Questions tagged [training]

For questions about training networks, rules systems, or other AI system components.

Filter by
Sorted by
Tagged with
0 votes
0 answers
6 views

Training and validation accuracy incerreases and decreases abruptly

What could be the cause of the unusual training accuracy pattern in my neural network, where accuracy increases rapidly and then drops abruptly, then rises again later? Is it related to the learning ...
2 votes
2 answers
101 views

What makes reproducing a model like GPT3/GPT3.5/ChatGPT difficult?

Is it difficult for other companies to train a model similar to ChatGPT, and what makes it difficult? What is challenging about reproducing the results obtained by OpenAI with ChatGPT/GPT3.5? Would it ...
0 votes
2 answers
20 views

When training a DNN on infinite samples, do ADAM or other popular optimization algorithms still work as intended?

When training a DNN on infinite samples, do ADAM or other popular optimization algorithms still work as intended? I have an DNN training from an infinite stream of samples, that most likely won't ...
  • 101
0 votes
1 answer
617 views

Does chatGPT learn or remember from (public) user input? Will it 'fess up to it? I could not get it to reveal [closed]

It started with a question inspired by this video: New Research Suggests to Put AI to Sleep https://youtu.be/0yuQlbCkTJ0 She says: "In this video I discuss a new research paper which suggest a ...
1 vote
1 answer
44 views

For specific tasks, is it better to fine-tune models on examples or just use prompting with the context of the task?

These days large language models cover a vast amount of topics and information, but I wanted to understand: For specific tasks, is it better to fine-tune models on examples or just use prompting with ...
  • 113
0 votes
0 answers
14 views

How many opposing users should be recorded exterior to the average data before being combined?

Outside of the programming toward AI, I am having difficulty putting together a plan on how this machine I hope to build would work. The basic question is: How should it handle user reviews / ...
  • 101
0 votes
0 answers
6 views

How does the 'moment's matching trick' in GAN training improve the diversity of the generated samples?

I was investigating the TimeGAN code, when i stumbled across the 'moments loss' component. In one of the issues, the author states that this is a 'moment's matching trick' used 'to improve the ...
0 votes
1 answer
50 views

Pretrain large model on single GPU [closed]

i want to pretrain some model on P100 which is provided by kaggle. Pretraining on 3 A100 is about 1.5 day. I have 2 questions: Can I put the same seed everywhere so that the results match, train the ...
0 votes
0 answers
29 views

RLLib seems to stop learning at certain point

I work on an AI system for queue management but the learning curve flattens at a certain point and the AI seems to stop learning at all, even if the learning before that point was very good. My model: ...
0 votes
0 answers
21 views

How is model training affected after randomizing the weights of an intermediate layer of a pre-trained model?

Assuming that I have a deep learning model (let's say a ResNet) pretrained on a given dataset (let's say it is ImageNet). I load that model and randomize the weights of one of the intermediate layers, ...
0 votes
0 answers
15 views

Should the number of training iterations of an instance segmentation model depend on the number of instances in the training dataset?

I need to train instance segmentation models on several different datasets. The datasets vary widely in how many instances each image contains. For example: ...
  • 1
2 votes
0 answers
57 views

Training tricks to improve stability of mixed precision

I would love to be able to use automatic mixed precision more extensively in my training, but I find that it is too unstable and often ends in NaNs. Are there any general tricks in training that ...
  • 21
9 votes
2 answers
1k views

How can I encode angle data to train neural networks?

I am training a neural network where the target data is a vector of angles in radians (between $0$ and $2\pi$). I am looking for study material on how to encode this data. Can you supply me with a ...
0 votes
1 answer
45 views

Why does data augmentation using synthetic data generated by one model improve the performance of another model?

I understand from articles like this one that synthetic data generated by one model based on real data can improve the performance of a second model. Can anyone help me understand the intuition behind ...
0 votes
0 answers
19 views

Is it possible or useful to perform quantization-aware training on an untrained model?

Having done some online searches on the topic of quantization, and more specifically, quantization-aware training, I have noticed that all guides and tutorials start by using a pre-trained model, or ...
0 votes
0 answers
17 views

How to speed up the learning process

I have built a network that performans pretty well on my data. The issue I have is that for a larger number of epochs at the start of the training process the val/train acc/loss are stagnating (for ...
1 vote
1 answer
37 views

Why training the same model on the same data can be slower on better card?

Can someone explain why training CNN model (in my case DenseNet201) on the same data, and the same data processing pipeline can be slower on better GPU (RTX3090) than worse one (RTX3060), with the ...
  • 1,220
2 votes
1 answer
308 views

Order of features learned by DNNs during training?

I'm looking for papers probing into the question of what features get learned when (or equivalently what subproblems get "solved" when) during the training process. For example, a paper ...
  • 121
1 vote
0 answers
15 views

Can I use a dataset with real-world images and corresponding actions that the expert took to train an IRL algorithm?

Offline Reinforcement Learning approaches like Inverse Reinforcement Learning/ Batch RL/ imitation learning/ behavior cloning allow us to use previous demonstrations by an expert to learn a policy. ...
0 votes
1 answer
40 views

Training loss decreases very fast after few epochs

I am implementing an ANN whose training loss is in Figure: As you can see training loss decreases very fast and it is approximately 3.2 at epochs 2, 3, ..., 8, ... 10, and so on. (batch learning) The ...
  • 37
-1 votes
1 answer
25 views

Is AI the right tool for solving walking?

Modern robots walk like they had a heart attack and then a stroke... twice. Everyone in AI fiercely believes that the higher number of neurons (or the quality of training with smaller number of ...
0 votes
1 answer
60 views

What to say about time complexity of SVM?

I've read it in some literature now that the training speed of SVM (in general) is very low. Why is that the case? What is to say about time complexity of SVM?
user avatar
0 votes
1 answer
28 views

Model Suggestions for Real Life local Hospital Data [closed]

I'm doing a machine learning project and was looking for suggestions. It's meant to get the date, household, age, sex, doctor, date of the medical appointment, and type of medical appointment of a ...
1 vote
1 answer
30 views

What loss function will be correlated with classification metrics?

Recently I developed a custom training algorithm for deep learning models, based on evolutionary algorithms. Details are not important, except that it also uses decreasing regular cross entropy loss ...
  • 1,220
1 vote
1 answer
72 views

Training Loss Value Increasing instead of Decreasing [closed]

I am developing my first feed-forward fully-connected ANN from scratch based on batch learning mode on a toy training set. I am using back-propagation for calculating the gradient of the loss ...
  • 37
0 votes
1 answer
26 views

How do you train a model with low success rate?

I'm training a model where the samples success rate is low. I mean how do I tackle such situation - maybe only show the samples which match but then the ones that doesn't may match too. But on the ...
0 votes
1 answer
94 views

How is reinforcement learning applied in the real industry?

I'm a newbie to reinforcement learning. While studying reinforcement learning, a question arose about how to apply reinforcement learning in the real world. Assuming that a reinforcement learning ...
0 votes
0 answers
16 views

Do we need negative samples for tuning extractive QA systems?

I am checking the API for the haystack reader: https://haystack.deepset.ai/reference/reader and in particular the fine-tuning method (.train). I was just wondering if for training/fine-tuning a QA ...
0 votes
1 answer
154 views

Is there a way to freeze training for weights, but not biases in PyTorch? [closed]

I'm constructing a neural network where the weights of my first hidden layer (connected to the input) are all 1 (identity matrix), but the biases are variable. Is there a way to "freeze" any ...
  • 133
0 votes
0 answers
12 views

Same 'area under curve' but different parameters?

For 'area under the curve' (AUC) calculations in machine learning, is it the case that if we are mapping the false positives (x-axis) against the true positives (y-axis), that this curve must be ...
0 votes
0 answers
10 views

Tree boosting additive loss

In the XGBoost documentation, they specify that the additive training is done given an objective $obj^{(t)}$ defined as $obj^{(t)} = \sum\limits_{i=1}^n \ell(y_i, \hat{y}_i^{(t-1)}+f_t(x_i)) + \sum\...
0 votes
0 answers
24 views

Why does providing an extra prediction output help stabilize training?

I am reading the PRIMAL: Pathfinding via Reinforcement and Imitation Multi-Agent Learning paper where they tackle the multi-agent path finding problem using reinforcement learning. The problem is ...
0 votes
1 answer
53 views

Can I ignore duplicate records of dataset for training?

I have a a dataset that most of records and their corresponding labels are the same and only timestamp of each records is different from other record. If I ignore duplicate records in for training ...
0 votes
1 answer
24 views

Adding several variables that could be important but can introduce overfitting

Sopose a productivity dataset, where day of the week and months day number are important. I'm thinking to encode these with a one-hot encoding. But if you have few years of data, that features might ...
1 vote
2 answers
42 views

Are RL algorithms suppose to keep learning?

I don't understand if the purposes of RL agents is simply optimizing a model with a reward instead of using labeled data (i.e. in a supervision fashion), or they have also the purpose of keep training ...
  • 41
0 votes
1 answer
54 views

Emergent behavior in AI models that looks similar to natural neural systems

"ImageNet Classification with Deep Convolutional Neural Networks" by Krizhevsky & Sutskever & Hinton describes very interesting emergent behavior of the AlexNet. It was trained on 2 ...
  • 103
7 votes
2 answers
995 views

Deep Learning with Best-so-far instead of Where-you-are

It is my understanding that when training a Deep NN in Tensorflow/PyTorch/... we only keep the current state of the network in memory, except perhaps when we manually decide to save the current ...
1 vote
2 answers
34 views

How to manually adjust output from model? [closed]

I wonder if it is possible to add manual inference to the output of a model? For example, I have a model called 'net', and the output value of 'net' is a vector called v = [v1, ... vn]. v is a binary ...
  • 111
0 votes
0 answers
23 views

Why would the training loss consistently increase over many epochs?

I am getting a very strange learning curve when I try training a neural network which I am not able to explain. I have never seen a learning curve that looks like this when it's a very simple and ...
  • 101
0 votes
0 answers
26 views

How can I estimate how long it will take to train a model with P params on D data points? [duplicate]

How can I estimate how long it will take to train a model with P params on D data points? Assuming my GPU can do T teraflops, and I have a neural network with P params, which I wish to train on D data ...
-1 votes
1 answer
241 views

Should I spend money on a machine-learning capable PC or just use Google CoLab? [closed]

Assuming I have internet access, should I spend money on a PC or just use Google Colab? I'll be doing deep-learning training. Google CoLab: https://colab.research.google.com/
0 votes
1 answer
29 views

Is batch size of 1 a valid choice for a very deep neural network with high memory requirement?

I am training a very deep neural network (Panoptic-DeepLab) with a ResNet34 backbone on Google Colab on CityScapes dataset for Panoptic Segmentation, and noticed that, with a big crop size, the batch ...
2 votes
0 answers
62 views

How does one detect training instabilities in DQN?

I am curious what training instabilities look like in a standard dqn, with or without a target network. I'm assuming the loss function would never converge since the difference between the predicted q-...
0 votes
0 answers
11 views

Approach for predictive model being trained while running at the same time

I have only surface level knowledge in all things AI but am thinking about tackling a specific use case with it in the future. I would like to predict user input, specifically a resident doing stuff ...
  • 1
1 vote
2 answers
511 views

How many training steps does it usually take to train an RL model?

This is my model average rewards as follow image. How to tell if it is undertrained or not convergent? How many training steps does it usually take to train an RL model? And I'm using PPO to train.
  • 223
0 votes
0 answers
26 views

How to setup and train a network with same structure but varying input and output size?

I am dealing with a problem of asset allocation. Asset price is input into the NN and the corresponding weight is output. Let's (to) be simple, assume the NN consists of three layers, the input layer ...
  • 1
1 vote
1 answer
24 views

What kind of NN I need to find ideal ranges and correlation between them?

I’m new to NN and I’m trying to collect material and study. I’m getting through a general high level book, but I’m still struggling understanding what kind of NN I should go ‘deeper into’ for what is ...
  • 113
1 vote
0 answers
43 views

What considerations should I take to train my transformer model?

I want to train my vision transformer model on a benchmark for an image segmentation task: (LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation) (GitHub), but I don't ...
0 votes
1 answer
556 views

Why do we train the discriminators k times but train the generator only 1 time in a iteration in GAN?

In this paper https://arxiv.org/abs/1406.2661 , the codes for training a gan are : Why do we train the discriminator for $k$ steps while the generator only for $1$ step? Why not the other way around?
  • 3
0 votes
0 answers
24 views

Is there any framework for generation of synthetic data where we can set how hard data is to classify?

I am comparing two different training methods of deep neural networks and I was wondering are there any frameworks that can generate x, y datasets but with some input parameter or parameters that can ...
  • 1,220

1
2 3 4 5
9