initial notebook

%% Cell type:code id: tags:

``` python

```

numpy==1.22.3

tensorflow==2.5.2

matplotlib==3.5.1

keras==2.7.0

 No newline at end of file

%% Cell type:code id: tags:

``` python

import numpy as np

import tensorflow as tf

import matplotlib.pyplot as plt

from keras.models import Model

ti_seq = [*list(range(50, 400, 25)),

          *list(range(400, 1000, 10)),

          1000, 1030, 1050, 1080, 1100, 1130, 1150, 1180, 1200, 1230, 1250, 1280, 1300, 1330, 1350, 1380,

          1400, 1450, 1500, 1550, 1600, 1650, 1700,

          1800, 1900, 2000, 2100, 2200, 2300, 2500, 3000]

# Load real shuffled data

data = np.load("data/RD_shuffled.npy")

print(data.shape)

data_with_tis = np.array([list(x) + ti_seq for x in data], dtype=float)

train_in = data_with_tis[:int(len(data_with_tis) * 0.85)]

train_out = data[:int(len(data) * 0.85)]

val_in = data_with_tis[int(len(data_with_tis) * 0.85):-50]

val_out = data[int(len(data) * 0.85):-50]

test_in = data_with_tis[-50:]

test_out = data[-50:]

print(len(data_with_tis), len(train_in), len(val_in), len(test_in))

print(train_in.shape)

new_train = True

if new_train:

    model = tf.keras.models.Sequential()

    # encoder

    model.add(tf.keras.layers.Dense(units=105, activation='relu', input_dim=105, name="in"))

    # model.add(tf.keras.layers.Dense(units=64, activation='relu'))

    # model.add(tf.keras.layers.Dense(units=32, activation='relu'))

    model.add(tf.keras.layers.Dense(units=14, activation='relu', name="bn"))

    # decoder

    # model.add(tf.keras.layers.Dense(units=32, activation='relu'))

    # model.add(tf.keras.layers.Dense(units=64, activation='relu'))

    model.add(tf.keras.layers.Dense(units=105, activation='linear', name="out"))

    mse = tf.keras.losses.MeanSquaredError(reduction="auto", name="mean_squared_error")

    model.compile(optimizer='adam', loss=mse, metrics=["accuracy"])

    m_hist = model.fit(train_out, train_out, validation_data=(val_out, val_out), batch_size=128, epochs=30)

    model.save("tf_models_BN14-W3")

    plt.plot(m_hist.history['accuracy'])

    plt.plot(m_hist.history['val_accuracy'])

    plt.title('Model accuracy')

    plt.ylabel('Accuracy')

    plt.xlabel('Epoch')

    plt.legend(['Train', 'Test'], loc='upper left')

    plt.show()

    plt.plot(m_hist.history['loss'])

    plt.plot(m_hist.history['val_loss'])

    plt.title('Model loss')

    plt.ylabel('Loss')

    plt.xlabel('Epoch')

    plt.legend(['Train', 'Test'], loc='upper left')

    plt.show()

else:

    model = tf.keras.models.load_model("tf_models_BN14-W3",

                                       custom_objects=None,

                                       compile=True)

    # train on another epochs

    # m_hist = model.fit(train, train, validation_data=(val, val), batch_size=32, epochs=30)

    # tf.keras.models.save_model(model, "tf_models_BT8-85-1")

    # plt.plot(m_hist.history['accuracy'])

    # plt.plot(m_hist.history['val_accuracy'])

    # get encoder -> the 14 params from bottle neck

    enc_model = Model(inputs=model.input,

                      outputs=model.get_layer('bn').output)

    enc_output = enc_model.predict(data)

    print("Cov")

    print(np.average(enc_output, axis=0))

    cov_m = np.cov(enc_output.T)

    # print(cov_m)

    eig = np.linalg.eigvals(cov_m)

    print(eig)

pred = model.predict(test_out)

for i in range(len(test_out)):

    plt.plot(ti_seq, test_out[i], color='red', marker="o", label="data")

    plt.plot(ti_seq, pred[i], color='blue', marker="o", label="pred")

    plt.ylim(0, 2000)

    plt.legend()

    plt.show()

```