Questions tagged [training]
For questions about training networks, rules systems, or other AI system components.
473
questions
1
vote
3
answers
62
views
Would maximizing (instead of minimizing) error of an LLM/HMM lead to complex behavior?
Imagine we have some sort of "next token predictor," either with transformer architecture, LSTM, or just a HMM (though the terminology I use here will be less aligned to HMMs, I believe the ...
- 113
1
vote
0
answers
35
views
Neural network learns to mimic distribution of classes in dataset instead of using signal from input
I'm trying to implement example from a classic AI paper named "Learning representations by back-propagating errors" by Hinton et al.
Example aims at training network able to predict third ...
- 11
0
votes
3
answers
58
views
My model is only improving when learning rate is 1. Should I be worried?
As the title says my GNN with three layers of GAT (Graph attention layers) is only moving the metrics when the learning rate is 1. As generally the learning rate is (0,1) should I be worried?
Also ...
- 1
0
votes
0
answers
28
views
How was the word2vec model trained?
Let's take the CBOW (continuous bag of words) model as the example.
Suppose that, there are $c$ context words, each of which is a one-hot encoding vector. So the total number of elements of input ...
- 1
0
votes
0
answers
10
views
Classes definition for detecting impervious surfaces on aerial photographies
My project is to use deep learning, essentially a UNET segmentation model, to detect impervious surfaces on high resolution aerial photographies.
I wonder if it's better to train the model with many ...
- 101
0
votes
0
answers
19
views
From Text-To-Speech to LLMs: Providing "writing style"
I've just recently learned about Text-To-Speech models and how they are trained. Unlike LLMs, to a provided pair (text, speech), a feature vector ${f}$, that was generated by more speech of that ...
- 143
0
votes
0
answers
26
views
use AI to creat images from texts
I want to use AI to create art images from a single text available in different languages, with the generated images being different depending on the language of the given text.
example 1:
step 1: the ...
0
votes
0
answers
17
views
Does the training time of a neural network increase more if we add a layer at the beginning or at the end?
Let's consider a fixed NN architecture, dataset and hardware. We add a layer, either at the beginning or at the end of the NN. In which case the training time will increase more? Intuitively, I ...
- 194
1
vote
0
answers
19
views
How do LGBM rankers train?
I'm looking into Learning to Rank models - specifically, the LGBMRanker model - and I want to understand how it's able to train. It takes in features, group sizes and labels, and optimizes for a ...
- 393
2
votes
1
answer
57
views
How to train a sample weight model for another ML model?
I'm trying to train a ML model, however the predictability of the different samples varies, i.e. some samples are inherently much harder to predict/estimate than others. Poorer predictions for these ...
- 123
0
votes
0
answers
25
views
Multi-Agent DQN not learning for Clean Up Game - Reward slowly decreasing
The environment of the Clean Up game is simple: in a 25*18 grid world, there's dirt spawning on the left side and apples spawning on the other. Agents get a +1 reward for eating an apple (by stepping ...
- 1
0
votes
0
answers
20
views
High Accuracy ML.NET Image differentiation model
I have a relatively big dataset (100+GB) that has 35 categories. All of them are microscopic images with slight differences. Although ML.NET documentation itself declares that training time should be ...
1
vote
2
answers
94
views
Tips and tricks when training a very large language model?
Have never trained a (very) large language model, so I am wondering if the process is the same as training a (regular) language model, i.e. you prepare the data, set up the architecture, ...
2
votes
1
answer
651
views
Why do transformers compute the loss over the prompt?
When fine-tuning large language models, which are commonly decoder transformer architectures, sometimes we want to compute the training loss over the entire prompt, sometimes just the completion ...
- 931
0
votes
1
answer
27
views
How to label missing/default values for a named entity recognition dataset
I am building the training dataset for a named entity recognition model, with 2 tags: Name and Category and I am using a pre-...
- 3
0
votes
0
answers
5
views
Mechanism of Prediction Readjustment in Supervised Learning and Role of Self-Attention in Sequence Data Relationships
In supervised learning, when the prediction deviates significantly from the expectation, how does it "readjust"?
And... LLMs are a subset of deep learning, just as generative AIs are.
Is the ...
2
votes
1
answer
148
views
Calculating mutual information between layer outputs and targets in a neural network
I've seen in several papers that it is possible to calculate the mutual information between a layer's outputs and the desired outputs. For example:
Source: https://www.ncbi.nlm.nih.gov/pmc/articles/...
- 73
0
votes
0
answers
27
views
Optimal Neural Network Structure for 'Crystal Rush': Single vs. Modular Networks?
Problem
I'm attempting to design a neural network for the bot programming game Crystal Rush. Given the game's complexity, I anticipate needing a vast neural network to manage bot movements, resource ...
user75399
0
votes
0
answers
31
views
Simple model to predict road traffic?
Given GPS tracking data of many trips belong to different drivers, different type of vehicles
Each GPC point recorded after 10 seconds
I can map matching the GPS data back to the roads sections
Right ...
- 101
0
votes
1
answer
33
views
YOLOv1 doesn't work for custom dataset
I am currently trying to train my own YOLOv1 network, based on this repository:
https://github.com/ivanwhaf/yolov1-pytorch
The images I want to train look like this:
Three classes I want to detect:
...
- 111
1
vote
1
answer
136
views
What causes my loss curve to consistently oscillate when training an LLM?
Why is my loss curve consistently oscillating? Every 50 steps it jumps back up. I'm assumming there's a bug in my data, since I'm using this colab notebook that shows a proper train/loss at the bottom....
- 11
0
votes
1
answer
46
views
Easiest way to train a neural-network with neurons that deviate from $f_{nl}(x \cdot A)$
I want to model how a neural network would behave for a system of input-output devices that are only approximately similar to a neuron. I think I have a resonable plan for how to do this, but I'm ...
- 221
0
votes
0
answers
19
views
Method of encoding vector graphic data for a GAN
I am working with vector data that I want to train a GAN on. The vector data is a combination of the following forms of vector primitives:
Lines - format: ...
- 143
0
votes
0
answers
16
views
How to Train Multiple Deep Learning Models on Multiple GPUs?
I have access to a GPU server with four gpus. Now I would like to train multiple models or folds of one model, one on each gpu. How can I schedule multiple trainings, so that a new training instance ...
- 186
0
votes
0
answers
10
views
how to train virtex automl to recognise 7 segment display numbers
We're working on a project which needs to read seven segment displays from photos. We've tried a few AI text recognisers but the 'font' is too tricky. So we thought given that we're looking for a ...
- 101
1
vote
1
answer
80
views
How to overcome symmetry in the solution space when learning a simple neural network?
What are good solution recipes to overcome the problem that in neural network learning, when the solution space has symmetries, learning may eventually stall due to the sum of the gradients over the ...
- 61
0
votes
0
answers
61
views
What would be an appropriate way of making an AI for Cell Machine?
The game Cell Machine is absolutely by itself not suitable for AI usage, but a certain niche within a mod of the game (Mystic Mod) known as "vaults" does seem like a suitable use case for AI....
user74135
0
votes
0
answers
18
views
What model should I use to find coordinate locations of objects on an image (not bounding boxes)
I am developing a neural network to count pits on an etched crystal wafer, and I created a dataset by placing points on each pit.
Each line in my annotation file looks like this:
Id, Filepath, ...
0
votes
0
answers
15
views
model versioning and validation of online models
While training an online model, usually we use progressive validation https://maxhalford.github.io/blog/online-learning-evaluation/. My question is what we should do when we detected performance ...
- 101
1
vote
0
answers
498
views
Relation between Batch Size and Micro Batch Size
In distributed training of large models (pipeline parallelism), a mini batch of training samples is divided into n-micro batches. Each device performs forward and backward passes for a micro batch.
...
- 111
0
votes
1
answer
41
views
When do you use epsilon in Reinforcement Learning?
In reinforcement learning (DQN) do I use epsilon when I am collecting examples from the environment or do I use epsilon when I am training the Q network and Target network ?
- 53
0
votes
0
answers
12
views
Consider an AI Image generation system AI IMAGe. How hard is it to remove specific images from the training dataset and all traces of the images?
Consider an AI image generation system called AI IMAGE. Suppose someone wanted to remove 100 of their photos that had been used to train AI IMAGE for a year in generating images. Would it be easy to ...
0
votes
1
answer
22
views
How to learn Categorial Embeddings in Unsupervised Learning?
I want to cluster mixed-type tabular data, for the categorial columns I want to use Categorial Embeddings and then an Autoencoder Network before clustering with KMeans or similar.
Now, when I want to ...
- 1
2
votes
1
answer
596
views
What if we drop the causal mask in auto-regressive Transformer?
I understand the triangular causal mask in the attention is used to prevent tokens from "looking into the future", but why do we want to prevent that?
Let's suppose we have a model with ...
- 271
1
vote
1
answer
94
views
Is it possible to write/build an AI model without using Frameworks? [closed]
I'm a relatively newbie in this world of Artificial Intelligence, although I am able to use frameworks such as Tensorflow and also understand the general concepts behind training weights and ...
- 113
0
votes
1
answer
34
views
What is the relation between any suitable measure of model complexity, number of training examples and network size in deep learning?
What is the relation between any suitable measure of model complexity, number of training examples and network size in deep learning?
- 153
0
votes
1
answer
26
views
For given units of a measure of model complexity, how many examples do we need to train a network to get the model right and generalize?
For given units of a measure of model complexity, how many examples do we need to train a network to get the model right and generalize?
- 153
0
votes
0
answers
28
views
i have encountered some issues while using Stable Diffusion. It generates some strange limbs for me
My generation method involves using the original image along with SAM to extract masks for the images. Then I use OpenPose to fix the pose of the person and keep the clothes unchanged while generating ...
-1
votes
1
answer
68
views
SWE vs. AI…career change [closed]
I’m looking to switch careers and really want to take the dive, but became hesitant after chatgpt. My question is will it be worth it to devote time, energy and money into the switch if it’s likely to ...
0
votes
1
answer
34
views
How to handle BatchNorm in the last layers of Neural Networks?
I am creating a neural network using batchnorm as a regularization method to enable deep models and prevent overfitting.
I understand that batchnorming supresses the internal covariance shift ...
- 121
1
vote
2
answers
55
views
How to apply backpropagation when one layer of the network is a call-only function (no gradient)?
I worked with Feed Forward Neural Network and VAE and understood backpropagation algorithm. Now I build a VAE network, one layer of it is a very complex vector-to-vector function $f(x)$ (a general '...
- 111
0
votes
0
answers
54
views
How to generate original training videos based on existing videoset?
I am a software engineer who is quickly ramping up on AI tech, but am nevertheless very new to the sector.
A collegue has an extensive collection of training videos, the vertical is wheelchair seating ...
3
votes
0
answers
179
views
Let's Verify Step by Step: Old wine in new bottles?
In their paper "Let's Verify Step by Step" OpenAI proudly presents a new way of reward learning which shall foster LLMs' capabilities of mathematical and logical reasoning:
We've trained a ...
0
votes
0
answers
16
views
How to extract body of a base-model and fine tune with that body on different shape dataset like this situation
In BERT like transformer model (I am not talking about BERT in this thread), it has 2 training objectives Masked Language Modeling and Next sentence prediction right? and BERT model is also supports ...
- 158
1
vote
1
answer
95
views
Where would If my goal was to do OCR pdf text extraction on 1000s of PDFs I have?
So I have lots of data (pdfs) that I want to train an AI model to extract info from. All of them are a little different but have the same key data points. Is it possible to train an AI on the pdfs I ...
- 11
0
votes
0
answers
15
views
Regression Model overestimates in train-mode
I have a Deep Learning Regression model to predict some values. The results are fine when I use the model in Evaluation Mode, but when I turn Training Mode on the model tends to overestimate the ...
- 1
0
votes
1
answer
106
views
gradient descent optimization to make orthogonal vectors
I am trying to use gradient descent to optimize 3 3-d vectors to be perpendicular. To do this I am summing the dot product of each pair of vectors. When they are all perpendicular the result should be ...
- 133
1
vote
1
answer
57
views
Can I speed up NN training by manually guiding training?
I have not found any neural network training methods that recommend manually intervening in the training process while it is happening. However, some experiments I've done seem to show this can be an ...
- 387
1
vote
3
answers
93
views
Why train models on all data available? Why not train on only the best?
Stable Diffusion for example.
Why train it on "ugly" pictures? Why not train a model on only the best and most awarded pictures with the best artist? wouldn't the final output have an ...
- 179
0
votes
0
answers
11
views
Final Model Training Problem - Overfitting
I am working on a CNN project for multiclass classification. I implemented hyperparameter optimization to find the most suitable model, during which I got a best accuracy of 97.38%. I then took this ...
- 11