simple_svm.py

#!/usr/bin/env python3.6
import numpy as np
import tensorflow as tf

# Mash-up of:
# - https://stackoverflow.com/a/49497004
# - https://ml-with-tensorflow.info/2017/03/01/svm-with-tensorflow/

X = np.zeros([157, 128])
Y = np.zeros([157], dtype=np.int32)
example_id = np.array(['%d' % i for i in range(len(Y))])

x_column_name = 'x'
example_id_column_name = 'example_id'

train_input_fn = tf.estimator.inputs.numpy_input_fn(
    x={x_column_name: X, example_id_column_name: example_id},
    y=Y,
    num_epochs=None,
    shuffle=True)

svm = tf.contrib.learn.SVM(
    example_id_column=example_id_column_name,
    feature_columns=(tf.contrib.layers.real_valued_column(
        column_name=x_column_name, dimension=128),),
    l2_regularization=0.1)

svm.fit(input_fn=train_input_fn, steps=10)

metrics = svm.evaluate(input_fn=train_input_fn, steps=1)
print("Loss", metrics['loss'], "\nAccuracy", metrics['accuracy'])

min_x = np.amin(X)
max_x = np.amax(X)
min_y = np.amin(Y)
max_y = np.amax(Y)

x_predict = np.random.uniform(min_x, max_x, (157, 128, 1))
y_predict = np.random.uniform(min_y, max_y, (157, 1))

def predict_fn():
  return {
    'x': tf.constant(x_predict),
    'y': tf.constant(y_predict),
  }

pred = list(svm.predict(input_fn=predict_fn))
predicted_class = map(lambda x: x['classes'], pred)
annotations = map(lambda x: '%.2f' % x['logits'][0], pred)
for a in annotations:
    print(a)