Search

Recent findings suggest that humans deploy cognitive mechanism of physics simulation engines to simulate the physics of objects. We propose a framework for bots to deploy probabilistic programming tools for interacting with intuitive physics environments. The framework employs a physics simulation in a probabilistic way to infer about moves performed by an agent in a setting governed by Newtonian laws of motion. However, methods of probabilistic programs can be slow in such setting due to their need to generate many samples. We complement the model with a model-free approach to aid the sampling procedures in becoming more efficient through learning from experience during game playing. We present an approach where combining model-free approaches (a convolutional neural network in our model) and model-based approaches (probabilistic physics simulation) is able to achieve what neither could alone. This way the model outperforms an all model-free or all model-based approach. We discuss a case study showing empirical results of the performance of the model on the game of Flappy Bird., Comment: Shorter version to appear in AAAI-21

The classification of streets on road networks has been focused on the vehicular transportational features of streets such as arterials, major roads, minor roads and so forth based on their transportational use. City authorities on the other hand have been shifting to more urban inclusive planning of streets, encompassing the side use of a street combined with the transportational features of a street. In such classification schemes, streets are labeled for example as commercial throughway, residential neighborhood, park etc. This modern approach to urban planning has been adopted by major cities such as the city of San Francisco, the states of Florida and Pennsylvania among many others. Currently, the process of labeling streets according to their contexts is manual and hence is tedious and time consuming. In this paper, we propose an approach to collect and label imagery data then deploy advancements in computer vision towards modern urban planning. We collect and label street imagery then train deep convolutional neural networks (CNN) to perform the classification of street context. We show that CNN models can perform well achieving accuracies in the 81% to 87%, we then visualize samples from the embedding space of streets using the t-SNE method and apply class activation mapping methods to interpret the features in street imagery contributing to output classification from a model., Comment: Paper to appear at IEEE Big Data 2019

This paper presents an example of how demographical characteristics of patients influence their susceptibility to certain medical conditions. In this paper, we investigate the association of health conditions to age of patients in a heterogeneous population. We show that besides the symptoms a patients is having, the age has the potential of aiding the diagnostic process in hospitals. Working with Electronic Health Records (EHR), we show that medical conditions group into clusters that share distinctive population age densities. We use Electronic Health Records from Brazil for a period of 15 months from March of 2013 to July of 2014. The number of patients in the data is 1.7 million patients and the number of records is 47 million records. The findings has the potential of helping in a setting where an automated system undergoes the task of predicting the condition of a patient given their symptoms and demographical information., Comment: NIPS 2016 Workshop on Machine Learning for Health

Understanding the dynamics by which urban areas attract visitors is significant for urban development in cities. In addition, identifying services that relate to highly attractive districts is useful to make policies regarding the placement of such places. Thus, we present a framework for classifying districts in cities by their attractiveness to visitors, and relating Points of Interests (POIs) types to districts' attraction patterns. We used Origin-Destination matrices (ODs) mined from cell phone data that capture the flow of trips between each pair of places in Riyadh, Saudi Arabia. We define the attraction profile for a place based on three main statistical features: The amount of visitors a place received, the distribution of distance traveled by visitors on the road network, and the spatial spread of where visitors come from. We use a hierarchical clustering algorithm to classify all places in the city by their features of attraction. We detect three types of Urban Attractors in Riyadh during the morning period: Global which are significant places in the city, Downtown which the central business district and Residential attractors. In addition, we uncover what makes these places different in terms of attraction patterns. We used a statistical significance testing approach to rigorously quantify the relationship between Points of Interests (POIs) types (services) and the 3 patterns of Urban Attractors we detected. The proposed framework can be used for detecting the attraction patterns given by type of services related to each pattern. This is a critical piece of information to inform trip distribution models.

Rapid urbanization places increasing stress on already burdened transportation systems, resulting in delays and poor levels of service. Billions of spatiotemporal call detail records (CDRs) collected from mobile devices create new opportunities to quantify and solve these problems. However, there is a need for tools to map new data onto existing transportation infrastructure. In this work, we propose a system that leverages this data to identify patterns in road usage. First, we develop an algorithm to mine billions of calls and learn location transition probabilities of callers. These transition probabilities are then upscaled with demographic data to estimate origin-destination (OD) flows of residents between any two intersections of a city. Next, we implement a distributed incremental traffic assignment algorithm to route these flows on road networks and estimate congestion and level of service for each roadway. From this assignment, we construct a bipartite usage network by connecting census tracts to the roads used by their inhabitants. Comparing the topologies of the physical road network and bipartite usage network allows us to classify each road's role in a city's transportation network and detect causes of local bottlenecks. Finally, we demonstrate an interactive, web-based visualization platform that allows researchers, policymakers, and drivers to explore road congestion and usage in a new dimension. To demonstrate the flexibility of this system, we perform these analyses in multiple cities across the globe with diverse geographical and sociodemographic qualities. This platform provides a foundation to build congestion mitigation solutions and generate new insights into urban mobility., Comment: submitted to KDD 2014