Digital Cities Roadmap. Группа авторов

learning method is facilitated across a number of ML frameworks and resources. The challenge of choosing the best platform in order to data analytics flow sharing can also be achieved from alternative viewpoints challenging despite the growing amount of such toolkits. There is generally no one toolkit that completely suits all challenges (Table 1.6) and includes remedies. Some of the toolkits available could overlap, with benefits and drawbacks.

      Table 1.6 Difference between deep learning and machine learning tools [56].

Tool	Creator	OS	Open source?	Written In	Interface	CUDA support?	Algorithms	Release date
Tensor Flow	Google Brain team	Linux. Mac OS X (Windows support on road map	Yes	C++, Python	Python, C/ C++	Yes	Deep learning algorithm: RNN, CN, RBM and DUN.	Novembeir 2015
Theano	Universit de Montral	Cross-platform	Yes	Python	Python	Yes	Deep learning algorithm: RNN, CN, RBM and DBN.	September 2007
H20	H20.ai	Linux, Mac OS, Microsoft Windows And Cross-platform inch Apache HDFS; Amazon EC2, Google Compute Engine, and Microsoft Azure.	Yes	Java, Scala, Python, R	Python, R	No	Algorithms for classification, clustering, generalized linear models, statistical analysis, ensembles, optimization tools, data pre-processing options and deep neural networks.	August 2011
Deeplearning4j	Various. Original author Adam Gibson	Linux, OSX, Windows, Android, CyanogenMod (Cross-platform)	Yes	Java, Scala, C, CUDA	Java, Scala, CIo-jure	Yes	Deep learning algorithms including: RBM, DBN. RNN. deep autoencoder	August 2013
MLlib Spark	Apache Software Foundation. UC Berkeley AMPLab, Databricks	Microsoft Windows, OS X, Linux	Yes	Scala. Java, Python, R	Scala, Java, Python. R	No	Classification, regression, clustering, dimensionality reduction, and collaborative filtering	May 2014
Azure	Dave Cutler from Microsoft	Microsoft Windows, Linux	No	C++	C++, Java. ASP.NET, PHP. Nodejs, Python	Yes	Classification, regression, clustering	October 2010
Torch	Ronan Collobert, Koray Kavukcuoglu, Clement Farabet	Linux. Android, Mac OS XTiOS	Yes	C, Lua	Lua, LuaJlT. C, utility library for C++/ OpenCL	Yes	Deep algorithms	October 2002
MOA	University of Waikato	Cross-platform	Yes	Java	GUI, the commandline. and Java	No	ML algorithms (classification, regression, clustering, outlier detection, concept drift detection and recommender systems)	November 2014
Caffe	Berkeley Vision and Learning Center. community contributors	Ubuntu, OS X, AWS, unofficial Android port, Windows support by Microsoft Research, unofficial Windows port	Yes	C++, Python	C++, command line. Python, MATLAB	Yes	Deep learning algorithms: CN, and RNN	December 2012

       1.10.7 Big Data Research Applications for SBs in Real-Time

Many systems need a stream data processing in real time, so it is not feasible for this form of program to wait for data to be stored so evaluated. Stream processing is typically configured to interpret large volumes of data and operate on data streaming in real time by using constant queries, including SQL queries, to manage streaming data in real time, using an interface that is robust, accessible and fault resistant (Table 1.7).

       1.10.8 Implementation of the ML Concept in the SB Context

      Figure 1.25 illustrates specific measures to forecast an event in the SB sense by utilizing ML methods.

      On the other hand, the aim of optimization is to optimize long-lasting gains by proper decisions. Strengthening learning with these issues can be used. Many optimization issues may be treated as predicting issues such that benefit is estimated for different activities and the activity with the largest income is chosen. The most important form in optimization is decision-making. A variety of factors and compromises about the effects of specific environmental locations need to be addressed.

       Smart Building Services Taxonomy

      The taxonomy of SB resources essential domains is shown in Figure 1.26. Lighting service connects the well-being of occupants in SBs that have sensors that save energy when lights are not needed, based on their operation. Power and electricity can supply a percentage of SB power consumption with renewable energy sources. HVAC implies the heating, ventilation and air conditioning device, built for the comfort of citizens and an efficient ambient contact. The water resources program aims at growing conservation and maximizing resource recovery for water supply.

      Smart building service taxonomy is related to the maintenance of electronic doors, biometrics and SB security cameras devices. The Control Center offers management and decision-making for apps. The automated apps