Sequence Models-Handson.txt - Sequence Data Sequence data is the one which has a temporal relationship within them This includes any time series data

Sequence Models-Handson.txt - Sequence Data Sequence data...

This preview shows page 1 - 2 out of 4 pages.

Sequence DataSequence data is the one which has a temporal relationship within them.This includes any time series data, spoken language, text and audio/video signals.To make any predictions or classification, the network has to go through the entire sequence of data to make sense out of it.Why Not Feed Forward Network?The main drawback of employing a feed forward network for sequential data is that they are limited to accept an only fixed length of data, i.e. the dimensions of the input feature always remain the same.The sequence data can vary in length, for example, the user can utter a very long speech data for translation or just a small sentence to ask for the currenttime.Sequential MemoryIn order to make predictions from a sequence data, the model has to remember theprevious inputs given the current input as shown in the GIF above.This property of keeping track of previous information in the sequence to predict the next output is known as sequential memory.Sequence ModelsTo model sequences, we need a special framework that deals with:features with a variable lengthmaintain sequence orderkeep track of long-term memorySequence model can address these issues.Some of the well-known sequence models are Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU), which you will learn in detail in the rest of this course.Types of Sequence ModelsSome of the variations in Sequence models are:One-to-Many: sequence output, for image captioning.Many-to-One: sequence input, for sentiment classification.Many-to-Many: sequence in and out, for machine translation.Synched Many-to-Many: synced sequences in and out, for video classification.
Background image
Image of page 2

You've reached the end of your free preview.

Want to read all 4 pages?

  • Summer '17
  • RNNs, Sequence Models

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture