Hello, I'm new to machine learning and I've been trying to build my first recurrent neural network with tensorflow. What I've been trying to do is make this model predict me one step forward (basically just return me one number). The problem I have is the shapes of my tensors.
I have formatted the X data so it would be a list of around 2000 numpy arrays with 5 featues each.
Example:
[
array(1,2,3,4,5),
...
array(3,4,5,6,7)
]
My placeholder for the X data looks like this:
data = tf.placeholder (tf.float32, [None, 5])
My Y data is around 2000 elements in a single list:
[
1,
...
2000
]
My Y data placeholder:
target = tf.placeholder (tf.float32)
My cell:
cell = tf.contrib.rnn.BasicRNNCell (num_units = 100)
My weights and biases:
weight = tf.Variable (tf.random_normal ([100, 1]))
bias = tf.Variable (tf.random_normal([1, 1]))
I tried playing around with the numbers that describe the shapes of the tensors but I can't seem to get it right. Whenever I run the code I get an error that tensor shapes are in some way incorrect (either it was expecting something else or it can't multiply with weights because of the incompatible shapes). Can you please help me find where am I wrong or how should I format the data to make it work?