Search for: All records

Systems consisting of interacting agents are prevalent in the world, ranging from dynamical systems in physics to complex biological networks. To build systems which can interact robustly in the real world, it is thus important to be able to infer the precise interactions governing such systems. Existing approaches typically dis- cover such interactions by explicitly modeling the feed-forward dynamics of the trajectories. In this work, we propose Neural Interaction Inference with Potentials (NIIP) as an alternative approach to discover such interactions that enables greater flexibility in trajectory modeling: it discovers a set of relational potentials, represented as energy functions, which when minimized reconstruct the original trajectory. NIIP assigns low energy to the subset of trajectories which respect the relational constraints observed. We illustrate that with these representations NIIP displays unique capabilities in test-time. First, it allows trajectory manipulation, such as interchanging interaction types across separately trained models, as well as trajectory forecasting. Additionally, it allows adding external hand-crafted potentials at test-time. Finally, NIIP enables the detection of out-of-distribution samples and anomalies without explicit training. 

This paper discusses the implementation of an introductory course to engineering established to provide students with knowledge about the roles of engineers, the engineering method, ethics, teamwork, and detailed information about each of the engineering majors offered in the College of Engineering (CoE) of the host institution. The course is offered as part of a larger initiative seeking to improve success indicators among low-income students. This paper provides details about the course structure, implementation context, metrics, and results measured via descriptive statistics among participant students. The results of a longitudinal implementation, suggest that early provision of career information and awareness can impact the engineering retention and persistence of students and their interest in their chosen majors, particularly in educational settings where students declare their major on the entrance to their first year. 

Improving the level of success of students from low socioeconomic backgrounds in science, technology, engineering, and mathematics (STEM) disciplines has been a prevailing concern for higher education institutions for many years. To address this challenge, a pilot initiative has been implemented with engineering students at the University of Puerto Rico Mayaguez, a recognized Hispanic-serving institution. Over the past four years, the Program for Engineering Access, Retention, and LIATS Success (PEARLS) has brought in an innovative intervention model that combines elements from socio-cognitive career theories and departure studies to impact students' success. PEARLS has established a comprehensive range of tools and services, including mentorship, professional readiness training, research opportunities, scholarships, and peer mentor activities. These efforts have led to impressive outcomes, including a significant increase in retention and persistence rates, increased graduation rates having quad-fold those observed in the general student population, and an impressive record of engagements in industry, research, and leadership experiences. This paper discusses the program structure and outcomes from five perspectives that include background experiences, the structure of provided services, the results of their execution, the elements of knowledge derived from its application, and the challenges experienced throughout its implementation. 

One of the long-term objectives of Machine Learning is to endow machines with the capacity of structuring and interpreting the world as we do. This is particularly challenging in scenes involving time series, such as video sequences, since seemingly different data can correspond to the same underlying dynamics. Recent approaches seek to decompose video sequences into their composing objects, attributes and dynamics in a self-supervised fashion, thus simplifying the task of learning suitable features that can be used to analyze each component. While existing methods can successfully disentangle dynamics from other components, there have been relatively few efforts in learning parsimonious representations of these underlying dynamics. In this paper, motivated by recent advances in non-linear identification, we propose a method to decompose a video into moving objects, their attributes and the dynamic modes of their trajectories. We model video dynamics as the output of a Koopman operator to be learned from the available data. In this context, the dynamic information contained in the scene is encapsulated in the eigenvalues and eigenvectors of the Koopman operator, providing an interpretable and parsimonious representation. We show that such decomposition can be used for instance to perform video analytics, predict future frames or generate synthetic video. We test our framework in a variety of datasets that encompass different dynamic scenarios, while illustrating the novel features that emerge from our dynamic modes decomposition: Video dynamics interpretation and user manipulation at test-time. We successfully forecast challenging object trajectories from pixels, achieving competitive performance while drawing useful insights. 

Systems consisting of interacting agents are prevalent in the world, ranging from dynamical systems in physics to complex biological networks. To build systems which can interact robustly in the real world, it is thus important to be able to infer the precise interactions governing such systems. Existing approaches typically discover such interactions by explicitly modeling the feed-forward dynamics of the trajectories. In this work, we propose Neural Interaction Inference with Potentials (NIIP) as an alternative approach to discover such interactions that enables greater flexibility in trajectory modeling: it discovers a set of relational potentials, represented as energy functions, which when minimized reconstruct the original trajectory. NIIP assigns low energy to the subset of trajectories which respect the relational constraints observed. We illustrate that with these representations NIIP displays unique capabilities in test-time. First, it allows trajectory manipulation, such as interchanging interaction types across separately trained models, as well as trajectory forecasting. Additionally, it allows adding external hand-crafted potentials at test-time. Finally, NIIP enables the detection of out-of-distribution samples and anomalies without explicit training. 

Industrial control systems (ICS) include systems that control industrial processes in critical infrastructure such as electric grids, nuclear power plants, manufacturing plans, water treatment systems, pharmaceutical plants, and building automation systems. ICS represent complex systems that contain an abundance of unique devices all of which may hold different types of software, including applications, firmware and operating systems. Due to their ability to control physical infrastructure, ICS have more and more become targets of cyber-attacks, increasing the risk of serious damage, negative financial impact, disruption to business operations, disruption to communities, and even the loss of life. Ethical hacking represents one way to test the security of ICS. Ethical hacking consists of using a cyber-attacker's perspective and a variety of cybersecurity tools to actively discover vulnerabilities and entry points for potential cyber-attacks. However, ICS ethical hacking represents a difficult task due to the wide variety of devices found on ICS networks. Most ethical hackers do not hold expertise or knowledge about ICS hardware, device computing elements, protocols, vulnerabilities found on these elements, and exploits used to exploit these vulnerabilities. Effective approaches are needed to reduce the complexity of ICS ethical hacking tasks. In this study, we use ontology modeling, a knowledge representation approach in artificial intelligence (AI), to model data that represent ethical hacking tasks of building automation systems. With ontology modeling, information is stored and represented in the form of semantic graphs that express individuals, their properties, and the relations between multiple individuals. Data are drawn from sources such as the National Vulnerability Database, ExploitDB, Common Weakness Enumeration (CWE), the Common Attack Pattern and Enumeration Classification (CAPEC), and others. We show, through semantic queries, how the ontology model can automatically link together entities such as software names and versions of ICS software, vulnerabilities found on those software instances, vulnerabilities found on the protocols used by the software, exploits found on those vulnerabilities, weaknesses that represent those vulnerabilities, and attacks that can exploit those weaknesses. The ontology modeling of ICS ethical hacking and the semantic queries run over the model can reduce the complexity of ICS hacking tasks. 

We performed a comprehensive demographic study of the CO extent relative to dust of the disk population in the Lupus clouds in order to find indications of dust evolution and possible correlations with other disk properties. We increased the number of disks of the region with measured R CO and R dust from observations with the Atacama Large Millimeter/submillimeter Array to 42, based on the gas emission in the 12 CO J = 2−1 rotational transition and large dust grains emission at ~0.89 mm. The CO integrated emission map is modeled with an elliptical Gaussian or Nuker function, depending on the quantified residuals; the continuum is fit to a Nuker profile from interferometric modeling. The CO and dust sizes, namely the radii enclosing a certain fraction of the respective total flux (e.g., R 68% ), are inferred from the modeling. The CO emission is more extended than the dust continuum, with a R 68% CO / R 68% dust median value of 2.5, for the entire population and for a subsample with high completeness. Six disks, around 15% of the Lupus disk population, have a size ratio above 4. Based on thermo-chemical modeling, this value can only be explained if the disk has undergone grain growth and radial drift. These disks do not have unusual properties, and their properties spread across the population’s ranges of stellar mass ( M ⋆ ), disk mass ( M disk ), CO and dust sizes ( R CO , R dust ), and mass accretion of the entire population. We searched for correlations between the size ratio and M ⋆ , M disk , R CO , and R dust : only a weak monotonic anticorrelation with the R dust is found, which would imply that dust evolution is more prominent in more compact dusty disks. The lack of strong correlations is remarkable: the sample covers a wide range of stellar and disk properties, and the majority of the disks have very similar size ratios. This result suggests that the bulk of the disk population may behave alike and be in a similar evolutionary stage, independent of the stellar and disk properties. These results should be further investigated, since the optical depth difference between CO and dust continuum might play a major role in the observed size ratios of the population. Lastly, we find a monotonic correlation between the CO flux and the CO size. The results for the majority of the disks are consistent with optically thick emission and an average CO temperature of around 30 K; however, the exact value of the temperature is difficult to constrain. 

We present new 890 μ m continuum ALMA observations of five brown dwarfs (BDs) with infrared excess in Lupus I and III, which in combination with four previously observed BDs allowed us to study the millimeter properties of the full known BD disk population of one star-forming region. Emission is detected in five out of the nine BD disks. Dust disk mass, brightness profiles, and characteristic sizes of the BD population are inferred from continuum flux and modeling of the observations. Only one source is marginally resolved, allowing for the determination of its disk characteristic size. We conduct a demographic comparison between the properties of disks around BDs and stars in Lupus. Due to the small sample size, we cannot confirm or disprove a drop in the disk mass over stellar mass ratio for BDs, as suggested for Ophiuchus. Nevertheless, we find that all detected BD disks have an estimated dust mass between 0.2 and 3.2 M ⊙ ; these results suggest that the measured solid masses in BD disks cannot explain the observed exoplanet population, analogous to earlier findings on disks around more massive stars. Combined with the low estimated accretion rates, and assuming that the mm-continuum emission is a reliable proxy for the total disk mass, we derive ratios of Ṁ acc ∕ M disk that are significantly lower than in disks around more massive stars. If confirmed with more accurate measurements of disk gas masses, this result could imply a qualitatively different relationship between disk masses and inward gas transport in BD disks.