Background - Smartphone-Based Human Activity Recognition - page 22

in various applications (Bishop 2006 ). For example, when data are assumed to
be Gaussian-distributed, its is possible to learn the model only by calculating the
mean and variance of the input data.

• Artificial Neural Networks (ANN): is an ML approach with a biological inspira-

tion. It simulates how the brain and its nervous system, composed of interconnected
neurons, is able to learn from experience and capture the underlying structure of
the data. Neurons are set in a layered structure and have associated weights which
are able to adapt based on the training data and the network output through a cost
function. This approach has shown to perform well in many applications (e.g.
(LeCun et al. 1989 ), including non-linear problems. Its main disadvantage relies
in the need of a large dataset for its training stage. Multilayer Perceptron (MLP)
is a popular ANN model that maps the input through multiple layers of neurons
in a fully connected directed graph until reaching the output.

• Logistic Regression (LR) is a probabilistic algorithm used for solving classifi-

cation and regression problems. It estimates the probability of a given sample
of belonging to a particular class. This is achieved through the use of a logistic
function which is modeled by fitting the training data generally using maximum
likelihood estimation (Landwehr et al. 2005 ).

2.5.3 Support Vector Machines

A Support vector machine is one of the most commonly used supervised ML algo-
rithms. It was initially proposed by Vladimir Vapnik and his colleagues in (Boser
et al. 1992 ; Cortes and Vapnik 1995 ) with the aim of solving linear and non-linear
binary classification problems. Afterward, this algorithm has been adapted for its
application in multiclass classification and regression analysis (Drucker et al. 1997 ;
Weston and Watkins 1998 ).
The SVM for classification is a deterministic approach that aims to find the hyper-
planes that best separate the data into classes. These subspaces are the ones that
provide the largest margin separation from the classes of the training data with the
intention of providing a model with low generalization error for its use with unseen
data samples.
SVMs are the basis for the classification of activities in this work. For this reason
we now introduce them, starting from the binary SVM model which is its simplest
representation, to the extended case that allows the classification of more than two
classes: the multiclass SVM. This algorithm will be further revised throughout the
development of this research to tackle specific requirements for our application in
aspects such as kernel type, arithmetic used and algorithm output type.