Week 1 assignment 3: inputs for LSTM

Course Q&A Deep Learning Specialization Sequence Models
1

hi community

for lstm the input function requires the batch,Tx and the “features” for each sample in batch. but why is it that the C1 initialises the input layer as X = Input(shape=(Tx, n_values)) but says that X is of shape a m,Tx , n isnt there some sort of reshaping required prior to the use of the input.

i tried referencing this video to check if my understanding was correct and it somewhat affirms what i know. can anyone tell me what i have missed out.

also 1 more thing does anyone have any useful links with regards to the loop creating layers in the djmodel function? i just want to know more about how it is done and the inner workings that make it work.

2

LSTM input has format (BATCH_SIZE, NUM_TIMESTEPS, NUM_FEATURES_PER_TIMESTEP). When specifying model input, you don’t have to specify the batch dimension as indicated in the exercise markdown.

image
image743×147 17.1 KB

Does this link on tensorflow functional api help understand how layers are chained?

3

Weird i dont seem to have that line but tks. Might have deleted it on accident.

4

Here are steps to refresh the workspace and get started from the starter code.

1 Like