Initialise.QLearningPersExpPath: Set changeable model variables
In MartinKies/RLR: Reinforcement Learning with R

Returns a list with the following items

epsilon [only in case of epsilon.greedy] Specifies how often the Algorithm tries a random move. Initialized with epsilon.start of model.par.
memory Memory of the Algorithm given past games. May be intitialised to give a starting point to learning.

Initialise.QLearningPersExpPath(
  game.object = NULL,
  algo.par,
  memory.init = NULL,
  memory.param = NULL,
  model.par = NULL
)

`game.object`	A Game Object as defined by `Get.Game.Object.<Name>`. Necessary in the case of memory intitialisation.
`memory.init`	Which type of initialization should take place? It `NULL`, the option `none` is used. The following types are supported none No initialization takes place. Memory is an empty list. self.play The other strategies play against themselves - to understand possible secret handshakes. The following `memory.param` are expected: no How often should the other strategies play against themselves? If combinations of different memories are needed, one can use the function `Extend.Memory.QLearningPersExpPath()`
`memory.param`	Parameters necessary for the chosen `memory.init`.
`model.par`	Parameters of model i.e. Neural Network. Currently only used for RNN.