'numpy.ndarray' object has no attribute 'tocoo'

[Anjs04]

In scoring.py, on this line:
cy = y_train_.tocoo()

where y_train = [[] for x in range(y_train_.shape[0])]

[GTmac]

what is the format of your label matrix and what is the dataset you are using? You need to ensure that the label matrix is a sparse matrix in order to call .tocoo()

[jml6m]

I'm getting this same error but in training.py

Has something to do with my input set containing hot encodings (created using keras.utils.to_categorical) which required me to add InputLayer to the beginning of my sequential model as so:

model = keras.Sequential([
        keras.layers.InputLayer(input_shape=(55,), sparse=True),
        keras.layers.Dense(1024, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
        keras.layers.Dense(512, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
        keras.layers.Dense(256, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
        keras.layers.Dense(128, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
        keras.layers.Dense(32, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
        keras.layers.Dense(1, activation=tf.nn.sigmoid)
    ])

AttributeError: 'numpy.ndarray' object has no attribute 'tocoo'

[GTmac]

Do you also get the error on the same line? (cy = y_train_.tocoo()) If so -- what is the format of your input matrix? It must be a sparse matrix defined in scipy.sparse (such as https://docs.scipy.org/doc/scipy/reference/generated/scipy.sparse.coo_matrix.html or https://docs.scipy.org/doc/scipy/reference/generated/scipy.sparse.csr_matrix.html) otherwise the tocoo() method is not appliable.

[jml6m]

@GTmac I am trying to pass a matrix (numpy array essentially) to a keras model that contains values that are both numeric and categorical. From what I'm reading this may not be possible, yet someone in another thread suggested that using a sparse matrix may work. If you have any information that could help please let me know.

But you're right, I did not use a scipy.sparse matrix, therefore it's throwing the error. You can keep the issue closed.