math.operator

class tinychain.math.operator.Abs(subject)

Bases: Unary

backward(_variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Acos(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Acosh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Add(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Asin(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Asinh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Atan(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Atanh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Cond(subject, args)

Bases: Operator

A boolean condition

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Cos(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Cosh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Custom(subject)

Bases: Unary

A custom operator

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Div(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Dual(subject, args)

Bases: Operator

A differentiable operator with two arguments

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.DualBroadcast(subject, args)

Bases: Operator

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Exp(subject)

Bases: Unary

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Gradients

Bases: dict

clear() → None.  Remove all items from D.

copy() → a shallow copy of D

fromkeys(value=None, /)

Create a new dictionary with keys from iterable and values set to value.

get(key, default=None, /)

Return the value for key if key is in the dictionary, else default.

items() → a set-like object providing a view on D's items

keys() → a set-like object providing a view on D's keys

pop(k[, d]) → v, remove specified key and return the corresponding value.

If the key is not found, return the default if given; otherwise, raise a KeyError.

popitem()

Remove and return a (key, value) pair as a 2-tuple.

Pairs are returned in LIFO (last-in, first-out) order. Raises KeyError if the dict is empty.

setdefault(key, default=None, /)

Insert key with a value of default if key is not in the dictionary.

Return the value for key if key is in the dictionary, else default.

update([E, ]**F) → None.  Update D from dict/iterable E and F.

If E is present and has a .keys() method, then does: for k in E: D[k] = E[k] If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v In either case, this is followed by: for k in F: D[k] = F[k]

values() → an object providing a view on D's values

class tinychain.math.operator.Log(subject, args)

Bases: Operator

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.LogicalAnd(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.LogicalNot(subject)

Bases: Unary

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.LogicalOr(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.LogicalXor(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.MatMul(subject, args)

Bases: Dual

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Mul(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Operator(subject, args)

Bases: Op

A differentiable operator like addition, multiplication, exponentiation, etc.

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Pow(subject, args)

Bases: Dual

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Sin(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Sinh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Sub(subject, args)

Bases: DualBroadcast

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Tan(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Tanh(subject)

Bases: Trig

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Trig(subject)

Bases: Unary

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

class tinychain.math.operator.Unary(subject)

Bases: Operator

backward(variable=None)

Return the derivative of this Operator (may be a numeric constant or itself an Operator).

If a variable is specified, this will be the partial derivative w/r/t the given variable.

forward()

Return the result of evaluating this Operator

gradients(loss)

Return the Gradients this Operator with respect to the given variables.

If no variables are specified, this will return the Gradients of each Variable that this Operator depends on.

property shape

The shape of the result of this operation

simplify()

Return a simplified but logically equivalent version of this Operator, if possible. For example, Mul(2, 1).simplify() will return 2.

IMPORTANT: don’t call simplify until after constructing an entire operator graph. This is because simplify may discard parts of the operator graph needed to apply the chain rule correctly.

tinychain.math.operator.chain_rule(op, *args)

Compute the chain rule coefficient of the given Operator.

This function will return 1 if the given numeric is constant, or has only constant inputs.

tinychain.math.operator.constant(numeric)

Return the given numeric state as a constant, i.e. not the result of a differentiable Operator.

tinychain.math.operator.derivative_of(state_or_function, variable=None, keepdims=False)

Find the derivative of the given state_or_function.

If a differentiable state is given, this will construct a new op to calculate it derivative, which can be a partial derivative if a variable is specified.

If a differentiable function is given, a new callable function will be returned which computes its derivative.

If state_or_function is not differentiable, a TypeError will be raised.

tinychain.math.operator.gradients(numeric, loss, variables=None)

Return the gradient of a numeric state with respect to the given loss.

If one variable is given, one gradient will be returned, or a KeyError will be raised if not present in the graph. If a list of variables is given, a corresponding list of gradients will be returned. If no variables are given, a Gradients object whose keys are the inputs of the graph will be returned.

tinychain.math.operator.is_constant(numeric)

Return False if the given numeric state is the result of an Operator, i.e. a differentiable function.

tinychain.math.operator.is_one(numeric)

Return True if the given numeric state is a constant with value one.

tinychain.math.operator.is_zero(numeric)

Return True if the given numeric state is a constant with value zero.

tinychain.math.operator.ones_like(numeric, keepdims=True)

Construct a constant with each element equal to one, with the same shape and data type as numeric.

tinychain.math.operator.operator(state_or_ref)

Return the Operator instance which produces the given state_or_ref, if any

tinychain.math.operator.simplify(state)

Simplify the given operator graph, if possible. For example, simplify(Add(0, 2)) will return 2.

tinychain.math.operator.zeros_like(state, keepdims=True)

Construct a constant with each element equal to zero, with the same shape and data type as numeric.