Tile coding is a coarse coding function approximation method that uses several overlapping offset grids (tilings) to approximate a continuous space.
- numpy
- matplotlib (to run the example)
from tilecoding import tilecoder
# grid dimensions and tilings
dims = [8, 10, 6, 10]
tilings = 10
# value limits of each dimension (min, max)
lims = [(3.0, 7.5), (-4.4, 4.2), (9.6, 12.7), (0.0, 1.0)]
# create tilecoder with step size 0.1
T = tilecoder(dims, lims, tilings, 0.1)
# training iteration with value 5.5 at location (3.3, -2.1, 11.1, 0.7)
T[3.3, -2.1, 11.1, 0.7] = 5.5
# get approximated value at (3.3, -2.1, 11.1, 0.7)
print T[3.3, -2.1, 11.1, 0.7]
8x8 tile coder with 8 tilings approximating f(x, y) = sin(x) + cos(y) + N(0, 0.1)
- Add function
void save_weights(file_path) - Add function
void load_weights(file_path) - Add function
weights get_weights() - Add function
void set_weights(weights) - Add function
np.ndarray predict(input_data) - Add function
loss train_on_batch(Xs, ys) - Predict multiple values.