scholarly journals InterTwin: Deep Learning Approaches for Computing Measures of Effectiveness for Traffic Intersections

2021 ◽  
Vol 11 (24) ◽  
pp. 11637
Author(s):  
Yashaswi Karnati ◽  
Rahul Sengupta ◽  
Sanjay Ranka
Keyword(s):  
Network Architecture ◽  
Probability Distributions ◽  
Wait Time ◽  
Signal Timing ◽  
Learning Approaches ◽  
Microscopic Simulation ◽  
Spatial Graph ◽  
Traffic Intersections ◽  
Good Signal ◽  
Measures Of Effectiveness

Microscopic simulation-based approaches are extensively used for determining good signal timing plans on traffic intersections. Measures of Effectiveness (MOEs) such as wait time, throughput, fuel consumption, emission, and delays can be derived for variable signal timing parameters, traffic flow patterns, etc. However, these techniques are computationally intensive, especially when the number of signal timing scenarios to be simulated are large. In this paper, we propose InterTwin, a Deep Neural Network architecture based on Spatial Graph Convolution and Encoder-Decoder Recurrent networks that can predict the MOEs efficiently and accurately for a wide variety of signal timing and traffic patterns. Our methods can generate probability distributions of MOEs and are not limited to mean and standard deviation. Additionally, GPU implementations using InterTwin can derive MOEs, at least four to five orders of magnitude faster than microscopic simulations on a conventional 32 core CPU machine.

Trajectory-Based Measurement of Signalized Intersection Capacity

2019 ◽  
Vol 2673 (10) ◽  
pp. 370-380 ◽  
Author(s):  
Walid Fourati ◽  
Bernhard Friedrich
Keyword(s):  
Cycle Time ◽  
Degree Of Saturation ◽  
Limiting Factor ◽  
Signal Timing ◽  
Flow Rates ◽  
Practical Applications ◽  
Road Intersections ◽  
Good Signal ◽  
Saturation Flow ◽  
Over Time

Capacities of road intersections are a limiting factor and crucial for the performance of road networks. Therefore, for purposes of intersection design and of optimal signal timing, numerous methodologies have been proposed to either estimate or directly measure the capacity of single movements at road intersections. However, both model-based estimation and direct measurement suffer from the large effort that is needed to gather the relevant data. Even worse, once the data are collected they only represent a snapshot of the capacity over time. This paper proposes an alternative approach to estimate capacity of signalized road intersections over time using only automatically generated trajectories of probe vehicles. The obtained capacity can be used to evaluate the effective degree of saturation using real demand, or to assess hypothetic different conditions in demand or signaling. The cyclic operation of signalized intersections allows for the accumulation of trajectories, and thus in practical applications for the compensation of potentially low penetration rates. Within a sequential process the intersection’s cycle time and the approach green time and saturation flow rates are determined. The determination of the cycle time and the green times is based on an existing approach. The derivation of the saturation flow rates relies on its direct dependency to the saturation time headway and uses two parameters to be calibrated. Testing with a commercial dataset on an intersection in Munich produced a good signal timing estimation and saturation flow values that are comparable to a calculation based on the German guideline.

Multi-Context Generation in Virtual Reality Environments Using Deep Reinforcement Learning

2020 ◽  
Author(s):  
James Cunningham ◽  
Christian Lopez ◽  
Omar Ashour ◽  
Conrad S. Tucker
Keyword(s):  
Virtual Reality ◽  
Reinforcement Learning ◽  
Virtual Environments ◽  
Probability Distributions ◽  
Automatic Generation ◽  
Grocery Store ◽  
Training Data ◽  
Learning Approaches ◽  
Common Concept

Abstract In this work, a Deep Reinforcement Learning (RL) approach is proposed for Procedural Content Generation (PCG) that seeks to automate the generation of multiple related virtual reality (VR) environments for enhanced personalized learning. This allows for the user to be exposed to multiple virtual scenarios that demonstrate a consistent theme, which is especially valuable in an educational context. RL approaches to PCG offer the advantage of not requiring training data, as opposed to other PCG approaches that employ supervised learning approaches. This work advances the state of the art in RL-based PCG by demonstrating the ability to generate a diversity of contexts in order to teach the same underlying concept. A case study is presented that demonstrates the feasibility of the proposed RL-based PCG method using examples of probability distributions in both manufacturing facility and grocery store virtual environments. The method demonstrated in this paper has the potential to enable the automatic generation of a variety of virtual environments that are connected by a common concept or theme.

Microscopic Simulation and Optimization of Signal Timing based on Multi-Agent: A Case Study of the Intersection in Tianjin

2018 ◽  
Vol 22 (9) ◽  
pp. 3373-3382
Author(s):  
Leiting Sun ◽  
Jianqiang Tao ◽  
Chunfa Li ◽  
Shengkai Wang ◽  
Ziqiang Tong
Keyword(s):  
Signal Timing ◽  
Microscopic Simulation ◽  
Simulation And Optimization ◽  
Multi Agent

scholarly journals PARROT is a flexible recurrent neural network framework for analysis of large protein datasets

eLife ◽  
2021 ◽  
Vol 10 ◽  
Author(s):  
Daniel Griffith ◽  
Alex S Holehouse
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Deep Learning ◽  
High Throughput ◽  
Recurrent Neural Network ◽  
Transcriptional Activation ◽  
Network Architecture ◽  
Learning Approaches ◽  
Large Protein ◽  
Protein Datasets

The rise of high-throughput experiments has transformed how scientists approach biological questions. The ubiquity of large-scale assays that can test thousands of samples in a day has necessitated the development of new computational approaches to interpret this data. Among these tools, machine learning approaches are increasingly being utilized due to their ability to infer complex nonlinear patterns from high-dimensional data. Despite their effectiveness, machine learning (and in particular deep learning) approaches are not always accessible or easy to implement for those with limited computational expertise. Here we present PARROT, a general framework for training and applying deep learning-based predictors on large protein datasets. Using an internal recurrent neural network architecture, PARROT is capable of tackling both classification and regression tasks while only requiring raw protein sequences as input. We showcase the potential uses of PARROT on three diverse machine learning tasks: predicting phosphorylation sites, predicting transcriptional activation function of peptides generated by high-throughput reporter assays, and predicting the fibrillization propensity of amyloid beta with data generated by deep mutational scanning. Through these examples, we demonstrate that PARROT is easy to use, performs comparably to state-of-the-art computational tools, and is applicable for a wide array of biological problems.

scholarly journals Design and implementation of convolutional neural network architectures for low-dose CT image noise reduction

2021 ◽  
Author(s):  
Seyyedomid Badretale
Keyword(s):  
Neural Network ◽  
Image Quality ◽  
Convolutional Neural Network ◽  
Network Architecture ◽  
Low Dose ◽  
Ct Images ◽  
Learning Approaches ◽  
Low Dose Ct ◽  
Image Noise Reduction ◽  
Low Dose Computed Tomography

An essential objective in low-dose Computed Tomography (CT) imaging is how best to preserve the image quality. While the image quality lowers with reducing the X-ray dosage, improving the quality is crucial. Therefore, a novel method to denoise low-dose CT images has been presented in this thesis. Different from the traditional algorithms which utilize similar shared features of CT images in the spatial domain, the deep learning approaches are suggested for low-dose CT denoising. The proposed algorithm learns an end-to-end mapping from the low-dose CT images for denoising the low-dose CT images. The first method is based on a fully convolutional neural network. The second approach is a deep convolutional neural network architecture consisting of five major sections. The results of two frameworks are compared with the state-of-the-art methods. Several metrics for assessing image quality are applied in this thesis in order to highlight the supremacy of the performed method.

scholarly journals Effective Training of Deep Convolutional Neural Networks for Hyperspectral Image Classification through Artificial Labeling

2020 ◽  
Vol 12 (16) ◽  
pp. 2653 ◽  
Author(s):  
Wojciech Masarczyk ◽  
Przemysław Głomb ◽  
Bartosz Grabowski ◽  
Mateusz Ostaszewski
Keyword(s):  
Neural Network ◽  
Transfer Learning ◽  
Network Architecture ◽  
Hyperspectral Image ◽  
Large Data ◽  
Learning Approaches ◽  
Classification Problems ◽  
Deep Convolutional Neural Networks ◽  
Additional Advantage ◽  
Hyperspectral Classification

Hyperspectral imaging is a rich source of data, allowing for a multitude of effective applications. However, such imaging remains challenging because of large data dimension and, typically, a small pool of available training examples. While deep learning approaches have been shown to be successful in providing effective classification solutions, especially for high dimensional problems, unfortunately they work best with a lot of labelled examples available. The transfer learning approach can be used to alleviate the second requirement for a particular dataset: first the network is pre-trained on some dataset with large amount of training labels available, then the actual dataset is used to fine-tune the network. This strategy is not straightforward to apply with hyperspectral images, as it is often the case that only one particular image of some type or characteristic is available. In this paper, we propose and investigate a simple and effective strategy of transfer learning that uses unsupervised pre-training step without label information. This approach can be applied to many of the hyperspectral classification problems. The performed experiments show that it is very effective at improving the classification accuracy without being restricted to a particular image type or neural network architecture. The experiments were carried out on several deep neural network architectures and various sizes of labeled training sets. The greatest improvement in overall accuracy on the Indian Pines and Pavia University datasets is over 21 and 13 percentage points, respectively. An additional advantage of the proposed approach is the unsupervised nature of the pre-training step, which can be done immediately after image acquisition, without the need of the potentially costly expert’s time.

Queue Jump Lane, Transit Signal Priority, and Stop Location Evaluation of Transit Preferential Treatments Using Microsimulation

2015 ◽  
Vol 2533 (1) ◽  
pp. 39-49 ◽  
Author(s):  
Burak Cesme ◽  
Selman Z. Altun ◽  
Barrett Lane
Keyword(s):  
Simulation Models ◽  
High Volume ◽  
Signal Timing ◽  
Transit Signal Priority ◽  
Preferential Treatment ◽  
Test Bed ◽  
Microscopic Simulation ◽  
Delay Reduction ◽  
Treatment Benefits ◽  
Detailed Simulation

Transit preferential treatments offer the potential to improve transit travel time and reliability. However, the benefits of these treatments vary greatly depending on the specific characteristics of the study area, including turning movement and pedestrian volumes, signal timing parameters, and transit stop location. To evaluate the performance of preferential treatments, practitioners typically rely on microscopic simulation models, which require a considerable amount of effort, or a review of previous studies, which may reflect a bias toward the area characteristics. This paper develops a test bed and a planning-level framework to help practitioners determine benefits offered by various preferential treatments without developing a detailed simulation model. To evaluate preferential treatment benefits, the authors performed extensive simulation runs under various scenarios at an isolated intersection with VISSIM. The analyses show that the greatest benefit comes from relocating a nearside stop to a farside stop, in which farside stops can reduce delay up to 30 s per intersection. The highest saving that could be obtained with a queue jump lane is approximately 9 s per intersection. As the number of right turns increases along with the number of conflicting pedestrians, the benefit of a queue jump lane disappears. Transit signal priority with 15 s of green extension and red truncation can offer up to 19 s of reduction in delay; the benefits become more pronounced with a high volume-to-capacity (v/c) ratio. With a low v/c ratio, granting 10 s of green extension without red truncation provides very marginal benefits; only a 2-s delay reduction per intersection is gained.

Use of Microsimulation to Evaluate Signal-Phasing Schemes at Diverging Diamond Interchanges

2017 ◽  
Vol 2620 (1) ◽  
pp. 10-19 ◽  
Author(s):  
Shannon Warchol ◽  
Thomas Chase ◽  
Christopher Cunningham
Keyword(s):  
Assessment Tool ◽  
Signal Timing ◽  
Pairwise Comparisons ◽  
Initial Signal ◽  
The Subject ◽  
Diamond Interchanges ◽  
Measures Of Effectiveness ◽  
Future Work ◽  
Dynamic Bandwidth ◽  
Critical Movement

Even though diverging diamond interchanges (DDIs) have been the subject of research for more than a decade, the effort to standardize interchange signal timing has developed only recently. A three-factor fully crossed experiment was conducted to investigate the influence of crossover spacing and increased volumes on the performance of DDI phasing schemes. PTV Vistro software and the dynamic bandwidth assessment tool were used to optimize the split, cycle length, and offset of each of the 72 treatments. PTV Vissim software was used to collect microsimulation data. Mean interchange delay and mean stops per vehicle were selected as measures of effectiveness. Pairwise comparisons were used to determine whether an existing preferred phasing scheme could minimize delays or stops under three cases: ( a) given spacing and increased volume, ( b) given volume independent of spacing, and ( c) given spacing independent of increased volume. The data revealed that a two- or three-critical-movement phasing scheme usually resulted in the lowest mean interchange delay and the fewest stops. Overall, the results provide an initial signal timing scheme for practitioners given a crossover spacing, an increased volume, or both. Future work will include exploring low volumes, balanced interchange volumes, and their effects on the four-critical-movement phasing scheme, as well as the effect of closely spaced adjacent intersections.

scholarly journals Cross-situational word learning with multimodal neural networks

2021 ◽  
Author(s):  
Wai Keen Vong ◽  
Brenden M. Lake
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Neural Networks ◽  
Word Learning ◽  
Network Architecture ◽  
Question Answering ◽  
Cognitive Models ◽  
Learning Approaches ◽  
Neural Network Learning ◽  
Single Epoch

In order to learn the mappings from words to referents, children must integrate co-occurrence information across individually ambiguous pairs of scenes and utterances, a challenge known as cross-situational word learning. In machine learning, recent multimodal neural networks have been shown to learn meaningful visual-linguistic mappings from cross-situational data, as needed to solve problems such as image captioning and visual question answering. These networks are potentially appealing as cognitive models because they can learn from raw visual and linguistic stimuli, something previous cognitive models have not addressed. In this paper, we examine whether recent machine learning approaches can help explain various behavioral phenomena from the psychological literature on cross-situational word learning. We consider two variants of a multimodal neural network architecture, and look at seven different phenomena associated with cross-situational word learning, and word learning more generally. Our results show that these networks can learn word-referent mappings from a single epoch of training, matching the amount of training found in cross-situational word learning experiments. Additionally, these networks capture some, but not all of the phenomena we studied, with all of the failures related to reasoning via mutual exclusivity. These results provide insight into the kinds of phenomena that arise naturally from relatively generic neural network learning algorithms, and which word learning phenomena require additional inductive biases.

Sign in / Sign up

Export Citation Format

Share Document