# Tag Info

Accepted

### What is the total number of actions and rewards count

TL;DR In the DQN paper, each environment was trained for 50 million frames, grouped in fours without overlap, so there were 12.5 million state, action, reward next-state records used. The above direct ...
• 23.1k
1 vote
Accepted

### Deep Features of Images - Better to use an unsupervised approach, or train a classifier with many classes?

This is exactly the problem I am currently working on. I don't suggest that you use a supervised method to learn latent representation, as the model might learn shortcuts or only a most meaningful ...
• 732

### How "Patch Merging" works in SWIN-Transformers?

What happens during the patch merging? Concatenationating is only one part of the whole operations. Below, I quote the code from the original implementation and explain step by step what happens. <...
1 vote

### What is the right way to find the alphas in this equation?

Equation (7) in the Grad-CAM++ paper is linear. In fact for a given class $c$ we have just one equation and many unknowns (the $\alpha_{ab}^{kc}$), hence the equation is underdetermined and will have ...
• 131

### Is the Machine Learning community going against Occam's razor?

Regularisation (at least, $L_1$ and $L_2$) can be viewed as an application of Occam's razor. Regularization is widely used in ML and studied in learning theory (see, for example, the structural risk). ...
• 33k

### Finding patterns in binary files using deep learning

I was writing a comment, but I'll write an answer instead. With a dense network your model has too many parameters an will overfit to the data. Even if it learns that "0 and 100 tend to appear ...
• 315
1 vote

### Why is it important/beneficial for an activation function to be zero-meaned?

The first part of this answer is regarding your concrete question and the second part summarizes things on activation functions in general because I believe that there are more important factors than ...
• 166
1 vote

$\tilde{A}$ is related to normalized Laplacian matrix that "shows many useful properties" of matrix $A$. Note that: Since the degree matrix $D$ is diagonal, its reciprocal square root $D^{-{... • 1,663 0 votes ### Does make sense to add an additional Attention layer while Fine-Tuning Bert? The usual practice is to the first token embedding as an input to the classifier, which forces the last layer to collect the relevant information from the previous layers to this particular embeddings.... • 141 1 vote Accepted ### Arcface implementation for image similarity produces opposite embeddings for positive negative image pairs Cosine similarity,$s$, is an angle distance. The nature of its distance is represented in the range$s \in [-1, 1]$being$s=-1$very dissimilar and$s=1\$ very similar. However the intuition of ...
• 998

Top 50 recent answers are included