More Like this

1. Sensitivity Analysis for Dynamic Control of PSTNs with Skewed Distributions

   https://doi.org/10.1609/icaps.v33i1.27183  

   Chen, Rosy; Ma, Yiran; Wu, Siqi; Boerkoel_Jr, James C (July 2023, Proceedings of the International Conference on Automated Planning and Scheduling)

   Koenig, Sven; Stern, Roni; Vallati, Mauro (Ed.)

   Probabilistic Simple Temporal Networks (PSTN) facilitate solving many interesting scheduling problems by characterizing uncertain task durations with unbounded probabilistic distributions. However, most current approaches assess PSTN performance using normal or uniform distributions of temporal uncertainty. This paper explores how well such approaches extend to families of non-symmetric distributions shown to better represent the temporal uncertainty introduced by, e.g., human teammates by building new PSTN benchmarks. We also build probability-aware variations of current approaches that are more reactive to the shape of the underlying distributions. We empirically evaluate the original and modified approaches over well-established PSTN datasets. Our results demonstrate that alignment between the planning model and reality significantly impacts performance. While our ideas for augmenting existing algorithms to better account for human-style uncertainty yield only marginal gains, our results surprisingly demonstrate that existing methods handle positively-skewed temporal uncertainty better. 

   more » « less
2. Improving structured population models with more realistic representations of non‐normal growth

   https://doi.org/10.1111/2041-210X.13240  

   DeMarche, Megan L.; Morris, William; Linares, Cristina; Doak, Daniel; Altwegg, ed., Res (July 2019, Methods in Ecology and Evolution)

   Abstract Structured population models are among the most widely used tools in ecology and evolution. Integral projection models (IPMs) use continuous representations of how survival, reproduction and growth change as functions of state variables such as size, requiring fewer parameters to be estimated than projection matrix models (PPMs). Yet, almost all published IPMs make an important assumption that size‐dependent growth transitions are or can be transformed to be normally distributed. In fact, many organisms exhibit highly skewed size transitions. Small individuals can grow more than they can shrink, and large individuals may often shrink more dramatically than they can grow. Yet, the implications of such skew for inference from IPMs has not been explored, nor have general methods been developed to incorporate skewed size transitions into IPMs, or deal with other aspects of real growth rates, including bounds on possible growth or shrinkage.Here, we develop a flexible approach to modelling skewed growth data using a modified beta regression model. We propose that sizes first be converted to a (0,1) interval by estimating size‐dependent minimum and maximum sizes through quantile regression. Transformed data can then be modelled using beta regression with widely available statistical tools. We demonstrate the utility of this approach using demographic data for a long‐lived plant, gorgonians and an epiphytic lichen. Specifically, we compare inferences of population parameters from discrete PPMs to those from IPMs that either assume normality or incorporate skew using beta regression or, alternatively, a skewed normal model.The beta and skewed normal distributions accurately capture the mean, variance and skew of real growth distributions. Incorporating skewed growth into IPMs decreases population growth and estimated life span relative to IPMs that assume normally distributed growth, and more closely approximate the parameters of PPMs that do not assume a particular growth distribution. A bounded distribution, such as the beta, also avoids the eviction problem caused by predicting some growth outside the modelled size range.Incorporating biologically relevant skew in growth data has important consequences for inference from IPMs. The approaches we outline here are flexible and easy to implement with existing statistical tools. 

   more » « less
3. Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability

   Ghosh, Dibya; Rahme, Jad; Kumar, Aviral; Zhang, Amy; Adams, Ryan P.; Levine, Sergey (January 2021, Advances in neural information processing systems)

   Generalization is a central challenge for the deployment of reinforcement learning (RL) systems in the real world. In this paper, we show that the sequential structure of the RL problem necessitates new approaches to generalization beyond the well-studied techniques used in supervised learning. While supervised learning methods can generalize effectively without explicitly accounting for epistemic uncertainty, we describe why appropriate uncertainty handling can actually be essential in RL. We show that generalization to unseen test conditions from a limited number of training conditions induces a kind of implicit partial observability, effectively turning even fully-observed MDPs into POMDPs. Informed by this observation, we recast the problem of generalization in RL as solving the induced partially observed Markov decision process, which we call the epistemic POMDP. We demonstrate the failure modes of algorithms that do not appropriately handle this partial observability, and suggest a simple ensemble-based technique for approximately solving the partially observed problem. Empirically, we demonstrate that our simple algorithm derived from the epistemic POMDP achieves significant gains in generalization over current methods on the Procgen benchmark suite. 

   more » « less
4. Memory for incidentally learned categories evolves in the post-learning interval

   https://doi.org/10.7554/eLife.81855  

   Gabay, Yafit; Karni, Avi; Holt, Lori L (March 2023, eLife)

   Humans generate categories from complex regularities evolving across even imperfect sensory input. Here, we examined the possibility that incidental experiences can generate lasting category knowledge. Adults practiced a simple visuomotor task not dependent on acoustic input. Novel categories of acoustically complex sounds were not necessary for task success but aligned incidentally with distinct visuomotor responses in the task. Incidental sound category learning emerged robustly when within-category sound exemplar variability was closely yoked to visuomotor task demands and was not apparent in the initial session when this coupling was less robust. Nonetheless, incidentally acquired sound category knowledge was evident in both cases one day later, indicative of offline learning gains and, nine days later, learning in both cases supported explicit category labeling of novel sounds. Thus, a relatively brief incidental experience with multi-dimensional sound patterns aligned with behaviorally relevant actions and events can generate new sound categories, immediately after the learning experience or a day later. These categories undergo consolidation into long-term memory to support robust generalization of learning, rather than simply reflecting recall of specific sound-pattern exemplars previously encountered. Humans thus forage for information to acquire and consolidate new knowledge that may incidentally support behavior, even when learning is not strictly necessary for performance. 

   more » « less
5. Getting Local and Personal: Toward Building a Predictive Model for COVID in Three United States Cities.

   Edwards, April; Metcalf, Leigh; Casey, William A; Chatterjee, Shirshendu; Janwa, Heeralal; Battifarano, Ernest (May 2023, Springer)

   Kacprzyk, Janusz; Pal, Nikhil R; Perez, Rafael B; Corchado, Emilio S; Hagras, Hani; Kóczy, László T; Kreinovich, Vladik; Lin, Chin-Teng; Lu, Jie; Melin, Patricia (Ed.)

   The COVID-19 pandemic was lived in real-time on social media. In the current project, we use machine learning to explore the relationship between COVID-19 cases and social media activity on Twitter. We were particularly interested in determining if Twitter activity can be used to predict COVID-19 surges. We also were interested in exploring features of social media, such as replies, to determine their promise for understanding the views of individual users. With the prevalence of mis/disinformation on social media, it is critical to develop a deeper and richer understanding of the relationship between social media and real-world events in order to detect and prevent future influence operations. In the current work, we explore the relationship between COVID-19 cases and social media activity (on Twitter) in three major United States cities with different geographical and political landscapes. We find that Twitter activity resulted in statistically significant correlations using the Granger causality test, with a lag of one week in all three cities. Similarly, the use of replies, which appear more likely to be generated by individual users, not bots or public relations operations, was also strongly correlated with the number of COVID-19 cases using the Granger causality test. Furthermore, we were able to build promising predictive models for the number of future COVID-19 cases using correlation data to select features for input to our models. In contrast, significant correlations were not identified when comparing the number of COVID-19 cases with mainstream media sources or with a sample of all US COVID-related tweets. We conclude that, even for an international event such as COVID-19, social media tracks closely with local conditions. We also suggest that replies can be a valuable feature within a machine learning task that is attempting to gauge the reactions of individual users. 

   more » « less