Benjamin Bettisworth is worrking on ESR Project 9: Auto-tuned & flexible phylogenetic likelihood calculations under the supervision of Alexandros Stamatakis at the Heidelberg Institute for Theoretical Studies. In his project Ben will be looking into optimizing the internals of the most recent version of RAxML, called RAxML-NG, to perform run time optimizations based on a heuristic approach. An example of such an optimization is to dynamically select the most efficient computational kernel for specific phases of the program (e.g., tree search versus model parameter optimization phase). This will involve runtime analysis of the data being processed in order to select the most efficient likelihood kernel based upon factors such as (i) the characteristics of the specific input dataset being analyzed, (ii) the computer architecture on which RAxML-NG is being executed, and (iii) the current program phase.
Therefore, Benjamin's project is split into two major parts.
1) The first is a working implementation of this dynamic likelihood kernel selection strategy in RAxML-NG.
2) The second part of the project is then to incorporate complex and more parameter-rich models (e.g., heterotachous models) into RAxML-NG and specifically investigate how difficult it will be to obtain reliable parameter estimates for these models. In general, direct numerical optimization of these parameters tends to be difficult and exhibits several local optima on a single fixed tree topology. In other words, it poses a numerical optimization challenge.
Before IGNITE, Benjamin obtained a degree from the University of Alaska Fairbanks (UAF). During his undergraduate studies, his main interest was in security, as there was a strong security program at the department. However, as Ben progressed in his studies, he became more and more interested in scientific computing, particularly phylogenetics. In his own words: "I was lucky because UAF also had a large biology program, as well as a super computing center on campus. Many graduate students in biology had phylogenetic analysis as part of their thesis, and I was fortunate enough to be able to help, in order to get hands on experience."
Immediately after graduating from the masters program, Ben joined a local computer vision startup, which focused on crowd metrics. There, he became the lead backend engineer and created a few models to track crowd engagement, attention, and dominant emotion. "From my previous job, I am still interested in computer vision, and the topics surrounding facial analysis. Other related topics include pose estimation, and human computer interaction."
Among ben's interest are also robotics and embedded devices, but these are more of a hobby. A personal goal of his is to make his house completely “smart” using electronics and software that he made himself!
Other personal hobbies include hiking, biking, and travel in general. "I am truly excited to travel around Europe as a part of the IGNITE project."
Therefore, Benjamin's project is split into two major parts.
1) The first is a working implementation of this dynamic likelihood kernel selection strategy in RAxML-NG.
2) The second part of the project is then to incorporate complex and more parameter-rich models (e.g., heterotachous models) into RAxML-NG and specifically investigate how difficult it will be to obtain reliable parameter estimates for these models. In general, direct numerical optimization of these parameters tends to be difficult and exhibits several local optima on a single fixed tree topology. In other words, it poses a numerical optimization challenge.
Before IGNITE, Benjamin obtained a degree from the University of Alaska Fairbanks (UAF). During his undergraduate studies, his main interest was in security, as there was a strong security program at the department. However, as Ben progressed in his studies, he became more and more interested in scientific computing, particularly phylogenetics. In his own words: "I was lucky because UAF also had a large biology program, as well as a super computing center on campus. Many graduate students in biology had phylogenetic analysis as part of their thesis, and I was fortunate enough to be able to help, in order to get hands on experience."
Immediately after graduating from the masters program, Ben joined a local computer vision startup, which focused on crowd metrics. There, he became the lead backend engineer and created a few models to track crowd engagement, attention, and dominant emotion. "From my previous job, I am still interested in computer vision, and the topics surrounding facial analysis. Other related topics include pose estimation, and human computer interaction."
Among ben's interest are also robotics and embedded devices, but these are more of a hobby. A personal goal of his is to make his house completely “smart” using electronics and software that he made himself!
Other personal hobbies include hiking, biking, and travel in general. "I am truly excited to travel around Europe as a part of the IGNITE project."
Insitution: Heidelberg Institute for Theoretical Studies
ESR Project 9: Auto-tuned & flexible phylogenetic likelihood calculations