From the article: Abstract: An overview of neural network architectures is presented. Some of these architectures have been created in recent years, whereas others originate from many decades ago. Apart from providing a practical tool for comparing deep learning models, the Neural Network Zoo also uncovers a taxonomy of network architectures, their chronology, and traces back lineages and inspirations for these neural information processing systems.
DOCUMENT
We present a novel architecture for an AI system that allows a priori knowledge to combine with deep learning. In traditional neural networks, all available data is pooled at the input layer. Our alternative neural network is constructed so that partial representations (invariants) are learned in the intermediate layers, which can then be combined with a priori knowledge or with other predictive analyses of the same data. This leads to smaller training datasets due to more efficient learning. In addition, because this architecture allows inclusion of a priori knowledge and interpretable predictive models, the interpretability of the entire system increases while the data can still be used in a black box neural network. Our system makes use of networks of neurons rather than single neurons to enable the representation of approximations (invariants) of the output.
LINK
A considerable amount of literature has been published on Corporate Reputation, Branding and Brand Image. These studies are extensive and focus particularly on questionnaires and statistical analysis. Although extensive research has been carried out, no single study was found which attempted to predict corporate reputation performance based on data collected from media sources. To perform this task, a biLSTM Neural Network extended with attention mechanism was utilized. The advantages of this architecture are that it obtains excellent performance for NLP tasks. The state-of-the-art designed model achieves highly competitive results, F1 scores around 72%, accuracy of 92% and loss around 20%.
DOCUMENT
Completeness of data is vital for the decision making and forecasting on Building Management Systems (BMS) as missing data can result in biased decision making down the line. This study creates a guideline for imputing the gaps in BMS datasets by comparing four methods: K Nearest Neighbour algorithm (KNN), Recurrent Neural Network (RNN), Hot Deck (HD) and Last Observation Carried Forward (LOCF). The guideline contains the best method per gap size and scales of measurement. The four selected methods are from various backgrounds and are tested on a real BMS and meteorological dataset. The focus of this paper is not to impute every cell as accurately as possible but to impute trends back into the missing data. The performance is characterised by a set of criteria in order to allow the user to choose the imputation method best suited for its needs. The criteria are: Variance Error (VE) and Root Mean Squared Error (RMSE). VE has been given more weight as its ability to evaluate the imputed trend is better than RMSE. From preliminary results, it was concluded that the best K‐values for KNN are 5 for the smallest gap and 100 for the larger gaps. Using a genetic algorithm the best RNN architecture for the purpose of this paper was determined to be Gated Recurrent Units (GRU). The comparison was performed using a different training dataset than the imputation dataset. The results show no consistent link between the difference in Kurtosis or Skewness and imputation performance. The results of the experiment concluded that RNN is best for interval data and HD is best for both nominal and ratio data. There was no single method that was best for all gap sizes as it was dependent on the data to be imputed.
DOCUMENT
Completeness of data is vital for the decision making and forecasting on Building Management Systems (BMS) as missing data can result in biased decision making down the line. This study creates a guideline for imputing the gaps in BMS datasets by comparing four methods: K Nearest Neighbour algorithm (KNN), Recurrent Neural Network (RNN), Hot Deck (HD) and Last Observation Carried Forward (LOCF). The guideline contains the best method per gap size and scales of measurement. The four selected methods are from various backgrounds and are tested on a real BMS and metereological dataset. The focus of this paper is not to impute every cell as accurately as possible but to impute trends back into the missing data. The performance is characterised by a set of criteria in order to allow the user to choose the imputation method best suited for its needs. The criteria are: Variance Error (VE) and Root Mean Squared Error (RMSE). VE has been given more weight as its ability to evaluate the imputed trend is better than RMSE. From preliminary results, it was concluded that the best K‐values for KNN are 5 for the smallest gap and 100 for the larger gaps. Using a genetic algorithm the best RNN architecture for the purpose of this paper was determined to be GatedRecurrent Units (GRU). The comparison was performed using a different training dataset than the imputation dataset. The results show no consistent link between the difference in Kurtosis or Skewness and imputation performance. The results of the experiment concluded that RNN is best for interval data and HD is best for both nominal and ratio data. There was no single method that was best for all gap sizes as it was dependent on the data to be imputed.
MULTIFILE
Machine learning models have proven to be reliable methods in classification tasks. However, little research has been done on classifying dwelling characteristics based on smart meter & weather data before. Gaining insights into dwelling characteristics can be helpful to create/improve the policies for creating new dwellings at NZEB standard. This paper compares the different machine learning algorithms and the methods used to correctly implement the models. These methods include the data pre-processing, model validation and evaluation. Smart meter data was provided by Groene Mient, which was used to train several machine learning algorithms. The models that were generated by the algorithms were compared on their performance. The results showed that Recurrent Neural Network (RNN) 2performed the best with 96% of accuracy. Cross Validation was used to validate the models, where 80% of the data was used for training purposes and 20% was used for testing purposes. Evaluation metrices were used to produce classification reports, which can indicate which of the models work the best for this specific problem. The models were programmed in Python.
DOCUMENT
The huge number of images shared on the Web makes effective cataloguing methods for efficient storage and retrieval procedures specifically tailored on the end-user needs a very demanding and crucial issue. In this paper, we investigate the applicability of Automatic Image Annotation (AIA) for image tagging with a focus on the needs of database expansion for a news broadcasting company. First, we determine the feasibility of using AIA in such a context with the aim of minimizing an extensive retraining whenever a new tag needs to be incorporated in the tag set population. Then, an image annotation tool integrating a Convolutional Neural Network model (AlexNet) for feature extraction and a K-Nearest-Neighbours classifier for tag assignment to images is introduced and tested. The obtained performances are very promising addressing the proposed approach as valuable to tackle the problem of image tagging in the framework of a broadcasting company, whilst not yet optimal for integration in the business process.
DOCUMENT
Current methods for energy diagnosis in heating, ventilation and air conditioning (HVAC) systems are not consistent with process and instrumentation diagrams (P&IDs) as used by engineers to design and operate these systems, leading to very limited application of energy performance diagnosis in practice. In a previous paper, a generic reference architecture – hereafter referred to as the 4S3F (four symptoms and three faults) framework – was developed. Because it is closely related to the way HVAC experts diagnose problems in HVAC installations, 4S3F largely overcomes the problem of limited application. The present article addresses the fault diagnosis process using automated fault identification (AFI) based on symptoms detected with a diagnostic Bayesian network (DBN). It demonstrates that possible faults can be extracted from P&IDs at different levels and that P&IDs form the basis for setting up effective DBNs. The process was applied to real sensor data for a whole year. In a case study for a thermal energy plant, control faults were successfully isolated using balance, energy performance and operational state symptoms. Correction of the isolated faults led to annual primary energy savings of 25%. An analysis showed that the values of set probabilities in the DBN model are not outcome-sensitive. Link to the formal publication via its DOI https://doi.org/10.1016/j.enbuild.2020.110289
DOCUMENT
In this article a generic fault detection and diagnosis (FDD) method for demand controlled ventilation (DCV) systems is presented. By automated fault detection both indoor air quality (IAQ) and energy performance are strongly increased. This method is derived from a reference architecture based on a network with 3 generic types of faults (component, control and model faults) and 4 generic types of symptoms (balance, energy performance, operational state and additional symptoms). This 4S3F architecture, originally set up for energy performance diagnosis of thermal energy plants is applied on the control of IAQ by variable air volume (VAV) systems. The proposed method, using diagnosis Bayesian networks (DBNs), overcomes problems encountered in current FDD methods for VAV systems, problems which inhibits in practice their wide application. Unambiguous fault diagnosis stays difficult, most methods are very system specific, and finally, methods are implemented at a very late stage, while an implementation during the design of the HVAC system and its control is needed. The IAQ 4S3F method, which solves these problems, is demonstrated for a common VAV system with demand controlled ventilation in an office with the use of a whole year hourly historic Building Management System (BMS) data and showed it applicability successfully. Next to this, the influence of prior and conditional probabilities on the diagnosis is studied. Link to the formal publication via its DOI https://doi.org/10.1016/j.buildenv.2019.106632
DOCUMENT
The Heating Ventilation and Air Conditioning (HVAC) sector is responsible for a large part of the total worldwide energy consumption, a significant part of which is caused by incorrect operation of controls and maintenance. HVAC systems are becoming increasingly complex, especially due to multi-commodity energy sources, and as a result, the chance of failures in systems and controls will increase. Therefore, systems that diagnose energy performance are of paramount importance. However, despite much research on Fault Detection and Diagnosis (FDD) methods for HVAC systems, they are rarely applied. One major reason is that proposed methods are different from the approaches taken by HVAC designers who employ process and instrumentation diagrams (P&IDs). This led to the following main research question: Which FDD architecture is suitable for HVAC systems in general to support the set up and implementation of FDD methods, including energy performance diagnosis? First, an energy performance FDD architecture based on information embedded in P&IDs was elaborated. The new FDD method, called the 4S3F method, combines systems theory with data analysis. In the 4S3F method, the detection and diagnosis phases are separated. The symptoms and faults are classified into 4 types of symptoms (deviations from balance equations, operating states (OS) and energy performance (EP), and additional information) and 3 types of faults (component, control and model faults). Second, the 4S3F method has been tested in four case studies. In the first case study, the symptom detection part was tested using historical Building Management System (BMS) data for a whole year: the combined heat and power plant of the THUAS (The Hague University of Applied Sciences) building in Delft, including an aquifer thermal energy storage (ATES) system, a heat pump, a gas boiler and hot and cold water hydronic systems. This case study showed that balance, EP and OS symptoms can be extracted from the P&ID and the presence of symptoms detected. In the second case study, a proof of principle of the fault diagnosis part of the 4S3F method was successfully performed on the same HVAC system extracting possible component and control faults from the P&ID. A Bayesian Network diagnostic, which mimics the way of diagnosis by HVAC engineers, was applied to identify the probability of all possible faults by interpreting the symptoms. The diagnostic Bayesian network (DBN) was set up in accordance with the P&ID, i.e., with the same structure. Energy savings from fault corrections were estimated to be up to 25% of the primary energy consumption, while the HVAC system was initially considered to have an excellent performance. In the third case study, a demand-driven ventilation system (DCV) was analysed. The analysis showed that the 4S3F method works also to identify faults on an air ventilation system.
DOCUMENT
