Research on: (i) COSMOS cloud connected vehicles, (ii) Monitoring of traffic intersections, using bird’s eye cameras, supported by ultra-low latency computational/communications hubs; (iii) Simultaneous video-based tracking of cars and pedestrians, and prediction of movement based on long-term observations of the intersection; (iv) Real-time computational processing, using deep learning, utilizing GPUs, in support of COSMOS applications; (v) Sub-10ms latency communication between all vehicles and the edge cloud computational/communication hub, to be used in support of autonomous vehicle navigation. The research is performed using the pilot node of project COSMOS infrastructure.

Continue reading

The objective of this project is to construct linkages across disparate public health data systems using machine learning tools and assess them for bias and equitable representation of subpopulations defined by demographic and socioeconomic factors.

Continue reading

42% of New York City greenhouse gas emissions result from on-site fossil fuel combustion in residential and commercial buildings; space heating is, by far, the majority contributor. Both New York State and NYC have policies to dramatically reduce emissions that will require a transformation in the way buildings are heated, including major efforts in existing buildings. This transition is inextricably linked to existing energy equity issues that we believe significantly overlap across NYC (and elsewhere). These include unreliable heating in the winter, susceptibility to extreme heat (an increasing occurrence with climate change) and struggles to afford energy needs. Various known data sources for NYC are available, though they are disparate and have not been analyzed holistically. Further, we believe there are potential engineering and policy solutions to these challenges. In this project, the DSI scholar will access (and search for where not yet known to qSEL researchers) relevant data sets, analyze those data sets to identify communities exposed to all or a subset of these issues, and assist qSEL researchers in developing models to evaluate possible solutions. The project has the possibility of extending through Summer 2020, subject to fundraising efforts and the success of the Spring 2020 project.

Continue reading

Contestation over language use is an unavoidable feature of American politics. Yet, despite the rise of language policing on both sides of the aisle, we know surprisingly little about how ordinary citizens respond to norms governing language use from both in-group and out-group members. Following Munger (2017), I would like to leverage social media platforms such as Reddit and Twitter to evaluate whether injunctions to use particular words (e.g., undocumented immigrant, Latinx) are effective. I plan to use an experimental approach, where conditional on mentions of “illegal alien” or “Hispanic/Latino,” users are randomly assigned to receive a “language correction.” Outcome measures would include subsequent use of corrected terms, valence of user responses, and upvoting/liking/RTing behavior.

Continue reading

Decoding behavioral signifiers for choice and memory can have far reaching implications for understanding actions and identifying disease. We use a four arm maze where we are able to observe choices and infer memory in mice, but have access to very few pre-determined behavioral signifiers. Several recent publications implemented computer vision to extract a variety of previously unreachable aspects of behavioral analysis, including animal pose estimation (Mathis et al., 2018) and distinguishable internal states (Calhoun et al., 2019). These descriptions allowed for the identification and characterization of dynamics, which then revealed an unprecedented richness to the behaviors that determine decision making. Applying such computational approaches to examine behavior in our maze in the context of behaviors that have been validated to measure choice and memory can reveal dimensions of behavior that predict or even determine these psychological constructs. DSI scholars would use pose estimation analysis to evaluate behavioral signifiers for choice and memory and relate it to our real time concurrent measures of neural activity and transmitter release. The students would also have opportunity to examine the effect of disease models known to impair performance on our maze task on any identified signifier.

Continue reading

The function for much of the 3 billion letters in the human genome remain to be understood. Advances in DNA sequencing technology have generated enormous amount of data, yet we don’t have the tool to extract rules of how the genome works. Deep learning holds great potential in decoding the genome, in particular due to the digital nature of DNA sequences and the ability to handle large data sets. However, like many other applications, the interpretability of deep learning models hampers its ability to help understand the genome. We are developing deep learning architectures embedded with the principles of gene regulation and we will be leveraging billions of existing measurements of gene activity to learn a mechanistic model of gene regulation in human cells.

Continue reading

This project will be focused on creating a deep learning framework for tracking individual molecules and proteins as they move within a cell under various conditions. Using total internal reflection (TIRF) microscopy, we have accumulated more than 10 million trajectories over dozens of experimental preparations with differences in both the imaging approaches as well as the biological context. In our experiments we have captured particles under a wide variety of conditions including increased protein expression level, and a range of drug concentrations. Our biggest challenge is being able to stably track the movement of a particle as it passes by other particles or groups of particles, and to do this in a way that generalizes over novel conditions. The Data Science Institute Scholar chosen for this project would work with scientists in the Javitch laboratory and others across the Columbia campus to conceive of an approach for efficiently and effectively tracking particles. The resulting work would be of great interest to an increasing number of scientists working in this field who currently rely on methods based on feature engineering that are often inaccurate or inflexible compared to modern deep learning methods.

Continue reading