Hyperbolic tangent

Average Accuracy: 9.5% → 97.4%

Time: 9.4s

Precision: binary64

Cost: 33472

• Unsound rule application detected in e-graph. Results may not be sound.

• Unsound rule application detected in e-graph. Results may not be sound.

• Unsound rule application detected in e-graph. Results may not be sound.

• Unsound rule application detected in e-graph. Results may not be sound.

• Unsound rule application detected in e-graph. Results may not be sound.

• Unsound rule application detected in e-graph. Results may not be sound.

Math

\[\frac{e^{x} - e^{-x}}{e^{x} + e^{-x}} \]

↓

\[\frac{2 \cdot x + \left(0.3333333333333333 \cdot {x}^{3} + \left(0.0003968253968253968 \cdot {x}^{7} + 0.016666666666666666 \cdot {x}^{5}\right)\right)}{e^{x} + e^{-x}} \]

FPCore

(FPCore (x)
 :precision binary64
 (/ (- (exp x) (exp (- x))) (+ (exp x) (exp (- x)))))

↓

(FPCore (x)
 :precision binary64
 (/
  (+
   (* 2.0 x)
   (+
    (* 0.3333333333333333 (pow x 3.0))
    (+
     (* 0.0003968253968253968 (pow x 7.0))
     (* 0.016666666666666666 (pow x 5.0)))))
  (+ (exp x) (exp (- x)))))

C

double code(double x) {
	return (exp(x) - exp(-x)) / (exp(x) + exp(-x));
}

↓

double code(double x) {
	return ((2.0 * x) + ((0.3333333333333333 * pow(x, 3.0)) + ((0.0003968253968253968 * pow(x, 7.0)) + (0.016666666666666666 * pow(x, 5.0))))) / (exp(x) + exp(-x));
}

Fortran

real(8) function code(x)
    real(8), intent (in) :: x
    code = (exp(x) - exp(-x)) / (exp(x) + exp(-x))
end function

↓

real(8) function code(x)
    real(8), intent (in) :: x
    code = ((2.0d0 * x) + ((0.3333333333333333d0 * (x ** 3.0d0)) + ((0.0003968253968253968d0 * (x ** 7.0d0)) + (0.016666666666666666d0 * (x ** 5.0d0))))) / (exp(x) + exp(-x))
end function

Java

public static double code(double x) {
	return (Math.exp(x) - Math.exp(-x)) / (Math.exp(x) + Math.exp(-x));
}

↓

public static double code(double x) {
	return ((2.0 * x) + ((0.3333333333333333 * Math.pow(x, 3.0)) + ((0.0003968253968253968 * Math.pow(x, 7.0)) + (0.016666666666666666 * Math.pow(x, 5.0))))) / (Math.exp(x) + Math.exp(-x));
}

Python

def code(x):
	return (math.exp(x) - math.exp(-x)) / (math.exp(x) + math.exp(-x))

↓

def code(x):
	return ((2.0 * x) + ((0.3333333333333333 * math.pow(x, 3.0)) + ((0.0003968253968253968 * math.pow(x, 7.0)) + (0.016666666666666666 * math.pow(x, 5.0))))) / (math.exp(x) + math.exp(-x))

Julia

function code(x)
	return Float64(Float64(exp(x) - exp(Float64(-x))) / Float64(exp(x) + exp(Float64(-x))))
end

↓

function code(x)
	return Float64(Float64(Float64(2.0 * x) + Float64(Float64(0.3333333333333333 * (x ^ 3.0)) + Float64(Float64(0.0003968253968253968 * (x ^ 7.0)) + Float64(0.016666666666666666 * (x ^ 5.0))))) / Float64(exp(x) + exp(Float64(-x))))
end

MATLAB

function tmp = code(x)
	tmp = (exp(x) - exp(-x)) / (exp(x) + exp(-x));
end

↓

function tmp = code(x)
	tmp = ((2.0 * x) + ((0.3333333333333333 * (x ^ 3.0)) + ((0.0003968253968253968 * (x ^ 7.0)) + (0.016666666666666666 * (x ^ 5.0))))) / (exp(x) + exp(-x));
end

Wolfram

code[x_] := N[(N[(N[Exp[x], $MachinePrecision] - N[Exp[(-x)], $MachinePrecision]), $MachinePrecision] / N[(N[Exp[x], $MachinePrecision] + N[Exp[(-x)], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

↓

code[x_] := N[(N[(N[(2.0 * x), $MachinePrecision] + N[(N[(0.3333333333333333 * N[Power[x, 3.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.0003968253968253968 * N[Power[x, 7.0], $MachinePrecision]), $MachinePrecision] + N[(0.016666666666666666 * N[Power[x, 5.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / N[(N[Exp[x], $MachinePrecision] + N[Exp[(-x)], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 11.4%

   \[\frac{e^{x} - e^{-x}}{e^{x} + e^{-x}} \]

2. Taylor expanded in x around 0 97.5%

   \[\leadsto \frac{\color{blue}{2 \cdot x + \left(0.3333333333333333 \cdot {x}^{3} + \left(0.0003968253968253968 \cdot {x}^{7} + 0.016666666666666666 \cdot {x}^{5}\right)\right)}}{e^{x} + e^{-x}} \]

3. Final simplification 97.5%

   \[\leadsto \frac{2 \cdot x + \left(0.3333333333333333 \cdot {x}^{3} + \left(0.0003968253968253968 \cdot {x}^{7} + 0.016666666666666666 \cdot {x}^{5}\right)\right)}{e^{x} + e^{-x}} \]

Alternatives

Alternative 1
Accuracy	97.3%
Cost	7296

\[\left(x \cdot x\right) \cdot \left(x \cdot -0.3333333333333333\right) + \left(x + {x}^{5} \cdot 0.13333333333333333\right) \]

Alternative 2
Accuracy	97.2%
Cost	1088

\[\frac{2 \cdot x + x \cdot \left(0.3333333333333333 \cdot \left(x \cdot x\right)\right)}{2 + x \cdot x} \]

Alternative 3
Accuracy	96.9%
Cost	576

\[x + \left(x \cdot x\right) \cdot \left(x \cdot -0.3333333333333333\right) \]

Alternative 4
Accuracy	7.6%
Cost	196

\[\begin{array}{l} \mathbf{if}\;x \leq -5 \cdot 10^{-310}:\\ \;\;\;\;-1\\ \mathbf{else}:\\ \;\;\;\;1\\ \end{array} \]

Alternative 5
Accuracy	4.9%
Cost	64

\[-1 \]

Alternative 6
Accuracy	96.6%
Cost	64

\[x \]

Reproduce

herbie shell --seed 2023160 
(FPCore (x)
  :name "Hyperbolic tangent"
  :precision binary64
  (/ (- (exp x) (exp (- x))) (+ (exp x) (exp (- x)))))