Python library for dot product classification-CodePudding

I have the following python pyseudo-code:

A1 = "101000001111"
A2 = "110000010101"
B2 = "000111010000"
B2 = "000110100000"
# TODO get X = [x1, x2, ..., x12]
assert(A1 * X > .5)
assert(A2 * X > .5)
assert(B1 * X < .5)
assert(B2 * X < .5)

So this will basically be a regression based classification.

0.5 is my threshold but how to get X?

CodePudding user response：

You need to find 12 coefficients. You can try to use LogisticRegression or LinearRegression
When you have linear coefficients you can use np.dot or @ operator to get a dot product.

Example:

import numpy as np
from sklearn.linear_model import LogisticRegression

A1 = "101000001111"
A2 = "110000010101"
B1 = "000111010000"
B2 = "000110100000"

A1 = np.array(list(A1), np.float32)
A2 = np.array(list(A2), np.float32)
B1 = np.array(list(B1), np.float32)
B2 = np.array(list(B2), np.float32)

X = np.array((A1, A2, B1, B2))
y = np.array([1, 1, 0, 0])
w = model = LogisticRegression().fit(X, y).coef_.flatten()

print(A1.dot(w))
print(A2.dot(w))
print(B1.dot(w))
print(B2.dot(w))

assert A1 @ w > 0.5
assert A2 @ w > 0.5
assert B1 @ w < 0.5
assert B2 @ w < 0.5

Results:

1.7993630995882384
1.5032155788245702
-1.0190643734998346
-1.0385501901808816