Computing in High-Energy AstroParticle Research (CHEAPR2016)
Scientific Program

 

The workshop will take place in the Physics Research Building.

Coffee breaks and snacks will be provided.

Wednesday August 24 - Day 1

First Activity:
The first activity will be a presentation on the brief history of Machine Learning, from the 1960s to the modern day. We will discuss supervised training with an emphasis on regression and classification. Bayesian analysis, Random Forests, and Monte Carlo simulations from the variety of machine learning algorithms will be reviewed. The diverse package of algorithms from SciKit Learn (scikit-learn.org) will be presented, leading into their applications. We will conclude with neural networks, convolutional neural networks, and deep learning versus genetic algorithms and genetic programming.


Coffee Break


Second Activity:
The second activity will be a hands-on, step-by-step walk through a single, real world problem from SciKit Learn (scikit-learn.org). Guests will work in small groups. The QueenBee laptops provided by the Connolly group here will have the necessary Python tools pre-installed, and no setup will be required. As a warm-up, we will work through some simple regression and/or classification examples, since this is a problem with which most scientists and engineers are familiar. We will then move to more complex examples within the Scikit Learn framework, and then break for lunch. The takeaway from this activity is that it is relatively easy to begin using evolutionary algorithms with open-source Python tools.


Break for Lunch


Third Activity:
We will spend the afternoon listening to lectures from workshop guests, updating us on various projects in which they are involved. The emphasis will be on guests not necessarily involved in large physics collaborations, as those presentations will take place on the second and third day.


Break for Dinner


Study Hall:
We would like to invite workshop guests to work together in small groups on problems and data sets they bring to the workshop. Examples might include a particular piece of your design process or analysis which could be robustly improved with evolutionary algorithms. This includes event geometric reconstruction, and event energy estimation. Conference rooms will be kept open in the physics research building to allow guest to sit down and solve problems. We will provide coffee and food throughout all of the study halls.

Thursday August 25 - Day 2

First Activity:
We will have a presentation from the Ohio Supercomputing Center (OSC), in which we are introduced to OSC resources and functionality, with a brief tutorial on high performance computing (HPC) in a multi-core, multi-node environment. The OSU physics department has purchased a "condo" on the Ruby cluster, meaning we have access to dedicated CPU-hours on that cluster free of charge. This allows testing of code and data sets in a HPC environment, with guest accounts provided by OSC. The takeaway from this activity is to experience the power of OSC HPC resources, and to learn how to collaborate with us at CCAPP via the Ruby cluster.


Coffee Break


Second Activity:
We will practice using genetic programming classification using ARA data. We will begin with a 30 minute tutorial on the software, Karoo GP, and then dive into the data. The ARA data set will be large, involving various sources of unphysical thermal noise, continuous-wave (CW) noise, and solar-flare events. The challenge will be to classify the data into these sets by training an evolutionary algorithm using Karoo GP. Karoo GP software has been deployed already in Square Kilometer Array analysis. The takeaway from this activity is that real high-energy astro-particle physics analysis can be performed efficiently with evolutionary algorithms, often using combinations of features from the data we do not anticipate.


Third Activity:
We will use the remainder of the morning to hear presentations from guests on their recent analysis progress, with an emphasis on the physics collaborations.


Break for Dinner


Study Hall:
(see above)

Friday August 26 - Day 3

First Activity:
The first activity for the day will center around practicing using evolutionary algorithms to classify ANITA data. ANITA data is unique, in that we know it has at least one high-energy astrophysical signal in it: the geomagnetically-induced cosmic ray events. The challenge will be to efficiently locate these events and tag them with some efficiency. The takeaway from this activity is simply to practice using a feature set to train different evolutionary algorithms to locate a small piece of interesting data in a large set filled with noise.


Coffee Break


Second Activity:
We will use the remainder of the morning to hear presentations from guests on their recent analysis progress, with an emphasis on the physics collaborations.


Break for Lunch


Third Activity:
We will spend the afternoon listening to any remaining science presentations, and conclude with study hall. We highly encourage guests to use this final study hall to engage in networking, and to sketch future analysis plans, in light of the knowledge and tools learned during the first two days. We will provide coffee and food throughout all of the study halls.

Department of Physics | Department of Astronomy | Postdoctoral Fellowships | Contact Us
191 West Woodruff Avenue, Columbus, OH 43210 Fax: (614) 292-7741