# BNNSActivationFunction Constants that describe activation functions. ``` struct BNNSActivationFunction ``` ## Topics ### Activation Functions [`abs`](/documentation/Accelerate/BNNSActivationFunction/abs) An activation function that returns the absolute value of its input. [`clamp`](/documentation/Accelerate/BNNSActivationFunction/clamp) An activation function that returns its input clamped to a specified range. [`identity`](/documentation/Accelerate/BNNSActivationFunction/identity) An activation function that returns its input. [`integerLinearSaturate`](/documentation/Accelerate/BNNSActivationFunction/integerLinearSaturate) An activation function that returns an arithmetic shift, preserving sign. [`integerLinearSaturatePerChannel`](/documentation/Accelerate/BNNSActivationFunction/integerLinearSaturatePerChannel) An activation function that returns an arithmetic shift, preserving sign for each channel. [`leakyRectifiedLinear`](/documentation/Accelerate/BNNSActivationFunction/leakyRectifiedLinear) An activation function that returns its input when that is greater than or equal to zero, otherwise it returns its input multiplied by a specified value. [`linear`](/documentation/Accelerate/BNNSActivationFunction/linear) An activation function that returns its input multiplied by a specified value. [`rectifiedLinear`](/documentation/Accelerate/BNNSActivationFunction/rectifiedLinear) An activation function that returns its input when that is greater than or equal to zero, otherwise it returns zero. [`scaledTanh`](/documentation/Accelerate/BNNSActivationFunction/scaledTanh) An activation function that returns the scaled hyperbolic tangent of its input. [`sigmoid`](/documentation/Accelerate/BNNSActivationFunction/sigmoid) An activation function that returns the sigmoid function of its input. [`softmax`](/documentation/Accelerate/BNNSActivationFunction/softmax) An activation function that returns the softmax function of its input. [`tanh`](/documentation/Accelerate/BNNSActivationFunction/tanh) An activation function that returns the hyperbolic tangent of its input. ### Raw Values [`init(_:)`](/documentation/Accelerate/BNNSActivationFunction/init(_:)) [`init(rawValue:)`](/documentation/Accelerate/BNNSActivationFunction/init(rawValue:)) [`rawValue`](/documentation/Accelerate/BNNSActivationFunction/rawValue) [`BNNSActivationFunctionAbs`](/documentation/Accelerate/BNNSActivationFunctionAbs) [`BNNSActivationFunctionCELU`](/documentation/Accelerate/BNNSActivationFunctionCELU) An activation function that evaluates the continuously differentiable exponential linear units (CELU) on its input. [`BNNSActivationFunctionClamp`](/documentation/Accelerate/BNNSActivationFunction/BNNSActivationFunctionClamp) An activation function that returns its input clamped to the specified range. [`BNNSActivationFunctionClampedLeakyRectifiedLinear`](/documentation/Accelerate/BNNSActivationFunctionClampedLeakyRectifiedLinear) An activation function that returns its input clamped to beta when that is greater than or equal to zero, otherwise it returns its input multiplied by alpha clamped to beta. [`BNNSActivationFunctionELU`](/documentation/Accelerate/BNNSActivationFunctionELU) An activation function that evaluates the exponential linear units (ELU) on its input. [`BNNSActivationFunctionErf`](/documentation/Accelerate/BNNSActivationFunctionErf) [`BNNSActivationFunctionGELU`](/documentation/Accelerate/BNNSActivationFunctionGELU) [`BNNSActivationFunctionGELUApproximation`](/documentation/Accelerate/BNNSActivationFunctionGELUApproximation) An activation function that evaluates the Gaussian error linear units (GELU) approximation on its input. [`BNNSActivationFunctionGELUApproximation2`](/documentation/Accelerate/BNNSActivationFunctionGELUApproximation2) An activation function that provides a fast evaluation of the Gaussian error linear units (GELU) approximation on its input. [`BNNSActivationFunctionGELUApproximationSigmoid`](/documentation/Accelerate/BNNSActivationFunctionGELUApproximationSigmoid) [`BNNSActivationFunctionGumbel`](/documentation/Accelerate/BNNSActivationFunctionGumbel) An activation function that returns random numbers from the Gumbel distribution. [`BNNSActivationFunctionGumbelMax`](/documentation/Accelerate/BNNSActivationFunctionGumbelMax) An activation function that returns random numbers from the Gumbel distribution. [`BNNSActivationFunctionHardShrink`](/documentation/Accelerate/BNNSActivationFunctionHardShrink) An activation function that returns zero when the absolute input is less than alpha, otherwise it returns its input. [`BNNSActivationFunctionHardSigmoid`](/documentation/Accelerate/BNNSActivationFunctionHardSigmoid) An activation function that returns the hard sigmoid function of its input. [`BNNSActivationFunctionHardSwish`](/documentation/Accelerate/BNNSActivationFunctionHardSwish) An activation function that returns the hard swish function of its input. [`BNNSActivationFunctionIdentity`](/documentation/Accelerate/BNNSActivationFunctionIdentity) [`BNNSActivationFunctionIntegerLinearSaturate`](/documentation/Accelerate/BNNSActivationFunction/BNNSActivationFunctionIntegerLinearSaturate) An activation function that returns an arithmetic shift, preserving sign. [`BNNSActivationFunctionIntegerLinearSaturatePerChannel`](/documentation/Accelerate/BNNSActivationFunction/BNNSActivationFunctionIntegerLinearSaturatePerChannel) An activation function that returns an arithmetic shift, preserving sign for each channel. [`BNNSActivationFunctionLeakyRectifiedLinear`](/documentation/Accelerate/BNNSActivationFunctionLeakyRectifiedLinear) [`BNNSActivationFunctionLinear`](/documentation/Accelerate/BNNSActivationFunction/BNNSActivationFunctionLinear) An activation function that returns its input multiplied by a specified value. [`BNNSActivationFunctionLinearWithBias`](/documentation/Accelerate/BNNSActivationFunctionLinearWithBias) An activation function that returns its input multiplied by a scale and added to a bias. [`BNNSActivationFunctionLogSigmoid`](/documentation/Accelerate/BNNSActivationFunctionLogSigmoid) An activation function that returns the logarithm of the sigmoid function of its input. [`BNNSActivationFunctionLogSoftmax`](/documentation/Accelerate/BNNSActivationFunctionLogSoftmax) An activation function that returns the logarithm of the softmax function of its input. [`BNNSActivationFunctionPReLUPerChannel`](/documentation/Accelerate/BNNSActivationFunctionPReLUPerChannel) An activation function provides per-channel alpha values to Leaky Rectified Linear. [`BNNSActivationFunctionRectifiedLinear`](/documentation/Accelerate/BNNSActivationFunctionRectifiedLinear) [`BNNSActivationFunctionReLU6`](/documentation/Accelerate/BNNSActivationFunctionReLU6) [`BNNSActivationFunctionScaledTanh`](/documentation/Accelerate/BNNSActivationFunctionScaledTanh) [`BNNSActivationFunctionSELU`](/documentation/Accelerate/BNNSActivationFunctionSELU) An activation function that evaluates the scaled exponential linear units (SELU) on its input. [`BNNSActivationFunctionSigmoid`](/documentation/Accelerate/BNNSActivationFunctionSigmoid) [`BNNSActivationFunctionSiLU`](/documentation/Accelerate/BNNSActivationFunctionSiLU) An activation function that returns the sigmoid linear unit (SiLU) function of its input. [`BNNSActivationFunctionSoftmax`](/documentation/Accelerate/BNNSActivationFunction/BNNSActivationFunctionSoftmax) An activation function that returns the softmax function of its input. [`BNNSActivationFunctionSoftplus`](/documentation/Accelerate/BNNSActivationFunctionSoftplus) An activation function that returns the softplus function of its input. [`BNNSActivationFunctionSoftShrink`](/documentation/Accelerate/BNNSActivationFunctionSoftShrink) An activation function that returns zero when the absolute input is less than alpha, otherwise it returns its input minus alpha. [`BNNSActivationFunctionSoftsign`](/documentation/Accelerate/BNNSActivationFunctionSoftsign) An activation function that returns the softsign function of its input. [`BNNSActivationFunctionTanh`](/documentation/Accelerate/BNNSActivationFunctionTanh) [`BNNSActivationFunctionTanhShrink`](/documentation/Accelerate/BNNSActivationFunctionTanhShrink) An activation function that returns its input minus the hyperbolic tangent of its input. [`BNNSActivationFunctionThreshold`](/documentation/Accelerate/BNNSActivationFunctionThreshold) An activation function that returns beta if its input is less than a specified threshold, otherwise it returns its input. --- Copyright © 2026 Apple Inc. All rights reserved. | [Terms of Use](https://www.apple.com/legal/internet-services/terms/site.html) | [Privacy Policy](https://www.apple.com/privacy/privacy-policy)