tswedish

Keybase proof

I hereby claim:

• I am tswedish on github.
• I am tswedish (https://keybase.io/tswedish) on keybase.
• I have a public key ASBmF9szOq6XhfbyhBf4NY1dTFdefEIf91vI5vstfqE7Ugo

To claim this, I am signing this object:

tswedish

def super_complicated_function(x):
	return x * sin(x+6.)**2. / 2. - 2. / 2.**x

tswedish

def fn(x):
    return x**2 + 0.2*(x-2)**5 + 2*x**3

# initialization
x = Variable(3.)
target = 42.
print('---- Initial Value ----')
print('fn(x): {}'.format(fn(x)))
print('Target: {}'.format(target))
print('intial guess for x: {}'.format(x))

tswedish

for x in [-1., 0.5, 3.]:
    print('Function Input: {}'.format(x))
    print('Function Value: {}'.format(super_complicated_function(x)))
    print('Function Symbolic Derivative: {}'.format(d_super_complicated_function(x)))
    x_v = Variable(x)
    L = super_complicated_function(x_v)
    print('Variable Function Output Value: {}'.format(L))
    L.backward()
    print('Input value: {}'.format(x_v))
    print('-'*32)

tswedish

def __add__(self, other):
        if isinstance(other, self.__class__):
            return Variable(lambda a,b : a+b, (self, other))
        else:
            return Variable(lambda a,b : a+b, (self, Variable(other)))

tswedish

def backward(self, output_gradient=1.):
        # combine gradients from other paths and propagate to children
        self.gradient += output_gradient
        local_gradient = self.derivative_op(*self.calc_input_values())
        for differential, input_variable in zip(local_gradient, self.input_variables):
            input_variable.backward(differential * output_gradient)

tswedish

    def calc_input_values(self):
        # calculate the real-valued input to operation
        return [v.value for v in self.input_variables]

    def forward(self):  
        # calculate the real-valued output of operation              
        return self.forward_op(*self.calc_input_values())

    @property
    def value(self):

tswedish

class Variable:
    def __init__(self, operation, input_variables=[]):
        # note the setter for @property value below
        self.value = operation
        self.input_variables = input_variables
        self.gradient = 0.

tswedish

# compose a graph of nodes

x = Variable(10.01)
y = Variable(0.05)

z = x * y
m = 1.3
q = m+y

L = (q - (z**(m/2.) + z**2. - 1./z))**2

tswedish

def forward_fn(x):
    for n in range(5):
        if n % 2 == 0:
            x = 3.*x
        else:
            x = x**(1./n) + 1./n

    return x

x = Variable(2.)