func q3() {
fns := []linreg.TransformFunc{phi0, phi1, phi2, phi3, phi4, phi5, phi6, phi7}
ks := []int{3, 4, 5, 6, 7}
data := getData("data/in.dta")
for _, k := range ks {
linreg := linreg.NewLinearRegression()
linreg.InitializeFromData(data[25:])
linreg.InitializeValidationFromData(data[:25])
linreg.TransformFunction = fns[k]
linreg.ApplyTransformation()
linreg.ApplyTransformationOnValidation()
linreg.Learn()
eIn := linreg.Ein()
eVal := linreg.EValIn()
eOut, _ := linreg.EoutFromFile("data/out.dta")
fmt.Printf("EVal = %f, for k = %d\n", eVal, k)
fmt.Printf("EIn = %f, for k = %d\n", eIn, k)
fmt.Printf("EOut = %f, for k = %d\n", eOut, k)
fmt.Println()
}
}
func q5() {
runs := 1000 // number of times we repeat the experiment
linreg := linreg.NewLinearRegression()
linreg.N = 100
var avgEin, avgEout float64
for run := 0; run < runs; run++ {
linreg.Initialize()
linreg.Learn()
avgEin += linreg.Ein()
avgEout += linreg.Eout()
}
avgEin = avgEin / float64(runs)
avgEout = avgEout / float64(runs)
fmt.Printf("average of In sample error 'Ein' for Linear regresion for N = 100 is %4.2f\n", avgEin)
fmt.Printf("average of Out of sample error 'Eout' for Linear regresion for N = 100 is %4.2f\n", avgEout)
}
func q7() {
runs := 1000 // number of times we repeat the experiment
linreg := linreg.NewLinearRegression()
linreg.N = 10
iterations := 0
for run := 0; run < runs; run++ {
linreg.Initialize()
linreg.Learn()
pla := pla.NewPLA()
pla.Initialize()
for i := 0; i < len(pla.Wn); i++ {
pla.Wn[i] = linreg.Wn[i]
}
iterations += pla.Converge()
}
avgIterations := float64(iterations) / float64(runs)
fmt.Printf("average for PLA to converge for N = %v with initial weight computed by linear regression is %4.2f\n", linreg.N, avgIterations)
}
func q2() {
linreg := linreg.NewLinearRegression()
linreg.InitializeFromFile("data/in.dta")
linreg.TransformFunction = nonLinearFeature
linreg.ApplyTransformation()
linreg.Learn()
ein := linreg.Ein()
eout, _ := linreg.EoutFromFile("data/out.dta")
fmt.Printf("Ein = %f, Eout = %f\n", ein, eout)
linreg.K = -3
linreg.LearnWeightDecay()
eAugIn := linreg.EAugIn()
eAugOut, _ := linreg.EAugOutFromFile("data/out.dta")
fmt.Printf("Ein = %f, Eout = %f for k = %d\n", eAugIn, eAugOut, linreg.K)
linreg.K = 3
linreg.LearnWeightDecay()
eAugIn = linreg.EAugIn()
eAugOut, _ = linreg.EAugOutFromFile("data/out.dta")
fmt.Printf("Ein = %f, Eout = %f for k = %d\n", eAugIn, eAugOut, linreg.K)
k := []int{2, 1, 0, -1, -2}
for _, ki := range k {
linreg.K = ki
linreg.LearnWeightDecay()
eAugIn = linreg.EAugIn()
eAugOut, _ = linreg.EAugOutFromFile("data/out.dta")
fmt.Printf("Ein = %f, Eout = %f for k = %d\n", eAugIn, eAugOut, ki)
}
minK := 0
minEAug := float64(1000)
for i := -10; i < 10; i++ {
linreg.K = i
linreg.LearnWeightDecay()
eAugOut, _ = linreg.EAugOutFromFile("data/out.dta")
if minEAug > eAugOut {
minK = i
minEAug = eAugOut
}
}
fmt.Printf("mininum Eout = %f with k = %d for k in {-10; 10}\n", minEAug, minK)
}
func q8() {
runs := 1000 // number of times we repeat the experiment
linreg := linreg.NewLinearRegression()
linreg.N = 1000
linreg.RandomTargetFunction = false
linreg.TwoParams = true
linreg.Noise = 0.1
var avgEin float64
for run := 0; run < runs; run++ {
linreg.TargetFunction = f //non linear function
linreg.Initialize()
linreg.Learn()
avgEin += linreg.Ein()
}
avgEin = avgEin / float64(runs)
fmt.Printf("average of In sample error 'Ein' for Linear regresion for N = 100 is %4.2f\n", avgEin)
}
func q9() {
runs := 1000 // number of times we repeat the experiment
linreg := linreg.NewLinearRegression()
linreg.N = 1000
linreg.RandomTargetFunction = false
linreg.TwoParams = true
linreg.Noise = 0.1
var aAvgIn, bAvgIn, cAvgIn, dAvgIn, eAvgIn float64
var aAvgOut, bAvgOut, cAvgOut, dAvgOut, eAvgOut float64
var AvgEout float64
for run := 0; run < runs; run++ {
linreg.TargetFunction = f //non linear function
linreg.Initialize()
linreg.TransformDataSet(nonLinearFeature, 6)
linreg.Learn()
aAvgIn += linreg.CompareInSample(a, 2)
bAvgIn += linreg.CompareInSample(b, 2)
cAvgIn += linreg.CompareInSample(c, 2)
dAvgIn += linreg.CompareInSample(d, 2)
eAvgIn += linreg.CompareInSample(e, 2)
aAvgOut += linreg.CompareOutOfSample(a, 2)
bAvgOut += linreg.CompareOutOfSample(b, 2)
cAvgOut += linreg.CompareOutOfSample(c, 2)
dAvgOut += linreg.CompareOutOfSample(d, 2)
eAvgOut += linreg.CompareOutOfSample(e, 2)
AvgEout += linreg.Eout()
}
aAvgIn = aAvgIn / float64(runs)
bAvgIn = bAvgIn / float64(runs)
cAvgIn = cAvgIn / float64(runs)
dAvgIn = dAvgIn / float64(runs)
eAvgIn = eAvgIn / float64(runs)
aAvgOut = aAvgOut / float64(runs)
bAvgOut = bAvgOut / float64(runs)
cAvgOut = cAvgOut / float64(runs)
dAvgOut = dAvgOut / float64(runs)
eAvgOut = eAvgOut / float64(runs)
AvgEout = AvgEout / float64(runs)
fmt.Printf("average of difference in sample between a() and the linear regression hypothesis is %5.3f\n", aAvgIn)
fmt.Printf("average of difference in sample between b() and the linear regression hypothesis is %5.3f\n", bAvgIn)
fmt.Printf("average of difference in sample between c() and the linear regression hypothesis is %5.3f\n", cAvgIn)
fmt.Printf("average of difference in sample between d() and the linear regression hypothesis is %5.3f\n", dAvgIn)
fmt.Printf("average of difference in sample between e() and the linear regression hypothesis is %5.3f\n", eAvgIn)
fmt.Printf("average of difference out of sample between a() and the linear regression hypothesis is %7.5f\n", aAvgOut)
fmt.Printf("average of difference out of sample between b() and the linear regression hypothesis is %7.5f\n", bAvgOut)
fmt.Printf("average of difference out of sampleb etween c() and the linear regression hypothesis is %7.5f\n", cAvgOut)
fmt.Printf("average of difference out of sample between d() and the linear regression hypothesis is %7.5f\n", dAvgOut)
fmt.Printf("average of difference out of sample between e() and the linear regression hypothesis is %7.5f\n", eAvgOut)
fmt.Printf("average Eout of the hypothesis learn through linear regression with noise of %4.2f is %4.3f\n", linreg.Noise, AvgEout)
}