of in-vitro diagnostics systems. With this pipetting system, we have successfully pipetted water between 0.5$\mu$1 and 2$\mu$1 with a coefficient of variation of down to 1\% (after two pipetting cycles). Two patents could be filed during this research project together with our industrial partners. This opens the door for an industrial realization of the pipetting system independent of the various patents existing in the field of liquid handling.
Stochastic Systems Seminar
(Joint work with Tom Kurtz)
Brown Applied Mathematics Pattern Theory and Vision Seminar
Abstract: Belief propagation (BP) was only supposed to work for tree-like networks but works suprisingly well in many applications involving networks with loops, including turbo codes. However, there has been little understanding of the algorithm or the nature of the solutions it finds for general graphs.
We have shown that BP can only converge to a stationary point of an approximate free energy, known as the Bethe free energy in statistical physics. This result characterizes BP fixed-points and makes connectons with variational approaches to approximate inference.
More importantly, our analysis lets us build on the progress made in statistical physics since Bethe's approximation was introduced in 1935. Kikuchi and others have shown how to construct more accurate free energy approximations, of which Bethe's approximation is the simplest. Exploiting the insights from our analysis, we derive generalized belief propagation (GBP) versions of these Kikuchi approximations. These new message passing algorithms can be significantly more accurate than ordinary BP, at an adjustable increase in complexity. We illustrate such a new GBP algorithm and show that it can give much more accurate marginal probabilities than those found using ordinary BP.
I'll show applications of GBP to error-correcting codes, and of BP to super-resolution (estimating missing high-resolution details from images).
(Joint work with Jonathan Yedidia (MERL) and Yair Weiss (U.C. Berkeley))
Brown Analysis Seminar
Scientific Computing Seminar
Abstract: In recent years there have been a number of substantial advances in numerical techniques for solving the wave equation. These include:
(i) Efficient methods for truncating unbounded domains which can provide arbitrary accuracy;
(ii) Fast methods for evaluating the standard integral solution formulas, akin to the fast multipole method in the frequency domain;
(iii) New integral solution formulas leading to two-step, explicit, unconditionally stable numerical methods with any order of accuracy;
I will review these new methods, and speculate on future developments.
<--- 2000 Index
DAM Home Page