reformatted learning, some fixes

Author: pascalwhoop

src/chaps/artificialintelligence.tex

@@ -31,6 +31,6 @@ \section{Artificial Intelligence}%
 \citep[p.24ff.]{russell2016artificial, arulkumaran2017brief}. The reasons for this are diverse but it can be argued that
 the combination of readily available computing power through cloud computing and advances in the mathematical
 underpinnings have allowed for fast-paced advances in recent years. Also, the currently very popular \acf {NN}
-architectures often require large amouts of data to learn which have lately been readily available for companies and
-:esearchers through the adoption of online technologies by the majority of the population
+architectures often require large amounts of data to learn which have lately been readily available for companies and
+researchers through the adoption of online technologies by the majority of the population
 \citep[p.27]{russell2016artificial}.

src/chaps/learning.tex

@@ -1,34 +1,29 @@
-According to \cite{russell2016artificial}, learning agents are those that
-\emph{improve their performance on future tasks after making observations about
-the world} \cite[p.693]{russell2016artificial}. Learning behavior is present in
-many species most notably humans. To create a learning algorithm means that the
-creator did not have to anticipate every potential variant of an environment
-that the agent is confronted with while still creating an agent that can act
-successfully in such environments. Cognitive Sciences define learning as the
-change of state due to experiences as a necessary requirement and often limit
-the recognition of learning to some observable behavior
-\cite[p.96f.]{cognition1999}. This applies to all known species and the same
-definition can easily be applied to a learning artificial agent. A learning
-agent that doesn't change its behavior is not very helpful and an agent that
+According to \cite{russell2016artificial}, learning agents are those that \emph{improve their performance on future
+tasks after making observations about the world} \cite[p.693]{russell2016artificial}. Learning behavior is present in
+many species most notably humans. To create a learning algorithm means that the creator did not have to anticipate every
+potential variant of an environment that the agent is confronted with while still creating an agent that can act
+successfully in such environments. Cognitive Sciences define learning as the change of state due to experiences as a
+necessary requirement and often limit the recognition of learning to some observable behavior
+\cite[p.96f.]{cognition1999}. This applies to all known species and the same definition can easily be applied to a
+learning artificial agent. A learning agent that doesn't change its behavior is not very helpful and an agent that
 doesn't change its state can hardly have learned something. 
 
-The \ac {AI} community has for many years employed a \emph{loss function} as a
-measure of learning progress. Loss functions describe the difference between the
-actual utility of the right actions versus the results of the agents learned
-actions. The exact loss function might be a mean squared error function or an
-absolute loss depending on the learning algorithm that is used. 
+The \ac {AI} community has for many years employed a \emph{loss function} as a measure of learning progress. Loss
+functions describe the difference between the actual utility of the right actions versus the results of the agents
+learned actions. The exact loss function might be a mean squared error function or an absolute loss depending on the
+learning algorithm that is used. 
 
-Computational learning theory looks at many different problems of learning: How
-to learn through a large number of examples, the effects of learning when the
-agent already knows something, how to learn without examples, how to learn
-through feedback from the environment and how to learn if the origin of the
-feedback is not deterministic \cite[]{russell2016artificial}. In this work, two
-of those problems are of special interest: The ability to learn from previously
-labelled examples and the ability to learn through feedback from the
-environment. The former is called \ac {SL}  and the latter is mostly
-referred to as \ac {RL}. 
+Computational learning theory looks at many different problems of learning: How to learn through a large number of
+examples, the effects of learning when the agent already knows something, how to learn without examples, how to learn
+through feedback from the environment and how to learn if the origin of the feedback is not deterministic
+\cite[]{russell2016artificial}. In this work, two of those problems are of special interest: The ability to learn from
+previously labelled examples and the ability to learn through feedback from the environment. The former is called \acl
+{SL}  and the latter is mostly referred to as \acl {RL}. To understand the difference, it is also important to
+understand algorithms that don't have access to labels for existing data, yet are still able to derive value from the
+information. These belong to the class of \acf {UL}. Although this class is not heavily relied upon in the
+implementation of the actual agent in the later practical implementation, it is crucial for many tasks in machine
+learning such as data exploration or anomality recognition. 
 
-The following sections will cover both mentioned forms of learning and
-Section~\ref{sec:neural_networks} will introduce an architecture that can be
-used as the learning function in these learning problems. Finally,
+The following sections will describe both \acl {SL} and \acl {UL} and Section~\ref{sec:neural_networks} will introduce
+an architecture that can be used as the learning function in these learning problems. Finally,
 Section~\ref{sec:Backpropagation} will explain how exactly \ac {NN} learn.