ultimate 2.62.2rc1

ultimate

ultimate

A scikit-learn compatible neural network implementation

Installation

pip install ultimate

Why Ultimate?

• Support feature importance
• Support missing values
• Support am2x/a2m2x activation functions
• Support softmax/hardmax/mse/hardmse loss functions
• Support fc/add/conv/star operations

How To Use?

# let's use a simple example to learn how to use
from ultimate.mlp import MLP
import numpy as np

# generate sample
X = np.linspace(-np.pi, np.pi, num=5000).reshape(-1, 1)
Y = np.sin(X)
print(X.shape, Y.shape)

# fit and predict
mlp = MLP(layer_size=[X.shape[1], 8, 8, 8, 1], rate_init=0.02, loss_type="mse", epoch_train=100, epoch_decay=10, verbose=1)

mlp.fit(X, Y)
pred = mlp.predict(X)

# show the result
import matplotlib.pyplot as plt  
plt.plot(X, pred)
plt.show()

Examples

• Feature Importance
• Image Regression
• Iris Classification
• MNIST Recognition