All Questions

Filter by
Sorted by
Tagged with
0 votes
1 answer
19 views

Why are there two different q-learning formulas?

I found the following q-learning formula: in this youtube video: https://www.youtube.com/watch?v=4C133ilFm3Q&t=521s I'm now a bit confused, since I thought, that the following one is the correct ...
Hans123's user avatar
  • 25
0 votes
0 answers
6 views

Train AI to schedule a league with certain constraints

I need some suggestions on options to look into to train an AI model to schedule sporting events based on certain constraints. Contraints like no back to back games, 2 games per day, 1 per week, keep ...
Mike Flynn's user avatar
0 votes
1 answer
11 views

How long do i need to train a really DEEP network?

When we are training a really DEEP and complicated network (CYCLEGAN, VQ-GAN, VQ-VAE2), how to estimate required time (to be accurate "training steps") for training process? Because from the ...
Тима 's user avatar
0 votes
0 answers
8 views

What is the exact purpose of input modulation gate in LSTMs?

Basically, I was learning about LSTMs where I found LSTMs are made up of three gates: The forget gate, input gate and output gate. However, I came across some sources that state there is a fourth gate ...
MrIzzat's user avatar
1 vote
0 answers
25 views

Why generative models produce mesh structure at the beginning?

I don't really understand the reason of this. I have listed the outputs of different models below. Gan (source: self made simple GAN for CIFAR10) Vq-vae (source: link) What i mean is illustrated ...
Тима 's user avatar
1 vote
1 answer
183 views

How to tell if a model is generative vs. predictive?

How does one tell if a given model is generative AI or predictive AI? Do generative models have more outputs than inputs and vice versa for predictive models?
Geremia's user avatar
  • 205
0 votes
0 answers
27 views

Relation between Q function and V function in Q learning

I'm trying to use RL to solve one problem using DQN. However, the action space depends on the current state and it is huge. Hence, I can't use the generic DQN where the output states equal the number ...
Amanli's user avatar
  • 1
0 votes
0 answers
17 views

Add limited metadata as input to semantic segmentation task

I do semantic segmentation on remote sensing images using UnetFormer model (https://github.com/WangLibo1995/GeoSeg). The input are tiles with several bands (12 or 13 channels). I would like to add ...
seb007's user avatar
  • 101
0 votes
0 answers
10 views

Why vq-vae sometimes struggles with homogenous background?

While training i noticed this problem pic1 pic2 pic3 pic4
Тима 's user avatar
0 votes
0 answers
9 views

Deep Learning: Architecture vs. Features for Performance?

In deep learning, when aiming for peak metric performance, is a well-designed architecture with imperfect features/dataset generally preferable to a poorly designed architecture with high-quality ...
Muhammad Ikhwan Perwira's user avatar
0 votes
0 answers
17 views

Distribution function for lstm and MLP

I have total of 6300 samples, 5800 of which are training data, and 500 of which are testing data. We compare the performance of LSTM and multilayer perceptron (MLP) with one hidden layer in terms of ...
HDD's user avatar
  • 101
0 votes
0 answers
18 views

Dismay felt by not looking at the "right" thing

Some people may feel dismay from time to time because they are not it feel like they are not looking at the right thing. I wonder whether this problem can be solved via wearing a set of glasses that ...
Joselin Jocklingson's user avatar
0 votes
1 answer
35 views

Vision transformer for image segmentation

I am working with vision transformers (ViT) for the task of image segmentation, but I am unsure of which segmentation head to use. I know I need a vision transformer as my backbone, and a segmentation ...
Alex's user avatar
  • 1
0 votes
1 answer
12 views

Attention Layer as input data filter

I can't find decisive sources on that matter. Is it possible to use attention layer as a sort of filter of input data before passing it further to the network? Is it possible to use it to reduce the ...
janek nowaczek's user avatar
2 votes
0 answers
39 views
+100

How to Create a Neural Network Model to Generate Dance Movements Based on Music in MMD Format

I am working on a project where I need to create a neural network model to generate dance movements based on music. My goal is to achieve results similar to this video: https://youtu.be/FrA7f5F9TsI ...
meow meow's user avatar
-1 votes
1 answer
30 views

Difference between .pt and .pth extensions in PyTorch model saving

I'm working on developing models using PyTorch and frequently experiment with pre-trained models. 've read through the PyTorch documentation. When saving models, I've noticed that sometimes the files ...
krp's user avatar
  • 9
0 votes
0 answers
12 views

How to choose the right size LLM for inference to balance latency and accuracy, among several LLMs from 3B through 140B?

I have multiple instances of a same LLM architecture in different sizes from 3B through 72B. I want to build a system that simple questions will be sent to smaller models, and hard questions to bigger ...
Xiao-Feng Li's user avatar
1 vote
1 answer
42 views

How are perplexities over multiple instance aggregated?

The perplexity of the $i^{th}$ token in the $k^{th}$ sequence is $$ P_{ki} = \frac{1}{p(t_{ki})} $$ The perplexity aggregated for the $k^{th}$ sequence is then $$ P_{k} = \left(\prod_{i=1}^N P_{ki}\...
Borun Chowdhury's user avatar
0 votes
1 answer
28 views

Modern methods for denoising and deblurring images with different image sizes

What modern methods exist for denoising and deblurring images? Yes, I haven’t forgotten how to Google and spent a lot of time searching for the code of a neural network that would not have been ...
Quark-Coder's user avatar
0 votes
1 answer
26 views

Why are neural networks optimized instead of just optimizing a high dimensional function?

I know that neural networks are universal approximators when given a sufficient number of neurons, but there are other things that can be universal approximators, such as a Taylor series with a high ...
Yash Nath's user avatar
1 vote
1 answer
31 views

100 layer neural network with 100 hidden units vs. 1 layer neural network with 100 hidden units

Suppose we have a neural network with 100 hidden layers. Each hidden layer has one hidden node, and the hidden nodes employ a universal basis function (e.g. tanh). Now we want to compare this network'...
Hector Auvinen's user avatar
0 votes
0 answers
19 views

Does anyone know how Apple Siri determines when to dispatch a user's question to ChatGPT in iOS 18?

(This is a general question regarding model inference dispatching, not specific to Apple or Siri. I use Apple Siri only as an example to make the question straightforward.) When prompted with a ...
Xiao-Feng Li's user avatar
1 vote
1 answer
44 views

What is the right DRL algorithm to use when the goal in an environment is not fixed?

Let's take the LunarLander environment from the package Gym as an example. In this case, one can run thousand of episodes until the agent learns a good policy. However, there is a condition: the goal ...
Dave's user avatar
  • 182
0 votes
1 answer
21 views

Should we define and name the form of interaction and connection between artificial intelligence and humans? Do we need a specific term for it?

The question of whether the form of interaction and connection between AI and humans should be defined and named, and the need for a specific term for it, arises from the growing role of AI in our ...
user avatar
0 votes
0 answers
18 views

Products built on Llama [closed]

Does anyone know any commercially available applications built on Llama in Europe? I can find lot of examples in academia but am struggling with commercial examples.
Ippolita Magrone's user avatar
0 votes
0 answers
13 views

PyTorch empty() function errors out for no obvious reasons

I am writing a Variational Recurrent AutoEncoder based model for stock returns prediction but just before instantiating the model, have stumbled upon a roadblock. The error thrown is: empty(): ...
insipidintegrator's user avatar
1 vote
0 answers
15 views

Geometry adaptive kernel estimation for crowd counting based on bounding box dimensions

I'm generating a ground truth density map set for an object counting task involving different object classes. The objects are labeled with bounding boxes (i.e [class, x, y, w, h]), which vary in size ...
yuki's user avatar
  • 11
2 votes
0 answers
24 views

A simple minesweeper game's rule in propositional and first order logic

The following picture is a snapshot of a $3 \times 3$ minesweeper game environment. According to the rules of the game, because of the middle 1, there is a mine in one of the other eight surrounding ...
user153245's user avatar
2 votes
0 answers
38 views

Can local learning rules minimize a global loss?

It is widely believed that synaptic plasticity is the way biological brains learn. Artificial implementations of this mechanism are for instance local weight-update rules in Spiking Neural Networks. ...
Alex's user avatar
  • 121
0 votes
0 answers
27 views

What does a feature's integrated gradient actually represent in the context of a regression task?

I've been reading about IGs, but all the articles I've read describe it in terms of a classification task. And in that context it makes a little more sense to me as the change in probability for a ...
Ryne C Johnston's user avatar
0 votes
1 answer
68 views

Which niche textbook(s) should I read to master the math of AI?

I am a SE, CS and math major who’s been casually studying AI for years and I’ve noticed that the math of AI is the hardest part. This is because AI math combines multiple branches of mathematics into ...
user3081098's user avatar
0 votes
0 answers
9 views

How do I continuously track the distance of an object from the camera during object detection in a video?

I'm trying to display the distance of an object from the camera, given a focal length. How do I ensure that the distance is continuously tracked in the video without much error? I'm currently using ...
Asmita Joshi's user avatar
0 votes
0 answers
27 views

CNN: Accuracy gap of 5-7 % between accuracy computed on-the-fly and separate model evaluation on the training set

I am training a CNN for some basic classification task. During training, I compute the training accuracy after every epoch. After the training has finished, I evaluate the model again on the entire ...
StrictlyStationaryPoster's user avatar
0 votes
0 answers
32 views

PixelCnn is not generating images

i have trained own pixelcnn and it is ok. but when i am trying to generate new image, something goes wrong ...
Тима 's user avatar
0 votes
0 answers
18 views

Can Diffusion Models denoise an unseen probability distribution during inference?

I am trying to understand if it is possible to condition the reverse condition process with a weight tensor. Normally this weight tensor is more restrictive (binary) and could be used for downstream ...
ElPotac's user avatar
0 votes
0 answers
16 views

Are more layers or a larger hidden state better for RNNs

Say I have some recurrent model, like Mamba, LSTM, Vanilla RNN,etc. Is an architecture with, for example, 2 layers and a very large hidden state going to do better than an architecture with a lot more ...
Jayson Meribe's user avatar
0 votes
1 answer
64 views

How to use the tf.image.SSIM function

Can anyone please help me understand how to use this SSIM function (https://www.tensorflow.org/api_docs/python/tf/image/ssim)? The filter_size parameter has a ...
Marco's user avatar
  • 119
0 votes
0 answers
17 views

QLora using peft from HF and custom class for binary classification

I am fine-tuning an mistral-7B LLM model for binary classification. I realize it may be an overkill; but we are running some experiments. So far, I have used HuggingFace libraries like peft and ...
kms's user avatar
  • 109
0 votes
0 answers
28 views

Gradient of Discrete Probability Function

I'm coding a gradient descent algorithm in PyTorch with the following loss function: $H(X*A)$, where $H(X) = -\sum_{i=1}^{n} P(x_i) \log_2 P(x_i)$ is the Entropy, and $X*A$ denotes a convolution ...
2 False's user avatar
0 votes
0 answers
11 views

What kind of search perform customGPTs?

Do they perform a similarity search between documents, a semantic search related to a query, nothing of both / other kind?
Evgeniy's user avatar
  • 101
0 votes
0 answers
14 views

CustomGPT and embeddings

Do customGPTs calculate embeddings for uploaded knowledge? Or is it recommended to calculate them and upload together with the knowledge?
Evgeniy's user avatar
  • 101
0 votes
1 answer
43 views

Can a neral network learn dependencies that can be expressed only by iterative approaches

There are many relationships that cannot be described by a "static function" in N dimensions because relationships of the type I am talking about have an iterative element when one of the ...
Igor's user avatar
  • 183
1 vote
0 answers
11 views

RNN Formulation equivalence not clear

In the paper https://arxiv.org/pdf/1211.5063 the authors provide an alternative equation for the more widely known equation to calculate the hidden state at timestep $t$ $$ x_t = σ(W_{rec}x_{t−1} + W_{...
greedsin's user avatar
  • 121
-1 votes
0 answers
28 views

Resizing Images Before Detection with YOLOv9

I would like to resize my images from 128x128 to 640x640 before running this line of code: !python detect.py --imgsz 640 --conf 0.5 --line-thickness 1 --device 0 --weights /content/drive/MyDrive/...
Cemre Aldoğan's user avatar
1 vote
1 answer
30 views

DDPG model outputting a fixed action at every timestep

I am trying to create a Car Following model, for which i am using DDPG. My action is acceleration bounded in a range of [-3,3] m/s2. While training the model, for every state it gives a single ...
Aditya Mishra's user avatar
0 votes
0 answers
32 views

Is a small transformer model able to effectively handle any input length provided it is fine-tuned on it?

Suppose we have a transformer LLM which can do a task such as summarising. I know transformer can technically handle any input length (assume we are not using learned positional embeddings) because ...
Emoticon's user avatar
1 vote
0 answers
29 views

How do I make an artificial intelligence for a game played on a continuous board?

There is a lot of introductory literature about artificial intelligence for time and space discrete games, like chess. A few fundamental methods are offered: Minimax to depth $n$ with an evaluation ...
Ignat Insarov's user avatar
1 vote
0 answers
18 views

What about the loss and custom metric with per-pair weights in multi-class classification?

Let's suppose that we have a multi-class classification problem with 5 classes: 0, 1, 2, 3, 4. The order is not random, they are neighbors. For example, imagine that a labelling is 1. If the ...
Konstantinos's user avatar
1 vote
0 answers
13 views

Best method for finding centroid of 3D object?

I'm somewhat new to machine learning and want to implement a manufacturing application that finds the centroid of an irregular object so that as little material as possible is removed during ...
dstoddard's user avatar
0 votes
0 answers
21 views

Analyzing a deep learning model by constructing a Matrix from input and output data

I am currently completing my bachelor’s thesis, and recently my supervisor suggested that I could strengthen my arguments by explaining why deep learning models perform so well in my case. To give you ...
glaand's user avatar
  • 101

15 30 50 per page
1
2 3 4 5
253