How to convert Scikit Learn logistic regression model to TensorFlow

Question

I would like to use existing Scikit Learn LogisticRegression model in the BigQuery ML. However, BQ ML currently has a hard limit of 50 unique labels and my model needs to handle more than that.

BQ accepts TensorFlow models, which do not seem to have this limit.

How can I convert existing Scikit logistic regression model to TensorFlow model?

Answer 1

Sure, here is a skeleton in TF 2.0.

import tensorflow as tf

weights = tf.Variable(tf.random.normal(shape=(784, 10), dtype=tf.float64))
biases  = tf.Variable(tf.random.normal(shape=(10,), dtype=tf.float64))

def logistic_regression(x):
    lr = tf.add(tf.matmul(x, weights), biases)
    return lr

def cross_entropy(y_true, y_pred):
    y_true = tf.one_hot(y_true, 10)
    loss = tf.nn.softmax_cross_entropy_with_logits(labels=y_true, logits=y_pred)
    return tf.reduce_mean(loss)

def grad(x, y):
    with tf.GradientTape() as tape:
        y_pred = logistic_regression(x)
        loss_val = cross_entropy(y, y_pred)
    return tape.gradient(loss_val, [weights, biases])