My loss value is nan and accuracy: 0.0000e+00 on numerical data

Question

I am working on the XAI model and implementing a simple model based on my data. While training, the loss, and accuracy are Nan and I am unable to find out the problem.

[XAI] https://github.com/EthicalML/xai/blob/master/examples/XAI%20Tabular%20Data%20Example%20Usage.ipynb

Below is my data, please do not include usage columns because these are prediction columns.

    import xai.data
    import pandas as pd
    
    df = pd.read_csv('/content/heat_pipe.csv') 
    categorical_cols = ["Date", "Temperature(degree)", "Wind_speed(m/s)","Humidity(%)","day_of_the_week", "sarturday", "sunday", "public_holidays_(weekdays)","total_holidays"]

_ = xai.correlations(df, include_categorical=True, plot_type="matrix")
_ = xai.correlations(df, include_categorical=True)


proc_df = xai.normalize_numeric(df)
proc_df = xai.convert_categories(proc_df)

x = proc_df.drop("usage_(total)", axis=1)
y = proc_df["usage_(total)"]


x_train, y_train, x_test, y_test, train_idx, test_idx = \
    xai.balanced_train_test_split(
            x, y, "Date", "Temperature(degree)", "Wind_speed(m/s)","Humidity(%)", "day_of_the_week", "sarturday", "sunday", "public_holidays_(weekdays)","total_holidays",
            min_per_group=300,
            max_per_group=300,
            categorical_cols=categorical_cols)

x_train_display = df[train_idx]
x_train_display
x_test_display = df[test_idx]

print("Total number of examples: ", x_test.shape[0])

df_test = x_test_display.copy()
df_test["usage_(total)"] = y_test

Model Code

import sklearn
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, mean_squared_error, roc_curve, auc
from tensorflow.keras.layers import Input, Dense, Flatten,Concatenate, concatenate, Dropout, Lambda, Embedding
from tensorflow.keras.models import Model, Sequential

def build_model(X):
    input_els = []
    encoded_els = []
    dtypes = list(zip(X.dtypes.index, map(str, X.dtypes)))
    for k,dtype in dtypes:
        input_els.append(Input(shape=(1,)))
        if dtype == "int8":
            e = Flatten()(Embedding(X[k].max()+1, 1)(input_els[-1]))
        else:
            e = input_els[-1]
        encoded_els.append(e)
    encoded_els = concatenate(encoded_els)

    layer1 = Dropout(0.5)(Dense(100, activation="relu")(encoded_els))
    out = Dense(1, activation='sigmoid')(layer1)

    # train model
    model = Model(inputs=input_els, outputs=[out])
    model.compile(optimizer="adam", loss='categorical_crossentropy', metrics=['accuracy'])
    return model


def f_in(X, m=None):
    """Preprocess input so it can be provided to a function"""
    if m:
        return [X.iloc[:m,i] for i in range(X.shape[1])]
    else:
        return [X.iloc[:,i] for i in range(X.shape[1])]

def f_out(probs, threshold=0.5):
    """Convert probabilities into classes"""
    return list((probs >= threshold).astype(int).T[0])


model = build_model(x_train)
model.fit(f_in(x_train), y_train, epochs=50, batch_size=512)

Error

Epoch 1/50
1/1 [==============================] - 1s 713ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 2/50
1/1 [==============================] - 0s 12ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 3/50
1/1 [==============================] - 0s 15ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 4/50
1/1 [==============================] - 0s 15ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 5/50
1/1 [==============================] - 0s 15ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 6/50
1/1 [==============================] - 0s 12ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 7/50
1/1 [==============================] - 0s 15ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 8/50
1/1 [==============================] - 0s 12ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 9/50
1/1 [==============================] - 0s 13ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 10/50
1/1 [==============================] - 0s 15ms/step - loss: nan - accuracy: 0.0000e+00
Epoch 11/50
1/1 [==============================] - 0s 13ms/step - loss: nan - accuracy: 0.0000e+00

Answer 1

For me this looks like a problem with your loss function. Categorical_crossentropy is normaly used in multiclass prediction. So to me this looks not like you have a classification problem but more like a regression problem. That's why you should use a loss for regression like mean_squared_error:

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

Another thing might be your activation function. Most of the time if you try to predict numerical values a linear activation function is used. A helpful read to choose the correct activation function might be this. That means you would have to change your activation function of the output layer to linear