We Keep Coding

How to solve the dimension problem for SpatialPyramidPooling2D

I'm currently working with the TensorFlow Addons SpatialPyramidPooling2D layer for image classification and I got the following error when I tried to fit the model.
ValueError: Dimensions must be equal, but are 8 and 20 for '{{node MatMul}} = BatchMatMulV2[T=DT_FLOAT, adj_x=false, adj_y=false](feature, transpose_1)' with input shapes: [?,20,8], [8,20,?]
I doubt that it's something to do with the output shape of the model. The last layer is supposed to be (None,<number_of_classes>) but I got (None,<number_of_channels>,<number_of_classes>). Because the output of SpatialPyraidPooling2D is a 3D tensor.
I tried to solve it by adding a Flatten layer right after SpatialPyramidPooling2D but it ends up the softmax layer giving me an error as below
ValueError: Input 0 of layer dense is incompatible with the layer: expected axis -1 of input shape to have value 1280 but received input with shape [None, 25600]

If you want output of shape (None, 8), I suggest you add a 1D pooling layer after the pyramid pooling thing.
import tensorflow as tf
x = tf.random.uniform((10, 20, 8), dtype=tf.float32)
pool = tf.keras.layers.GlobalAveragePooling1D()
print(pool(x).shape)
TensorShape([10, 8])

Keras error : Input 0 of layer dense_3 is incompatible with the layer: : expected min_ndim=2, found ndim=1. Full shape received: [16]

I want to compare two images represented by features vectors. For this, I used some existing code of siamese neural networks, with just the last part that computes the absolute difference of the two features vectors and then gives a similarity score (using sigmoid function).
So I tried to adapt the code, here is what I've got :
# Add a customized layer to compute the absolute difference between the encodings
L1_layer = tensorflow.keras.layers.Lambda(lambda tensors:K.abs(tensors[0] - tensors[1]), output_shape=(1,16))
L1_distance = L1_layer(K.constant(vectors(x_train, 0)))
# Add a dense layer with a sigmoid unit to generate the similarity score
prediction = Dense(1,activation='sigmoid', bias_initializer=initializers.Constant(0.1))(L1_distance)
# Connect the inputs with the outputs
siamese_net = Model(inputs= keras.layers.Input(shape=(2, 16)),outputs=prediction)
The function vectors(x_train, 0) computes a numpy array containing the two features vectors (each feature vector is also an array of length 16, so the final result of the function is [[vector1],[vector2]].)
When I execute the code, I encounter the following error, at the line where I create the Dense layer :
ValueError: Input 0 of layer dense_3 is incompatible with the layer: : expected min_ndim=2, found ndim=1. Full shape received: [16]
I don't really understand, since I've explicitely specified that the output size of my Lambda layer is (1,16), which has 2 dimensions as expected by the Dense layer.
Thanks for your help.

ValueError: Error when checking target: expected (keras Sequence model layer) to have n dimensions, but got array with shape

I have loaded images to train my model on recognizing one feature in those images.
Xtrain is a numpy ndarray of shape (1380,200,200,3 ) containing 1380 images sized 200 by 200pixels in RGB format
Ytrain has targets. shape (1380,2)
When I train my model (model.fit(Xtrain,Ytrain)) I seem to get a value error on everyone of the layers. As if the input was both Xtrain then Ytrain...
ValueError: Error when checking target: expected batch_normalization_24 to have 4 dimensions, but got array with shape (1380, 2)
Image:

The shape of Keras's batch normalizer layer's output is the same as its input. Since you have only two labels, your final layer in a Sequential model should generate two outputs. You can consider adding a Dense layer like:
model.add(Dense(2), activation='relu')
I also recommend to check your model's architecture using print(model.summary()) and make sure that inputs and outputs match with your dataset and vice versa.

Getting an error while giving input to Conv1D layer in a Keras model

I am using tf-idf vector data as an input for my Keras model. tf-idf vectors has the following shape:
<class 'scipy.sparse.csr.csr_matrix'> (25000, 310617)
Code:
inputs = Input((X_train.shape[1],))
convnet1=Conv1D(128,3,padding='same',activation='relu')(inputs)
Error:
ValueError: Input 0 is incompatible with layer conv1d_25: expected ndim=3, found ndim=2
When I am converting the input to Input(None,X_train.shape[1],) then I am getting an error while fitting because the input dimension has been changed to 3.

As mentioned in the error (i.e. expected ndim=3, found ndim=2), Conv1D takes a 3D array as input. So if you would like to feed this array to it you first need to reshape it:
X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1))
And also set the input layer shape accordingly:
inputs = Input(X_train.shape[1:])
However, Conv1D is usually used for processing sequences (like sequence of words in a sentence) or temporal data (like timeseries of weather temperature). And that's exactly why it takes inputs with shape of (num_samples, num_timesteps or sequence_len, num_features). Applying it on a tf-idf representation, which does not have any sequential order, may not be efficient that much. Instead, I suggest you to use a Dense layer. Or alternatively, instead of using tf-idf you can also feed the raw data (i.e. texts or sentences) directly into an Embedding layer and use Conv1D or LSTM layer(s) after it.

Keras SimpleRNN confusion

...coming from TensorFlow, where pretty much any shape and everything is defined explicitly, I am confused about Keras' API for recurrent models. Getting an Elman network to work in TF was pretty easy, but Keras resists to accept the correct shapes...
For example:
x = k.layers.Input(shape=(2,))
y = k.layers.Dense(10)(x)
m = k.models.Model(x, y)
...works perfectly and according to model.summary() I get an input layer with shape (None, 2), followed by a dense layer with output shape (None, 10). Makes sense since Keras automatically adds the first dimension for batch processing.
However, the following code:
x = k.layers.Input(shape=(2,))
y = k.layers.SimpleRNN(10)(x)
m = k.models.Model(x, y)
raises an exception ValueError: Input 0 is incompatible with layer simple_rnn_1: expected ndim=3, found ndim=2.
It works only if I add another dimension:
x = k.layers.Input(shape=(2,1))
y = k.layers.SimpleRNN(10)(x)
m = k.models.Model(x, y)
...but now, of course, my input would not be (None, 2) anymore.
model.summary():
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) (None, 2, 1) 0
_________________________________________________________________
simple_rnn_1 (SimpleRNN) (None, 10) 120
=================================================================
How can I have an input of type batch_size x 2 when I just want to feed vectors with 2 values to the network?
Furthermore, how would I chain RNN cells?
x = k.layers.Input(shape=(2, 1))
h = k.layers.SimpleRNN(10)(x)
y = k.layers.SimpleRNN(10)(h)
m = k.models.Model(x, y)
...raises the same exception with incompatible dim sizes.
This sample here works:
x = k.layers.Input(shape=(2, 1))
h = k.layers.SimpleRNN(10, return_sequences=True)(x)
y = k.layers.SimpleRNN(10)(h)
m = k.models.Model(x, y)
...but then layer h does not output (None, 10) anymore, but (None, 2, 10) since it returns the whole sequence instead of just the "regular" RNN cell output.
Why is this needed at all?
Moreover: where are the states? Do they just default to 1 recurrent state?

The documentation touches on the expected shapes of recurrent components in Keras, let's look at your case:
Any RNN layer in Keras expects a 3D shape (batch_size, timesteps, features). This means you have timeseries data.
The RNN layer then iterates over the second, time dimension of the input using a recurrent cell, the actual recurrent computation.
If you specify return_sequences then you collect the output for every timestep getting another 3D tensor (batch_size, timesteps, units) otherwise you only get the last output which is (batch_size, units).
Now returning to your questions:
You mention vectors but shape=(2,) is a vector so this doesn't work. shape=(2,1) works because now you have 2 vectors of size 1, these shapes exclude batch_size. So to feed vectors of size to you need shape=(how_many_vectors, 2) where the first dimension is the number of vectors you want your RNN to process, the timesteps in this case.
To chain RNN layers you need to feed 3D data because that what RNNs expect. When you specify return_sequences the RNN layer returns output at every timestep so that can be chained to another RNN layer.
States are collection of vectors that a RNN cell uses, LSTM uses 2, GRU has 1 hidden state which is also the output. They default to 0s but can be specified when calling the layer using initial_states=[...] as a list of tensors.
There is already a post about the difference between RNN layers and RNN cells in Keras which might help clarify the situation further.