New answers tagged

2 votes

Generating synthetic data with multiple records per ID

The simplest hierarchical data generating process could use a parameter $p_i$ for the $i$-th product and another parameter $s_j$ for the $j$-th store. Then you could generate data with mean $p_i+s_j$ ...
Stephan Kolassa's user avatar
0 votes

Question on the Partial Derivative of the Cross-Entropy Loss in SGD for Neural Networks

The short answer is your intuition is correct only when the true probability of every class is never 0. In single-label classification, we are dealing with almost the opposite case, every class except ...
pvelayudhan's user avatar
1 vote

Why do we work with factor of likelihoods instead of e.g. a sum for a batch in the negative log likelihood loss function?

No, you cannot simply drop the negative logarithm. This will give you a completely different loss function with unpleasant properties, not only in practice, but also in theory. Assume we make a ...
picky_porpoise's user avatar
0 votes

Relationship Between Neural Network Distances and Performance

You can have a single network structure, with randomly initialized weights for each instance, and randomly shuffled training samples that end up with substantial spread in the "distance" ...
EngrStudent's user avatar
  • 9,375
0 votes

results of a regression predictor

When there is an assumption of a Gaussian-distributed error term (not a Gaussian distribution of all $y$ values, which is pretty much never important (common misconception that it is)), then ...
Dave's user avatar
  • 62.4k
0 votes

Should the testing data be uniquely distinct and come from different source/dataset than the training data?

If you are going to use your algorithm to classify out in the wild, then you will be interested in OOS performance on real-world data. The hope is that the lessons from the synthetic data generalize ...
dimitriy's user avatar
  • 35.5k
1 vote

Is there a standard way for training neural networks with negative-labeled data?

You're describing complementary label classification. This is discussed in Chapter 9 of Machine Learning from Weak Supervision: An Empirical Risk Minimization Approach (2023) by Masashi Sugiyama, Han ...
Sycorax's user avatar
  • 91k

Top 50 recent answers are included