2019.11 News Weekly 2

Uber says it may have to pay Waymo for self-driving car technology

Liu, Yang, et al. “Attention-Based Deep Ensemble Net for Large-Scale Online Taxi-Hailing Demand Prediction.” IEEE Transactions on Intelligent Transportation Systems (2019).

• Motivation: effectively ensemble different base models is a challenging but extremely valuable task.
• Methods:
  • construction of an ensemble framework designed for spatio-temporal data to predict large-scale online taxi-hailing demand, where an attention-based deep ensemble net is designed to enhance the prediction accuracy.
  • Three attention blocks to model the inter-channel relationship, inter=spatial relationship and position relationship of the feature maps.
  • The proposed method is a kind of commonly used ensemble method which applies to large-scale spatio-temporal prediction.

Gohorbani, Amirata, et al. “Towards automatic concept-based explanations.” (2019).

• Motivation: interpretability has become an important topic of research
• Research gap:
  • Most of the current explanation methods provide explanations through feature importance scores, which identify features that are important for each individual input.
  • However, how to systematically summarize and interpret such pre sample feature importance scores itself is challenging.
• Methods:
  • propose principles and desiderata for concept based explanation, which goes beyond per-sample features to identify heigher level human-undersantable concepts that apply across the entire dataset.
  • develop a new algorithm, ACE, to automatically extract visual concepts.

科学家心忧：电动车电池未来将引发电子垃圾危机

Strale, Mathieu. “Sustainable urban logistics: What are we talking about?.” Transportation Research Part A: Policy and Practice 130 (2019): 745-751.

• Motivation: sustainable urban logistics is a trendy research theme that addresses societal, environmental and industrial challenges.
• Research goal: to examine how the academic literature investigates sustainable urban logistics by using a survey.
• Results:
  • Technical problems, the search for optimization and solutions and the dissemination of good practices dominate the topics and methods mobilized.
  • the literature pays scant attention to environmental and social issues, which are an important part of sustainable development.
• sustainable urban logistics – Reducing transport-related impacts – Reducing nuisances linked to logistics buildings – Reducing energy and material consumption – Developing logistics activities as a tool for economic growth.

Kindermans, Pieter-Jan, et al. “The (un) reliability of saliency methods.” Explainable AI: Interpreting, Explaining and Visualizing Deep Learning. Springer, Cham, 2019. 267-280.

• Motivation: Saliency methods aim to explain the predictions of deep neural networks.
• Results:
  • These methods lack reliability when the explanation is sensitive to factors that do not contribute to the model prediction.
  • we use a simple and common pre-processing step - adding a constant shift to the input data - to show that a transformation with no effect on the model can cause numerous methods to incorrectly attribute.
  • In order to guarantee reliability, we posit that methods should fulfill input invariance, the requirement that a saliency method mirror the sensitivity of the model with respect to transformations of the input.

Waymo重组在美业务，加速无人驾驶出租车落地

清华大学苏州汽车研究院院长成波：智能网联汽车六大趋势及中国方案

Cruise 每周两次更新自动驾驶汽车的 AI 大脑