Traffic Updates: Saying a Lot While Revealing a Little
Authors: John Krumm, Eric Horvitz986-995
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental data consists of freeway speeds and flows from the state of California in the USA. ... We used data from the first six months of 2017 for training and the last three months for testing. ... Our results show the inference error over 10,000 independent tests. |
| Researcher Affiliation | Industry | John Krumm, Eric Horvitz Microsoft Research Redmond, Washington USA jckrumm@microsoft.com, horvitz@microsoft.com |
| Pseudocode | No | The paper describes its methods verbally and mathematically but does not include any explicit pseudocode blocks or algorithms. |
| Open Source Code | No | The paper does not provide any statement or link indicating the release of source code for the described methodology. |
| Open Datasets | Yes | Our experimental data consists of freeway speeds and flows from the state of California in the USA. Traffic on roads is often characterized by its speed and flow, where flow indicates the number of vehicles passing a certain point in a given amount of time. Such data is available free via the California Performance Measurement System (Pe MS), which provides a wide variety of real-time and historical data for freeways in California (California Department of Transportation (Caltrans) 2018). |
| Dataset Splits | No | The paper states 'We used data from the first six months of 2017 for training and the last three months for testing.' and 'The median absolute V OI prediction error over all 109 stations was 0.058 mph*flow, based on an 80/20 train/test split of the Pe MS data from months 7-9.' It describes train and test splits, but does not explicitly mention a separate validation split for the main experiments. |
| Hardware Specification | No | The paper mentions 'On a conventional desktop PC' for running the BP code, but provides no specific hardware details like CPU/GPU models, memory, or other specifications for the experiments. |
| Software Dependencies | No | The paper describes the algorithms and models used but does not provide specific software dependencies with version numbers. |
| Experiment Setup | Yes | For all our MRF inferences, we discretized speeds into 5 mph bins, and we discretized flows into 25 vehicles/5 minutes bins. ... We stopped our program's message passing after the mean absolute difference in each message over time dropped below 0.1 or if the number of messaging iterations exceeded 100. ... We approximate the standard deviation of GPS as σg = 3 meters. Taking the vector difference of two measurements gives the velocity vector: ... For our simulation, we take ti = 2 seconds. |