rt-activation 1.0.0

Rectified Tangent Activation (RTA) function for Keras/TensorFlow

RT Activation

A custom activation function for Keras/TensorFlow implementing Rectified Tangent Activation (RTA).

Formula

f(x) = max(x, tanh(x))

Installation

pip install rt-activation

Usage

Simple Usage (String-based)

import keras
from keras import layers
import rt_activation  # This registers the activation function

model = keras.Sequential([
    keras.Input(shape=input_shape),
    layers.Conv2D(32, kernel_size=(3, 3), activation="RTA"),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Conv2D(64, kernel_size=(3, 3), activation="RTA"),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Flatten(),
    layers.Dropout(0.5),
    layers.Dense(num_classes, activation="softmax"),
])

Function-based Usage

import keras
from keras import layers
from rt_activation import RTA

model = keras.Sequential([
    keras.Input(shape=input_shape),
    layers.Conv2D(32, kernel_size=(3, 3), activation=RTA),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Conv2D(64, kernel_size=(3, 3), activation=RTA),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Flatten(),
    layers.Dropout(0.5),
    layers.Dense(num_classes, activation="softmax"),
])

Properties

• Smooth: Differentiable everywhere
• Non-saturating: Linear growth for large positive values
• Bounded for negatives: tanh behavior for negative inputs
• Zero-centered: Output can be negative

License

MIT License