The DNA-binding protein from starved cells (Dps) plays a crucial role in maintaining bacterial cell viability during periods of stress. Dps is a nucleoid-associated protein that interacts with DNA to create biomolecular condensates in live bacteria. Purified Dps protein can also rapidly form large complexes when combined with DNA in vitro. However, the mechanism that allows these complexes to nucleate on DNA remains unclear. Here, we examine how DNA topology influences the formation of Dps–DNA complexes. We find that DNA supercoils offer the most preferred template for the nucleation of condensed Dps structures. More generally, bridging contacts between different regions of DNA can facilitate the nucleation of condensed Dps structures. In contrast, Dps shows little affinity for stretched linear DNA before it is relaxed. Once DNA is condensed, Dps forms a stable complex that can form inter-strand contacts with nearby DNA, even without free Dps present in solution. Taken together, our results establish the important role played by bridging contacts between DNA strands in nucleating and stabilizing Dps complexes.
We present a general formal model called MODIA that can tackle a central challenge for autonomous vehicles (AVs), namely the ability to interact with an unspecified, large number of world entities. In MODIA, a collection of possible decision-problems (DPs), known a priori, are instantiated online and executed as decision-components (DCs), unknown a priori. To combine their individual action recommendations of the DCs into a single action, we propose the lexicographic executor action function (LEAF) mechanism. We analyze the complexity of MODIA and establish LEAF’s relation to regret minimization. Finally, we implement MODIA and LEAF using collections of partially observable Markov decision process (POMDP) DPs, and use them for complex AV intersection decision-making. We evaluate the approach in six scenarios within an industry-standard vehicle simulator, and present its use on an AV prototype.
more » « less- Award ID(s):
- 1724101
- PAR ID:
- 10063272
- Date Published:
- Journal Name:
- International Joint Conference on Artificial Intelligence
- Page Range / eLocation ID:
- 4768 to 4774
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
more » « less
A bottleneck in high-throughput nanomaterials discovery is the pace at which new materials can be structurally characterized. Although current machine learning (ML) methods show promise for the automated processing of electron diffraction patterns (DPs), they fail in high-throughput experiments where DPs are collected from crystals with random orientations. Inspired by the human decision-making process, a framework for automated crystal system classification from DPs with arbitrary orientations was developed. A convolutional neural network was trained using evidential deep learning, and the predictive uncertainties were quantified and leveraged to fuse multiview predictions. Using vector map representations of DPs, the framework achieves a testing accuracy of 0.94 in the examples considered, is robust to noise, and retains remarkable accuracy using experimental data. This work highlights the ability of ML to be used to accelerate experimental high-throughput materials data analytics.
-
null (Ed.)Internet of Things (IoT) sensors often operate in unknown dynamic environments comprising latency-sensitive data sources, dynamic processing loads, and communication channels of unknown statistics. Such settings represent a natural application domain of reinforcement learning (RL), which enables computing and learning decision policies online, with no a priori knowledge. In our previous work, we introduced a post-decision state (PDS) based RL framework, which considerably accelerates the rate of learning an optimal decision policy. The present paper formulates an efficient hardware architecture for the action evaluation step, which is the most computationally-intensive step in the PDS based learning framework. By leveraging the unique characteristics of PDS learning, we optimize its state value expectation and known cost computational blocks, to speed-up the overall computation. Our experiments show that the optimized circuit is 49 times faster than its software implementation counterpart, and six times faster than a Q-learning hardware accelerator.more » « less
-
more » « less
Abstract Background Electrocorticography (ECoG) language mapping is often performed extraoperatively, frequently involves offline processing, and relationships with direct cortical stimulation (DCS) remain variable. We sought to determine the feasibility and preliminary utility of an intraoperative language mapping approach guided by real-time visualization of electrocorticograms.
Methods A patient with astrocytoma underwent awake craniotomy with intraoperative language mapping, utilizing a dual iPad stimulus presentation system coupled to a real-time neural signal processing platform capable of both ECoG recording and delivery of DCS. Gamma band modulations in response to 4 language tasks at each electrode were visualized in real-time. Next, DCS was conducted for each neighboring electrode pair during language tasks.
Results All language tasks resulted in strongest heat map activation at an electrode pair in the anterior to mid superior temporal gyrus. Consistent speech arrest during DCS was observed for Object and Action naming tasks at these same electrodes, indicating good correspondence with ECoG heat map recordings. This region corresponded well with posterior language representation via preoperative functional MRI.
Conclusions Intraoperative real-time visualization of language task-based ECoG gamma band modulation is feasible and may help identify targets for DCS. If validated, this may improve the efficiency and accuracy of intraoperative language mapping.
-
null (Ed.)This work presents a novel prototype autonomous vehicle (AV) human-machine interface (HMI) in virtual reality (VR) that utilizes a human-like visual embodiment in the driver’s seat of an AV to communicate AV intent to pedestrians in a crosswalk scenario. There is currently a gap in understanding the use of virtual humans in AV HMIs for pedestrian crossing despite the demonstrated efcacy of human-like interfaces in improving human-machine relationships. We conduct a 3x2 within-subjects experiment in VR using our prototype to assess the efects of a virtual human visual embodiment AV HMI on pedestrian crossing behavior and experience. In the experiment participants walk across a virtual crosswalk in front of an AV. How long they took to decide to cross and how long it took for them to reach the other side were collected, in addition to their subjective preferences and feelings of safety. Of 26 participants, 25 preferred the condition with the most anthropomorphic features. An intermediate condition where a human-like virtual driver was present but did not exhibit any behaviors was least preferred and also had a signifcant efect on time to decide. This work contributes the frst empirical work on using human-like visual embodiments for AV HMIs.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.24963/ijcai.2017/664
