We won’t spend time on the theory of RNNs here but for a clear and concise introduction to this topic consult the series of posts in. Roughly speaking, this falls into a category of problems which also contains the problem of machine translation the structure of that model serves as the inspiration here. This is an instant cue that recurrent neural nets (RNNs) could be helpful (the dictum being RNNs for speech and language data, CNNs for image data and combined RNNs with CNNs for Image captioning). Unlike the usual regression problems where we predict numerical outcomes, we have a sequence-to-sequence learning problem at hand, where there is temporal structure in the data. Using these examples we must learn some patterns and use that information to predict what new coded tokens (words) could be. Here is the problem at hand we have a few examples of coded sequences and their intelligible counterparts. With that said, let us spur Rocinante and embark on our journey to fight the windmills.
Or - we could construct an encoder-decoder model to guess (almost) all the words!! Being masochists, we will of course choose the latter. Now, we could potentially guess that it is a substitution cipher and then eventually figure out the codes for each alphabet thereby decoding the messages. Suppose also that we had a list of examples of some codes and their corresponding words.
#MRP40 MORSE DECODER CRACK HOW TO#
Let’s suspend our skepticism and suppose we received messages in Morse code, but we didn’t know how to read them.
0 kommentar(er)
