This network learns to map a unit vector to a number corresponding to the position of the '1' in the unit vector.

KerasUnitLearner.py

from keras.layers.core import Dense
from keras.models import Sequential
from numpy import array
import numpy as np
import os
import argparse
from scipy import signal
 
N = 10

def create_model():
    model = Sequential()
    model.add(Dense(1, input_shape=(1, N)))
    model.add(Dense(10))
    model.add(Dense(1))
    model.compile(optimizer='adam', loss='mean_squared_error', metrics=['acc'])
    return model

def create_data(n):
    X = list()
    y = list()
    for i in range(n):
        X.append(array([array(signal.unit_impulse(N, i%N))]))
        y.append(array([array([i%N])]))
    return array(X), array(y)

def create_test():
    a = array([0, 0, 4, 5, 5, 6, 8, 4, 1, 4, 9, 8, 1, 1, 7, 9, 9, 3, 6, 7])
    X = list()
    y = list()
    for i in a:
        X.append(array([array(signal.unit_impulse(N, i%N))]))
        y.append(array([array([i % N])]))
    return array(X), array(y)

model = create_model()

X,y = create_data(1000)
X_test, y_test = create_test()
epochs = 200
model.fit(X, y, epochs=epochs, verbose=0)
e = model.evaluate(X_test, y_test, verbose=0)
print("%s: %.0f%%" % (model.metrics_names[1],  e[1]*100))