0
$\begingroup$

Several terms are used to describe the structure of instances of a dataset. For example, an instance can be 'univariate' if it depends on only one variable or 'multivariate' if it depends on 2 or more variables. I wonder if there are technical terms to distinguish the structure of a, let's call it a 'tuple', and a 'time series' or 'sequence'. Here are two examples that I am trying to describe or distinguish structurally: enter image description here

The left example is a time series, where all 11 samples belong to one instance. The right example is a collection of 'tuples' (if that is the correct term), where each row represents a separate instance.

I always refer to the term 'tuple' here, because the concept seems similar to me with the record in a database or the representation of points from mathematics. I am not sure if 'time series' and 'tuple' are the correct terms to describe how many samples an instance consists of. The terms should be analogous to uni/multivariate to describe the number of features in an instance.

Hence my question:
If univariate/multivariate describes an instance in the horizontal, what terms are used to describe an instance vertically?

$\endgroup$
1
$\begingroup$

We have a Dataset
Dataset has Instances

Instances can be a Vector Or a Sequence.

  • If it is a vector,
    It has Features
    Examples - Image, Tabular data
  • If it is a Sequence,
    It has Time-steps**
    Each Time-step can have Features i.e. Uni/Multi-Variate
    Examples - Video, Audio, Time-series data

**When we use text-data as a sequence, we call the words as "Tokens"

$\endgroup$
2
  • $\begingroup$ I conclude from the example with the image for the vectors that these are not only limited to 1-dimensional vectors, but can also be multi-dimensional (height x width x RGB channel) as in the case of colored images. Time steps, on the other hand, consist of a one-dimensional feature vector. Is that correct? $\endgroup$
    – JohnDizzle
    Mar 19 at 11:24
  • $\begingroup$ Time steps too can have multiple features e.g. If we measure Temperature, Pressure, Humidity from sensors. Similarly, think of video data. It's just has a temporal information across time steps. We can't shuffle the time steps randomly and make a sequence. $\endgroup$
    – 10xAI
    Mar 19 at 15:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.