3

This could be a reasonable approach, if it is credible that facial features could correlate with personality traits and if the method for deriving personality traits from twitter posts is reliable. I'd guess a lot of people would rightly ask a lot of questions on these points. You would have to be very cautious though that whatever produces your facial ...


2

In that context, ground truth and gold labels are synonymous. LSTMs don't require any additional supervision compared to other supervised machine learning models.


2

Specific classification behaviour will depend on the particular model form underlying a classification method. The exact response of a model to additional object classes can be derived mathematically in particular cases, though this may be complicated. Since you have not given details of a particular method, I will assume that you are more interested in ...


2

Why not use more thresholds? Like: If probability < 0.25 then class = "weak" If 0.25 <= probability <= 0.75 then class = "mediocre" If probability > 0.75 then class = "strong" But remember that if your interest is to predict correctly (in this case classify) new observations, you won't be able to make a comparison between the truth (2 classes) and ...


2

This is really more a question of English usage than statistics, but I'd say neither of the choices that you list is optimal. If I had to pick from those two, I'd pick 2, but I prefer: % of children under 5 who are febrile.


2

To answer my own question, the optimal way to pick an initial sample according to information criteria such as entropy is a notorious problem called maximal entropy sampling. This turns out to be NP-hard, so I will probably select a small uniform sample of the data and then try to apply maximal entropy sampling afterwards. For approximations, this post ...


2

Since you seem to have enough data, RANSAC could be a good option. That is, we treat the noisy data as outliers. The idea of RANSAC is to iteratively detect those outliers and discard them from the fitting. It is extensively used in image processing for tasks like camera calibration and segmenting primitives like planes in a 3D point cloud, with very good ...


1

Considering the size of the Dataset Mentioned by you can use Heterogenous Ensemble, Homogenous Ensemble or Edited Nearest Neighbors for Filtering the noise in the Data. The homogeneous ensemble is inspired by Cross-Validated Committees Filter[1]. At noise levels approximately more than 10% Heterogenous Ensemble performs worse than Homogenous Ensemble. So ...


1

It makes more sense if you label it by the name of the method which produces the result. For example, "Linear Regression".


1

There is no way to truly answer this question since it will be highly dependent on your use case but things like "target replication" have been shown to improve the performance in some cases. However, I would definitely try the weighted classes first since it doesn't really change your target (though it will affect calibration of your model!)


1

Here is a detailed theoretical analysis on this topic. https://arxiv.org/pdf/1506.01567.pdf. I think that depends on the specific problem and model. The mathematical propositions of the answer above can only be said about general statistical models. In an image data, we are looking at very high dimensions, and mathematics at that level (the extreme non-...


1

Your sliding window will generate something at every time period, assuming you slide it by one time slice at a time. So the feature you're generating from the sliding window will have a value at each time point, which corresponds to your target state. The question is: which time slice that's contained in your window is the one that you attach the value to? ...


1

It seems you're facing a one-class-classification (OCC) problem, where you found "good" features to model your target class $C$ and now asking if it is possible to model the "outlier" class with other features, correct me if i'm right. Under the assumption that your classification problem is an instance of OCC problems: It is not possible to model the ...


1

It's really hard to say. The problem is that what might be considered a strong correlation varies across disciplines. For example: I ask two people to measure the height of a group of children and correlate them. If I find a correlation of 0.7 I'm going to consider that very weak, and think that something has gone wrong somewhere. Similarly, if I give ...


1

(3) doesn't have to be bad if you have some prior about what the clusters might look like, however you wouldn't be using your labelled data optimally. As you point out, you can iteratively train a classifier on its own output. (2) isn't that different from (3) really, it'll depend on how good your metric is (1) is what I would recommend, though it doesn't ...


Only top voted, non community-wiki answers of a minimum length are eligible