Search for: All records

ABSTRACT The compaction measurements of Quaternary and Tertiary Gulf Coast aquifer system sediments in the Houston-Galveston region (TX) show spatially variable compression of 0.08 to 8.49 mm/yr because of geohistorical overburden pressure when groundwater levels in the aquifer system were stable after about the year 2000. An aquifer-system creep equation is developed for evaluating this variable compression, with a thickness-weighted average creep coefficient based on Taylor's (1942) secondary consolidation theory. The temporal variation of aquifer system creep can be neglected in a short-term observation period (such as a decade) after a long-term creep period (such as over 1,000 years) in geohistory. The creep coefficient of the Gulf Coast aquifer system is found to be in a range of 8.74 × 10−5 to 3.94 × 10−3 (dimensionless), with an average of 1.38 × 10−3. Moreover, for silty clay or clay-dominant aquitards in the Gulf Coast aquifer system the creep coefficient value varies in the range of 2.21 × 10−4 to 3.94 × 10−3, which is consistent with values found by Mesri (1973) for most soils, which vary in the range of creep coefficient, 1 × 10−4 to 5 × 10−3. Land subsidence due to secondary consolidation of the Gulf Coast aquifer system is estimated to be 0.04 to 4.33 m in the 20th century and is projected to be 0.01 to 0.64 m in the 21st century at the 13 borehole extensometer locations in the Houston-Galveston region. The significant creep should be considered in the relative sea level rise, in addition to tectonic subsidence and primary consolidation. 

Existing adversarial algorithms for Deep Reinforcement Learning (DRL) have largely focused on identifying an optimal time to attack a DRL agent. However, little work has been explored in injecting efficient adversarial perturbations in DRL environments. We propose a suite of novel DRL adversarial attacks, called ACADIA, representing AttaCks Against Deep reInforcement leArning. ACADIA provides a set of efficient and robust perturbation-based adversarial attacks to disturb the DRL agent's decision-making based on novel combinations of techniques utilizing momentum, ADAM optimizer (i.e., Root Mean Square Propagation, or RMSProp), and initial randomization. These kinds of DRL attacks with novel integration of such techniques have not been studied in the existing Deep Neural Networks (DNNs) and DRL research. We consider two well-known DRL algorithms, Deep-Q Learning Network (DQN) and Proximal Policy Optimization (PPO), under Atari games and MuJoCo where both targeted and non-targeted attacks are considered with or without the state-of-the-art defenses in DRL (i.e., RADIAL and ATLA). Our results demonstrate that the proposed ACADIA outperforms existing gradient-based counterparts under a wide range of experimental settings. ACADIA is nine times faster than the state-of-the-art Carlini & Wagner (CW) method with better performance under defenses of DRL. 

Community–academic partnerships (CAPs) are being increasingly used to study and address health disparity issues. CAPs help to create new bodies of knowledge and innovative solutions to community problems, which benefits the community and academia. Supported by a grant, a partnership was formed between an academic research team and a community health organization to analyze and interpret data collected from the caregivers of asthmatic African American children living in urban low-income households. Using a case study approach, we discuss how we built a healthy CAP and the lessons learned from the process. Our analysis was guided by the six main factors that facilitate success in developing collaborative relationships, including (1) environment; (2) membership; (3) process and structure; (4) communication; (5) purpose; and (6) resources. Based on these six factors, we describe our collaboration process, challenges, and areas for improvement. We aimed to provide a “points-to-consider” roadmap for academic and community partners to establish and maintain a mutually beneficial and satisfactory relationship. Collaborating with community members and organizations provides unique opportunities for researchers and students to apply their skills and knowledge from textbooks and the classroom, engage with community members, and improve real-life community needs. Building a constructive CAP involves efforts, energy, and resources from both parties. The six major themes derived from our project offer suggestions for building a healthy, collaborative, and productive relationship that best serves communities in the future. 

We report a new neural backdoor attack, named Hibernated Backdoor, which is stealthy, aggressive and devastating. The backdoor is planted in a hibernated mode to avoid being detected. Once deployed and fine-tuned on end-devices, the hibernated backdoor turns into the active state that can be exploited by the attacker. To the best of our knowledge, this is the first hibernated neural backdoor attack. It is achieved by maximizing the mutual information (MI) between the gradients of regular and malicious data on the model. We introduce a practical algorithm to achieve MI maximization to effectively plant the hibernated backdoor. To evade adaptive defenses, we further develop a targeted hibernated backdoor, which can only be activated by specific data samples and thus achieves a higher degree of stealthiness. We show the hibernated backdoor is robust and cannot be removed by existing backdoor removal schemes. It has been fully tested on four datasets with two neural network architectures, compared to five existing backdoor attacks, and evaluated using seven backdoor detection schemes. The experiments demonstrate the effectiveness of the hibernated backdoor attack under various settings.