Skip to main content

New answers tagged

0 votes

Training a model to perform mouse movement like a human

For creating a model to predict human-like mouse movements, i think you should try sequential models because they can handle the time-dependent nature of the task as per your requirements You should ...
Keval's user avatar
  • 191
0 votes

How to classify data which is spiral in shape?

I’ve experimented with multilayer perceptrons (MLPs) to understand their efficacy in handling complex, non-linear patterns like spirals. Here are some insights and configurations that I’ve found ...
Novalis133's user avatar
0 votes

Can the quality of randomness in neural network initialization affect model fitting?

I think theoretically yes, using a TRNG instead of a pseudo one obviously yields better values in anything that relies on randoms (if speed is not a concern). However, I think there's a lot going on ...
talles's user avatar
  • 116
0 votes

Is there a mathematical formula that describes the learning curve in neural networks?

I don't think there's a theoretical reason for the shape, it's just showing the diminishing returns as training goes. As for a mathematical function that describes this, exponential decay may fit well ...
talles's user avatar
  • 116
0 votes

How to classify human actions?

In response to your question, I recommend using YOLOv8 for pose detection. It's a highly efficient and low-cost model for pose estimation, and it performs well in real-time applications. You can use ...
Keval's user avatar
  • 191
2 votes
Accepted

Representing scalars as vectors for the network output

By representing the Rewards and Values as vectors, the network is able to model uncertainty. Instead of choosing one specific reward it can give multiple possible rewards, a non-zero probability. This ...
Lynix's user avatar
  • 33
0 votes

Loss Function not Decreasing

Your problem type is more suited for classification than regression as you want to assign a class to the second largest number and not predict a value. Check out loss functions like CrossEntropyLoss ...
renderbender's user avatar
2 votes

Make a neural network automatically discover classes?

To see the difficulty of this, let’s look at two dog pictures and two cat pictures. Your machine learning model might pick up on the fact that the first two are dogs and second two are cats. It, ...
Dave's user avatar
  • 658
1 vote

An anti-theft project that should identify user's driving pattern

What you’re looking for is outlier detection. Telling users apart may not be super easy in this scenario. The team could proceed from the assumption the users can be identified otherwise (what ...
foreverska's user avatar
  • 1,477
0 votes

If I freeze pre-trained model weights and than train a classifier on top of its embeddings does that called fine-tunning?

I'd like to add another angle to @kostya's excellent answer. I see the lack of use of additional input as a key distinction for not calling this adaptation fine-tuning. Moreover, as in today's ...
David Khosid's user avatar
2 votes
Accepted

Make a neural network automatically discover classes?

You're describing unsupervised learning. Neural networks require a loss function and if you don’t know what you’re learning that can be tough. That’s not to say this isn’t possible, autoencoders ...
foreverska's user avatar
  • 1,477
1 vote
Accepted

Which batch size is optimal for my neural network?

Whilst minibatch size can make a difference to eventual neural network performance, when combined with changes to other hyperparameters such as learning rate, in my experience the impact is relatively ...
Neil Slater's user avatar

Top 50 recent answers are included