refactor

2019-09-07 14:45:59 +02:00 · 2019-09-07 14:45:59 +02:00 · 2abe1dc891
commit 2abe1dc891
parent b0306e3231
5 changed files with 0 additions and 46 deletions
--- a/neural_networks/init.py
+++ b/neural_networks/init.py
--- a/neural_networks/functions.py
+++ b/neural_networks/functions.py
--- a/neural_networks/simple_mlp.py
+++ b/neural_networks/simple_mlp.py
--- a/neural_networks/vanilla_mlp.py
+++ b/neural_networks/vanilla_mlp.py
--- a/vanilla_logistic_regression.py
+++ b/vanilla_logistic_regression.py
@ -1,46 +0,0 @@
-import numpy as np
-
-"""
-# Logistic Regression
-
-w^T · x = w0 + w1x1 + w2x2 + w3x3
-
-P(A|x) = σ(w^Tx)
-
-# Given data:
-
-D = ((x1, y1), ..., (xn, yn))
-x ∈ R
-y ∈ {0, 1}
-
-
-# Pro's:
-
- interpretable
- small # params (d + 1)
- computationally efficient
- extensible to multi-class
-
-# Cons
-
- performance as good as other models
-
-# Maximum Likelihood Estimation MLE
-
-labels y
-input  x
-
-likelihood lw = P(y[i]| x[i], w) * P(y[j] | x[j], w) ... P(y[n] | x[n],w)
-
-in product notation
-
-lw = 
-"""
-
-
-def sigmoid(z):
-    """
-    :param z: w^Tx (scalar)
-    :return: σ(w^Tx) (scalar)
-    """
-    return 1 / (1 + np.exp(-z))