import numpy as np
import matplotlib.pyplot as plt

number_of_noise_steps = 20
noise_scale = np.arange(0, number_of_noise_steps + 1) / float(
    number_of_noise_steps
)

# 1x

data = np.load("avg_pooling_mlp/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    label="MLP"
)

data = np.load("basis_mlp/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    label="MLP Basis"
)


data = np.load("avg_pooling_nnmf/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    label="NNMF"
)

data = np.load("avg_pooling_nnmf_sp1.01/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    label="NNMF Sparse 1.01"
)

data = np.load("basis_nnmf/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    label="NNMF Basis"
)

# 2x

data = np.load("avg_pooling_mlp_x2/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    ":",
    label="MLP x2"
)

data = np.load("basis_mlp_x2/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    ":",
    label="MLP Basis x2"
)


data = np.load("avg_pooling_nnmf_x2/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    ":",
    label="NNMF x2"
)

data = np.load("avg_pooling_nnmf_sp1.01_x2/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    ":",
    label="NNMF Sparse 1.01 x2"
)

data = np.load("basis_nnmf_x2/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    ":",
    label="NNMF Basis x2"
)

# 4x

data = np.load("avg_pooling_mlp_x4/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    "--",
    label="MLP x4"
)

data = np.load("basis_mlp_x4/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    "--",
    label="MLP Basis x4"
)


data = np.load("avg_pooling_nnmf_x4/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    "--",
    label="NNMF x4"
)

data = np.load("avg_pooling_nnmf_sp1.01_x4/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    "--",
    label="NNMF Sparse 1.01 x4"
)

data = np.load("basis_nnmf_x4/noise_holes_w_noise_results.npy")
plt.plot(
    noise_scale,
    data,
    "--",
    label="NNMF Basis x4"
)

plt.legend()
plt.xlabel("eta")
plt.ylabel("Performance [%]")
plt.title("CIFAR10 Random Holes filled with rand")
plt.show()