Learning Deconvolution Network for Semantic Segmentation

Today, intelligent systems that offer artificial intelligence capabilities often rely on machine learning. Machine learning describes the capacity of systems to learn from problem-specific training data to automate the process of analytical model building and solve associated tasks. Deep learning is a machine learning concept based on artificial neural networks. For many applications, deep learning models outperform shallow machine learning models and traditional data analysis approaches. In this article, we summarize the fundamentals of machine learning and deep learning to generate a broader understanding of the methodical underpinning of current intelligent systems. In particular, we provide a conceptual distinction between relevant terms and concepts, explain the process of automated analytical model building through machine learning and deep learning, and discuss the challenges that arise when implementing such intelligent systems in the field of electronic markets and networked business. These naturally go beyond technological aspects and highlight issues in human-machine interaction and artificial intelligence servitization.

/pdf/machine-learning-and-deep-learning-mwd39usczr.pdf

Machine learning and deep learning

An Archetypal Analysis on

The growing size of cities and increasing population mobility have determined a rapid increase in the number of vehicles on the roads, which has resulted in many challenges for road traffic management authorities in relation to traffic congestion, accidents, and air pollution. Over the recent years, researchers from both industry and academia have been focusing their efforts on exploiting the advances in sensing, communication, and dynamic adaptive technologies to make the existing road traffic management systems (TMSs) more efficient to cope with the aforementioned issues in future smart cities. However, these efforts are still insufficient to build a reliable and secure TMS that can handle the foreseeable rise of population and vehicles in smart cities. In this survey, we present an up-to-date review of the different technologies used in the different phases involved in a TMS and discuss the potential use of smart cars and social media to enable fast and more accurate traffic congestion detection and mitigation. We also provide a thorough study of the security threats that may jeopardize the efficiency of the TMS and endanger drivers' lives. Furthermore, the most significant and recent European and worldwide projects dealing with traffic congestion issues are briefly discussed to highlight their contribution to the advancement of smart transportation. Finally, we discuss some open challenges and present our own vision to develop robust TMSs for future smart cities.

/pdf/a-communications-oriented-perspective-on-traffic-management-3peyi5cewi.pdf

A Communications-Oriented Perspective on Traffic Management Systems for Smart Cities: Challenges and Innovative Approaches

Heating, cooling and ventilation accounts for 35% energy usage in the United States. Currently, most modern buildings still condition rooms assuming maximum occupancy rather than actual usage. As a result, rooms are often over-conditioned needlessly. Thus, in order to achieve efficient conditioning, we require knowledge of occupancy. This paper shows how real time occupancy data from a wireless sensor network can be used to create occupancy models which in turn can be integrated into building conditioning system for usage based demand control conditioning strategies. Using strategies based on sensor network occupancy model predictions, we show that it is possible to achieve 42% annual energy savings while still maintaining American Society of Heating, Refrigerating and Air-Conditioning (ASHRAE) comfort standards.

http://faculty.ucmerced.edu/mcarreira-perpinan/papers/ipsn11.pdf

OBSERVE: Occupancy-based system for efficient reduction of HVAC energy

We demonstrate that CNN deep neural networks can not only be used for making predictions based on multivariate time series data, but also for explaining these predictions. 
This is important for a number of applications where predictions are the basis for decisions and actions. Hence, confidence in the prediction result is crucial. 
We design a two stage convolutional neural network architecture which uses particular kernel sizes.
This allows us to utilise gradient based techniques for generating saliency maps for both the time dimension and the features.
These are then used for explaining which features during which time interval are responsible for a given prediction, as well as explaining during which time intervals was the joint contribution of all features most important for that prediction. 
We demonstrate our approach for predicting the average energy production of photovoltaic power plants and for explaining these predictions.

/pdf/explainable-deep-neural-networks-for-multivariate-time-smlbjaohl9.pdf

Explainable Deep Neural Networks for Multivariate Time Series Predictions.

The Internet of Things is one of the next big changes in which devices, objects, and sensors are getting linked to the semantic web. However, the increasing availability of generated data leads to new integration problems. In this paper we present an architecture and approach that illustrates how semantic sensor networks, semantic web technologies, and reasoning can help in real-world applications to automatically derive complex models for analytics tasks such as prediction and diagnostics. We demonstrate our approach for buildings and their numerous connected sensors and show how our semantic framework allows us to detect and diagnose abnormal building behavior. This can lead to not only an increase of occupant well-being but also to a reduction of energy use. Given that buildings consume 40% of the world's energy use we therefore also make a contribution towards global sustainability. The experimental evaluation shows the benefits of our approach for buildings at IBM's Technology Campus in Dublin.

Adapting Semantic Sensor Networks for Smart Building Diagnosis

Almost 32% of the total energy consumption in industrialized countries is used for electricity, heating, ventilation, and air-conditioning (HVAC) in buildings. Deploying Energy Management Systems (EMS) helps reducing energy use. Unfortunately it is a complex task that requires to identify the EMS inputs among thousands of sensors in a building. Since most of these sensors lack any labeling standard this is currently done manually. We aim to semi-automate this mapping task and address the problem of identifying EMS inputs with minimal user involvement. This is achieved by utilizing linguistic and semantic techniques for computing similarity values between labels of sensors and EMS inputs. Experiments show that our approach can be successfully applied to real-world data.

Towards automating the deployment of energy saving approaches in buildings

Diagnosis, or the method to connect causes to its effects, is an important reasoning task for obtaining insight on cities and reaching the concept of sustainable and smarter cities that is envisioned nowadays. This paper, focusing on transportation and its road traffic, presents how road traffic congestions can be detected and diagnosed in quasi real-time. We adapt pure Artificial Intelligence diagnosis techniques to fully exploit knowledge which is captured through relevant semantics-augmented stream and static data from various domains. Our prototype of semantic-aware diagnosis of road traffic congestions, experimented in Dublin Ireland, works efficiently with large, heterogeneous information sources and delivers value-added services to citizens and city managers in quasi real-time.

/pdf/applying-semantic-web-technologies-for-diagnosing-road-2b0c7mm34r.pdf

Applying semantic web technologies for diagnosing road traffic congestions

In this work we present MTEX-CNN, a novel explainable convolutional neural network architecture which can not only be used for making predictions based on multivariate time series data, but also for explaining these predictions. The network architecture consists of two stages and utilizes particular kernel sizes. This allows us to apply gradient based methods for generating saliency maps for both the time dimension and the features. The first stage of the architecture explains which features are most significant to the predictions, while the second stage explains which time segments are the most significant. We validate our approach on two use cases, namely to predict rare server outages in the wild, as well as the average energy production of photovoltaic power plants based on a benchmark data set. We show that our explanations shed light over what the model has learned. We validate this by retraining the network using the most significant features extracted from the explanations and retaining similar performance to training with the full set of features.

Anika Schumann

Papers

Explainable Deep Neural Networks for Multivariate Time Series Predictions.

Adapting Semantic Sensor Networks for Smart Building Diagnosis

Towards automating the deployment of energy saving approaches in buildings

Applying semantic web technologies for diagnosing road traffic congestions

MTEX-CNN: Multivariate Time Series EXplanations for Predictions with Convolutional Neural Networks