SPEAKERS

Learning Representations of Human Behavior

Advances in computational behavior analyses and machine learning provide an opportunity to profoundly impact a wide range of professions that heavily rely on the interpretation of human behavior (e.g., customer profiling, targeted advertising, interpersonal skill training, and mental health screening). For example, automatic algorithms to detect impairments in social-emotional functioning from individuals’ behavior have the potential to increase the objectivity, accessibility, and efficiency of mental health care. In other words, automatic techniques can provide professionals with a different set of eyes and ears that produce quantified and objective assessments where otherwise only subjective information would be available.

Within this presentation I discuss how we can automatically learn meaningful and discriminatory representations of human behavior while leveraging both (1) explicit descriptors of human behavior motivated by top-down knowledge about human nature as well as (2) advanced machine learning techniques that derive representations automatically and directly from the data; combining the best of both worlds (1) better interpretability of human behavior and (2) exceptional performance in tasks with large quantities of data.

Quantifying the Acetone-Butanol-Ethanol Fermentation Pathway Using NMR Spectroscopy

Biofuels have long been considered a clean alternative to fossil fuels with ethanol a key focus. However, ethanol has several disadvantages as a fossil fuel replacement including the repurposing arable land for corn production specifically to convert to ethanol (limiting corn available for food) and the need to modify gasoline engines to cope with high amounts of ethanol to prevent engine damage. Butanol, another biofuel, is a promising alternative to ethanol as it less hygroscopic than ethanol and can directly be used as a gasoline replacement without engine modification. Additionally, butanol can be produced through acetone-butanol-ethanol (ABE) pathways in some anaerobic bacteria, such as Clostridium beijerinkii, which does not require the reappropriation of arable land for fuel production. Understanding how these pathways interact to produce butanol are important for optimizing the production of the biofuel. Herein, we propose a novel approach using one-dimensional and two-dimensional NMR to quickly quantify the major components of ABE fermentation, including acetone, butanol, ethanol, isopropanol, and acetic acid. 1D NMR can quickly quantify most major components in less than 8 min using standard quantitative techniques, however 2D NMR is required to quantify trace amounts of butyric acid due to spectral convolution. We explored TOCSY and 2D-JRES for suitability when simultaneously quantifying the major products (butanol) while also quantifying the minor products (butyric acid). We developed a simple calibration curve to explore and found that the linearity of both 2D methods was at least than 0.998 and RSD of both 2D methods was less than 8%. Quantification using TOCSY was found to be best, with a RSD less than 5%, however the 2D-JRES method was significantly faster (34 min vs. 161 min). Currently, we are expanding this method to include the other major components and are applying it directly to the effluent produced by Clostridium beijerinkii strains.

Aerial Co-robots of the Future: Safety, Intelligence, Certification

This presentation discusses the key challenges of the 21st century and puts forward the right perspective for development of aerial co-robots of the future by emphasizing safety, intelligence and certification. Each of these three pillars hinges on fundamental theoretical developments for support. Challenges with flight control, cyber-resilience, cooperative path planning, intelligent control, and certification are discussed, and fundamental limitations of feedback loops are revisited for development of safe intelligent control. The new metrics for certification, important in the era of the fourth industrial revolution and requiring new paradigms for certification, are presented. Applications in elderly care, scalable e-commerce, and precision agriculture are discussed.

Light, Electrons, Protons: Lessons from Model Systems and Potentials for Photocatalysis

The inspiration for this talk comes from the photoelectrochemical interface, which is a place rich with unknowns and unrealized potentials. Electrons are excited by light either in the electrode or in the adsorbed molecules, charges traverse the electrode-electrolyte interface, protons flow from the bulk to complete redox reactions, and interfacial electric fields develop to balance chemical potential differences between the opposing phases. In this talk, the complex chemistry at the interface will be used as a point of reference to motivate several chemical dynamics studies in small molecules, solids, and interfaces with the goal of generating new directions and ideas for understanding and driving interfacial reactions. New concepts that will be discussed are basicity in the excited state, solvation near an interface, electronic-vibrational dynamics in a solid made of a redox couple, and influencing proton conductivity with light. Several avenues on how to use this knowledge will be proposed.

Lifting the Cosmic Veil: Spitzer Observations from our own Backyard to the Edge of the Universe

Since its launch in 2003, NASA’s Spitzer Space Telescope has used the infrared part of the spectrum to study the Universe near and far. In doing so it has discovered objects as wide raging from a previously unknown ring of Saturn to young galaxies at the edge of the observable Universe. Its greatest legacy though may be the study of the atmospheres of planets around other stars. Come hear about what Spitzer has accomplished and what is in store for its future before the mission is set to come to an end in January of 2020.