verve Namespace Reference

The main namespace that contains everything. More...

Classes
class	ActiveTDConnectionList
	A dynamic list of "active" TDConnections (those with a non-zero eligibility trace). More...
class	Agent
	An Agent is an autonomous entity that learns from direct with its environment. More...
struct	ContinuousSensorDescriptor
	A data structure describing a continuous sensor. More...
struct	DiscreteSensorDescriptor
	A data structure describing a discrete sensor. More...
class	AgentDescriptor
	A data structure used for Agent creation. More...
class	Connection
	A synaptic connection between two Neurons. More...
class	Logger
	This class is used to log any events, errors, or warnings that may come up. More...
class	Neuron
	The basic Neuron class. More...
class	Observation
	A simple data structure containing arrays of discrete and/or continuous sensory input data. More...
class	Population
	An interface for a group of Neurons of similar function. More...
class	PredictiveModel
	A PredictiveModel learns a predictive model of the environment dynamics (transitions) from direct experience. More...
class	Projection
	An interface for a group of Connections from one group of Neurons to another. More...
struct	RBFInputData
	A convenient data structure for passing around common sets of data. More...
class	RBFNeuron
	A "radial basis function" Neuron. More...
class	RBFPopulation
	A group of RBFNeurons. More...
class	RLModule
	An RLModule learns from reinforcements to improve its action selection in order to increase its future reinforcement intake. More...
class	TDConnection
	A Connection that is trainable via temporal difference learning. More...
class	TDProjection
	A Projection of TDConnections. More...
class	UltraSparseCodePopulation
	A Population with only one active Neuron. More...
class	WinnerTakeAllPopulation
	A Population with a single active Neuron. More...
Namespaces
namespace	defaults
namespace	globals
Typedefs
typedef float	real
Enumerations
enum	AgentArchitecture { RL, MODEL_RL, CURIOUS_MODEL_RL }
enum	InitialWeightMethod { IDEAL_NOISE, WEIGHTS_NEAR_0, WEIGHTS_NEAR_1 }
enum	RBFActivationCode { HIGH_ACTIVATION, LOW_ACTIVATION, NO_ACTIVATION }
enum	TDConnectionType { VALUE_FUNCTION_TDCONNECTION, POLICY_TDCONNECTION }
Functions
VERVE_EXPORT_FUNCTION Agent *VERVE_CALL	createAgent (const AgentDescriptor &desc)
Variables
const real	VERVE_E = (real)2.71828182845904523536

---

Detailed Description

The main namespace that contains everything.

---

Typedef Documentation

typedef float verve::real

Definition at line 40 of file Defines.h.

---

Enumeration Type Documentation

enum verve::AgentArchitecture

Various Agent architectures. Enumerator: RL  Reinforcement learning only. MODEL_RL  Reinforcement learning with a learned predictive environment model. CURIOUS_MODEL_RL  Reinforcement learning, a predictive model, and a curiosity component that drives the Agent to experience new situations (hopefully improving its predictive model faster). Definition at line 46 of file Defines.h.

enum verve::InitialWeightMethod

Various methods used to set the weights of new Connections. Enumerator: IDEAL_NOISE  Apply Gaussian noise using an ideal variance level. WEIGHTS_NEAR_0  Set new Connection weights to random values very close to 0. WEIGHTS_NEAR_1  Set new Connection weights to random values very close to 1. Definition at line 36 of file Projection.h.

enum verve::RBFActivationCode

Different values used as a coarse representation of how much a data point activates the RBF. This is useful for keeping track of active RBFNeurons and deciding when to create new RBFs. Enumerator: HIGH_ACTIVATION  The data point is close enough to give the RBF high activation, so we can avoid creating any new RBFs. LOW_ACTIVATION  The data point is close enough to affect the RBF's activation, but a new RBF might need to be created. NO_ACTIVATION  The data point is far enough away from the RBF that it does not affect its activation at all. We might need to create a new RBF. Definition at line 36 of file RBFNeuron.h.

enum verve::TDConnectionType

Different types of TDConnections that must be handled differently. Enumerator: VALUE_FUNCTION_TDCONNECTION  A TDConnection used in the value function representation. POLICY_TDCONNECTION  A TDConnection used in the policy representation. Definition at line 34 of file TDConnection.h.

---

Function Documentation

VERVE_EXPORT_FUNCTION Agent *VERVE_CALL verve::createAgent (  const AgentDescriptor &  desc  )

Dynamically creates a new Agent using the given AgentDescriptor. Always use this instead of "new Agent" to ensure that memory is allocated from the correct heap. Definition at line 31 of file Agent.cpp.

---

Variable Documentation

const real verve::VERVE_E = (real)2.71828182845904523536

Definition at line 43 of file Defines.h. Referenced by verve::globals::calcDecayConstant().

---