braincog.base.strategy package
Submodules
braincog.base.strategy.LateralInhibition module
- class braincog.base.strategy.LateralInhibition.LateralInhibition(node, inh, mode='constant')
基类:
Module
侧抑制 用于发放脉冲的神经元抑制其他同层神经元 在膜电位上作用
- forward(x: Tensor, xori=None)
Defines the computation performed at every call.
Should be overridden by all subclasses.
备注
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- training: bool
braincog.base.strategy.surrogate module
- class braincog.base.strategy.surrogate.AtanGrad(alpha=2.0, requires_grad=True)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.BackEIGateGrad(alpha=2.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.EIGrad(alpha=2.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.GateGrad(alpha=2.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.QGateGrad(alpha=2.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.ReLUGrad(alpha=2.0, requires_grad=False)
-
使用ReLU作为代替梯度函数, 主要用为相同结构的ANN的测试
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.STDPGrad(alpha=2.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.SigmoidGrad(alpha=1.0, requires_grad=False)
-
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- training: bool
- class braincog.base.strategy.surrogate.SurrogateFunctionBase(alpha, requires_grad=True)
基类:
Module
Surrogate Function 的基类 :param alpha: 为一些能够调控函数形状的代理函数提供参数. :param requires_grad: 参数
alpha
是否需要计算梯度, 默认为False
- static act_fun(x, alpha)
- 参数
x – 膜电位的输入
alpha – 控制代理梯度形状的变量, 可以为
NoneType
- 返回
激发之后的spike, 取值为
[0, 1]
- forward(x)
- 参数
x – 膜电位输入
- 返回
激发之后的spike
- training: bool
- class braincog.base.strategy.surrogate.atan(*args, **kwargs)
基类:
Function
使用 Atan 作为代理梯度函数 对应的原函数为:
\[g(x) = \frac{1}{\pi} \arctan(\frac{\pi}{2}\alpha x) + \frac{1}{2}\]反向传播的函数为:
\[g'(x) = \frac{\alpha}{2(1 + (\frac{\pi}{2}\alpha x)^2)}\]- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, inputs, alpha)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.backeigate(*args, **kwargs)
基类:
Function
- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, input)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.ei(*args, **kwargs)
基类:
Function
- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, input)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.gate(*args, **kwargs)
基类:
Function
使用 gate 作为代理梯度函数 对应的原函数为:
\[g(x) = \mathrm{NonzeroSign}(x) \log (|\alpha x| + 1)\]反向传播的函数为:
\[g'(x) = \frac{\alpha}{1 + |\alpha x|} = \frac{1}{\frac{1}{\alpha} + |x|}\]- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, x, alpha)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- braincog.base.strategy.surrogate.heaviside(x)
- class braincog.base.strategy.surrogate.quadratic_gate(*args, **kwargs)
基类:
Function
使用 quadratic_gate 作为代理梯度函数 对应的原函数为:
\[\begin{split}g(x) = \begin{cases} 0, & x < -\frac{1}{\alpha} \\ -\frac{1}{2}\alpha^2|x|x + \alpha x + \frac{1}{2}, & |x| \leq \frac{1}{\alpha} \\ 1, & x > \frac{1}{\alpha} \\ \end{cases}\end{split}\]反向传播的函数为:
\[\begin{split}g'(x) = \begin{cases} 0, & |x| > \frac{1}{\alpha} \\ -\alpha^2|x|+\alpha, & |x| \leq \frac{1}{\alpha} \end{cases}\end{split}\]- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, x, alpha)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.relu_like(*args, **kwargs)
基类:
Function
- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, x, alpha)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.sigmoid(*args, **kwargs)
基类:
Function
使用 sigmoid 作为代理梯度函数 对应的原函数为:
\[g(x) = \mathrm{sigmoid}(\alpha x) = \frac{1}{1+e^{-\alpha x}}\]反向传播的函数为:
\[g'(x) = \alpha * (1 - \mathrm{sigmoid} (\alpha x)) \mathrm{sigmoid} (\alpha x)\]- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, x, alpha)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.stdp(*args, **kwargs)
基类:
Function
- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, inputs)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.
- class braincog.base.strategy.surrogate.straight_through_estimator(*args, **kwargs)
基类:
Function
使用直通估计器作为代理梯度函数 http://arxiv.org/abs/1308.3432
- static backward(ctx, grad_output)
Defines a formula for differentiating the operation with backward mode automatic differentiation (alias to the vjp function).
This function is to be overridden by all subclasses.
It must accept a context
ctx
as the first argument, followed by as many outputs as theforward()
returned (None will be passed in for non tensor outputs of the forward function), and it should return as many tensors, as there were inputs toforward()
. Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input. If an input is not a Tensor or is a Tensor not requiring grads, you can just pass None as a gradient for that input.The context can be used to retrieve tensors saved during the forward pass. It also has an attribute
ctx.needs_input_grad
as a tuple of booleans representing whether each input needs gradient. E.g.,backward()
will havectx.needs_input_grad[0] = True
if the first input toforward()
needs gradient computated w.r.t. the output.
- static forward(ctx, inputs)
Performs the operation.
This function is to be overridden by all subclasses.
It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).
The context can be used to store arbitrary data that can be then retrieved during the backward pass. Tensors should not be stored directly on ctx (though this is not currently enforced for backward compatibility). Instead, tensors should be saved either with
ctx.save_for_backward()
if they are intended to be used inbackward
(equivalently,vjp
) orctx.save_for_forward()
if they are intended to be used for injvp
.