Documentation
¶
Overview ¶
Package example_models provides factory functions that return fully parameterized, well-known Bayesian networks from the literature. Each function creates the network structure, adds CPDs with real probability values, sets state names, and validates the model.
Index ¶
- func Alarm() *models.BayesianNetwork
- func AlarmFull() *models.BayesianNetwork
- func Asia() *models.BayesianNetwork
- func Barley() *models.BayesianNetwork
- func Cancer() *models.BayesianNetwork
- func Child() *models.BayesianNetwork
- func CoinToss() *models.BayesianNetwork
- func DogProblem() *models.BayesianNetwork
- func Earthquake() *models.BayesianNetwork
- func FraudDetection() *models.BayesianNetwork
- func Get(name string) (*models.BayesianNetwork, error)
- func Hailfinder() *models.BayesianNetwork
- func Hepar2() *models.BayesianNetwork
- func Insurance() *models.BayesianNetwork
- func List() []string
- func MedicalDiagnosis() *models.BayesianNetwork
- func Mildew() *models.BayesianNetwork
- func MontyHall() *models.BayesianNetwork
- func Pathfinder() *models.BayesianNetwork
- func Pigs() *models.BayesianNetwork
- func Sachs() *models.BayesianNetwork
- func Student() *models.BayesianNetwork
- func Survey() *models.BayesianNetwork
- func VisitAsia() *models.BayesianNetwork
- func Water() *models.BayesianNetwork
- func WaterSprinkler() *models.BayesianNetwork
- func Win95pts() *models.BayesianNetwork
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func Alarm ¶
func Alarm() *models.BayesianNetwork
Alarm returns a simplified Alarm Bayesian network with 5 nodes. This is a compact version of the classic Burglary-Alarm network from Pearl (1988).
Nodes:
Burglary (B): {No, Yes}
Earthquake (E): {No, Yes}
Alarm (A): {No, Yes}
JohnCalls (J): {No, Yes}
MaryCalls (M): {No, Yes}
Edges: B->A, E->A, A->J, A->M
func AlarmFull ¶
func AlarmFull() *models.BayesianNetwork
AlarmFull returns the full ALARM (A Logical Alarm Reduction Mechanism) network with 37 nodes and 46 edges. From Beinlich et al. (1989). Structure only — CPDs too large to hardcode.
func Asia ¶
func Asia() *models.BayesianNetwork
Asia returns the Asia (chest clinic) Bayesian network with 8 nodes. This network was introduced by Lauritzen & Spiegelhalter (1988).
Nodes:
Asia (A): visit to Asia — {No, Yes}
Tub (T): tuberculosis — {No, Yes}
Smoke (S): smoker — {No, Yes}
Lung (L): lung cancer — {No, Yes}
Bronc (B): bronchitis — {No, Yes}
Either (E): tuberculosis or lung cancer — {No, Yes}
Xray (X): positive X-ray — {No, Yes}
Dysp (D): dyspnoea — {No, Yes}
Edges: A->T, S->L, S->B, T->E, L->E, E->X, E->D, B->D
func Barley ¶
func Barley() *models.BayesianNetwork
Barley returns the Barley crop yield network with 48 nodes and 84 edges. From Kristensen & Rasmussen (2002). Structure only — CPDs too large to hardcode.
func Cancer ¶
func Cancer() *models.BayesianNetwork
Cancer returns the simple Cancer diagnosis Bayesian network with 5 nodes.
Nodes:
Pollution (P): {Low, High}
Smoker (S): {No, Yes}
Cancer (C): {No, Yes}
Xray (X): {Negative, Positive}
Dyspnoea (D): {No, Yes}
Edges: P->C, S->C, C->X, C->D
func Child ¶
func Child() *models.BayesianNetwork
Child returns the Child congenital heart disease network with 20 nodes and 25 edges. From Spiegelhalter et al. (1992). Structure only — CPDs too large to hardcode.
func CoinToss ¶
func CoinToss() *models.BayesianNetwork
CoinToss returns a simple 3-node Bayesian network modeling a biased coin toss.
Nodes:
Bias: {Fair, Biased} — whether the coin is biased
FirstToss: {Heads, Tails}
SecondToss: {Heads, Tails}
Edges: Bias->FirstToss, Bias->SecondToss
func DogProblem ¶
func DogProblem() *models.BayesianNetwork
DogProblem returns a textbook Bayesian network about a dog's behavior. From Charniak (1991) "Bayesian Networks without Tears."
Nodes:
BowelProblem: {true, false}
DogOut: {true, false}
FamilyOut: {true, false}
HearBark: {true, false}
LightOn: {true, false}
Edges: BowelProblem->DogOut, FamilyOut->DogOut, FamilyOut->LightOn, DogOut->HearBark
func Earthquake ¶
func Earthquake() *models.BayesianNetwork
Earthquake returns the classic 5-node Earthquake/Burglary/Alarm network. This is an alias for the same model as Alarm() but with node names matching the bnlearn "earthquake" dataset. The structure is identical to the classic Pearl (1988) example.
This function is provided for compatibility with the bnlearn repository naming.
func FraudDetection ¶
func FraudDetection() *models.BayesianNetwork
FraudDetection returns a simple fraud detection Bayesian network with 6 nodes. This is a textbook-style network for credit card fraud detection.
Nodes:
Fraud: {No, Yes}
Age: {Young, Middle, Old}
Sex: {Male, Female}
ForeignPurchase: {No, Yes}
HighAmount: {No, Yes}
Alert: {No, Yes}
Edges: Fraud->ForeignPurchase, Fraud->HighAmount, Age->Fraud, Sex->Fraud,
ForeignPurchase->Alert, HighAmount->Alert
func Get ¶
func Get(name string) (*models.BayesianNetwork, error)
Get returns a BayesianNetwork for the given model name (case-insensitive). Returns an error if the model name is not recognized.
func Hailfinder ¶
func Hailfinder() *models.BayesianNetwork
Hailfinder returns the Hailfinder weather network with 56 nodes and 66 edges. From Abramson et al. (1996). Structure only — CPDs too large to hardcode.
func Hepar2 ¶
func Hepar2() *models.BayesianNetwork
Hepar2 returns the HEPAR II liver disease network with 70 nodes and 123 edges. From Onisko et al. (2001). Structure only — CPDs too large to hardcode.
func Insurance ¶
func Insurance() *models.BayesianNetwork
Insurance returns the Insurance evaluation network with 27 nodes and 52 edges. From Binder et al. (1997). Structure only — CPDs too large to hardcode.
func List ¶
func List() []string
List returns the names of all available example models, sorted alphabetically.
func MedicalDiagnosis ¶
func MedicalDiagnosis() *models.BayesianNetwork
MedicalDiagnosis returns a simple medical diagnosis Bayesian network with 8 nodes. A textbook model for reasoning about patient symptoms and diseases.
Nodes:
Smoking: {No, Yes}
Pollution: {Low, High}
LungCancer: {No, Yes}
Bronchitis: {No, Yes}
Fatigue: {No, Yes}
ChestPain: {No, Yes}
Cough: {No, Yes}
ShortnessOfBreath: {No, Yes}
Edges: Smoking->LungCancer, Smoking->Bronchitis, Pollution->LungCancer,
LungCancer->Fatigue, LungCancer->ChestPain, Bronchitis->Cough, Bronchitis->ShortnessOfBreath, LungCancer->ShortnessOfBreath
func Mildew ¶
func Mildew() *models.BayesianNetwork
Mildew returns the Mildew crop disease network with 35 nodes and 46 edges. From Kjærulff (1992). Structure only — CPDs too large to hardcode.
func MontyHall ¶
func MontyHall() *models.BayesianNetwork
MontyHall returns the Monty Hall problem as a Bayesian network with 3 nodes.
Nodes:
Prize: {Door1, Door2, Door3} — which door hides the prize
Guest: {Door1, Door2, Door3} — which door the guest picks
Host: {Door1, Door2, Door3} — which door the host opens
Edges: Prize->Host, Guest->Host
func Pathfinder ¶
func Pathfinder() *models.BayesianNetwork
Pathfinder returns the Pathfinder lymph-node disease network with 109 nodes and 195 edges. From Heckerman et al. (1992). Structure only — CPDs too large to hardcode.
func Pigs ¶
func Pigs() *models.BayesianNetwork
Pigs returns the Pigs genetic pedigree network with 441 nodes and 592 edges. From Cowell et al. (1999). Structure only — CPDs too large to hardcode.
func Sachs ¶
func Sachs() *models.BayesianNetwork
Sachs returns the Sachs protein signaling Bayesian network with 11 nodes. This network was learned from flow cytometry data by Sachs et al. (2005). Only the structure is provided (no CPDs) since the full parameterization is too large to hardcode.
Nodes: Raf, Mek, Plcg, PIP2, PIP3, Erk, Akt, PKA, PKC, P38, Jnk
Note: CheckModel() will fail on this network because CPDs are not set. Use it for structure-only tasks such as structure learning evaluation.
func Student ¶
func Student() *models.BayesianNetwork
Student returns the classic Student Bayesian network with 5 nodes:
D (Difficulty): 2 states {Easy, Hard}
I (Intelligence): 2 states {Low, High}
G (Grade): 3 states {A, B, C}
L (Letter): 2 states {Weak, Strong}
S (SAT): 2 states {Low, High}
Edges: D->G, I->G, I->S, G->L CPD values from Koller & Friedman (2009).
func Survey ¶
func Survey() *models.BayesianNetwork
Survey returns the Survey Bayesian network with 6 nodes. This network models the relationship between age, sex, education, occupation, residence, and travel mode. CPD values from the bnlearn repository (Scutari, 2010).
Nodes:
A (Age): {young, adult, old}
S (Sex): {M, F}
E (Education): {high, uni}
O (Occupation): {emp, self}
R (Residence): {small, big}
T (Travel): {car, train, other}
Edges: A->E, S->E, E->O, E->R, O->T, R->T
func VisitAsia ¶
func VisitAsia() *models.BayesianNetwork
VisitAsia returns the Asia network under an alternative name for compatibility. Some libraries refer to this model as "VisitAsia" rather than "Asia."
func Water ¶
func Water() *models.BayesianNetwork
Water returns the Water purification network with 32 nodes and 66 edges. From Jensen et al. (1989). Structure only — CPDs too large to hardcode.
func WaterSprinkler ¶
func WaterSprinkler() *models.BayesianNetwork
WaterSprinkler returns the classic Rain/Sprinkler/WetGrass Bayesian network.
Nodes:
Cloudy (C): {No, Yes}
Sprinkler (S): {Off, On}
Rain (R): {No, Yes}
WetGrass (W): {No, Yes}
Edges: Cloudy->Sprinkler, Cloudy->Rain, Sprinkler->WetGrass, Rain->WetGrass
func Win95pts ¶
func Win95pts() *models.BayesianNetwork
Win95pts returns the Win95pts printer troubleshooting network with 76 nodes and 112 edges. From Heckerman et al. (1995). Structure only — CPDs too large to hardcode.
Types ¶
This section is empty.
Source Files