7
votes
Accepted
Do we need automatic hyper-parameter tuning when we have a large enough dataset?
Unfortunately, even with large amounts of training data, hyperparameter choices can strongly influence the performance of a trained model.
What you can usually drop when you have large amounts of ...
- 29k
6
votes
Accepted
How was ChatGPT trained?
The key ingredient is called Reinforcement Learning from Human Feedback (RLHF), that is having humans rate the model answers and use the feedback to guide the model training.
The official blog ...
- 1,131
6
votes
Accepted
What are "development test sets" used for?
In machine learning, you normally split your data into 3 parts (80-10-10%).
The first part (80% of your initial data) is for the training of your ML model: this is known as the training dataset.
The ...
- 273
6
votes
Accepted
How can I estimate how many photos I need to train ResNet-50 for image classification?
What you want to calculate/estimate is known as the sample complexity in computational learning theory. If you knew the VC dimension of the neural network, you may be able to estimate the sample ...
- 39.1k
5
votes
Can people use neural networks without providing the set of training data?
You cannot train a neural network without training data. It would be like training a football player without making him/her play/watch football or anything that resembles football: it's simply not ...
- 39.1k
4
votes
Accepted
What happens to the training data after your machine learning model has been trained?
In many cases, a production-ready model has everything it needs to make predictions without retaining training data. For example: a linear model might only need the coefficients, a decision tree just ...
- 291
4
votes
Accepted
How do I select the (number of) negative cases, if I'm given a set of positive cases?
Short answer
To select the proper dataset to construct, you should first figure out a metric to use to compare, and then select the dataset construction that gives the better metric. There is no ...
- 1,510
4
votes
Accepted
What causes ChatGPT to generate responses that refer to itself as a bot or LM?
I don't work for openai, so I have no insight into exactly what works behind the scenes to make ChatGPT exhibit this behavior. However, in my opinion this is pretty clearly an example of prompt ...
- 194
4
votes
Accepted
Why does MNIST provide only a training and a test set and not a validation set as well?
The test set should never be seen and ran once at the end of training.
The validation set is used to help you select hyperparameters and it would be cheating to tune your model on the test set because ...
- 56
3
votes
Accepted
Why "large set of training data" is needed in Neural Network AI training?
You are not training the model to give the optimum result for one input; You want the model to produce the minimum loss for the whole population of samples that model may be given. The inputs are only ...
- 476
3
votes
Do we need automatic hyper-parameter tuning when we have a large enough dataset?
You don't NEED a hyperparameter tuner, but it can help in various situations. For example, if your model is not training well, perhaps using a tuner can help.
It's hard to say in which ...
- 118
2
votes
Accepted
How to source training data in ML for information security?
You have implicitly assumed that supervised learning is being used, given the assumption that labels are needed. But this might lead to the following potential problems:
Log file data tends to be ...
- 401
2
votes
How to split data for meta-learning?
I assume in your case what you need to be doing is to collate your 3 datasets together - these would form the training dataset, and then leave the testing dataset aside.
During Meta-Training, the code ...
- 131
2
votes
Accepted
During neural network training, can gradients leak sensitive information in case training data fed is encrypted (homomorphic)?
It will recover the encrypted inputs.
The algorithm starts with dummy data and dummy labels, and then iteratively optimizes the dummy gradients to be close as to the original. This makes the dummy ...
- 334
2
votes
Accepted
What is the effect of training a neural network with randomly generated fake data that satisfies certain constraints?
This is not advisable. If you train your model with random data your model is not learning anything useful, because there is no information to gain from those examples. Even worse it may (and likely ...
- 587
2
votes
Accepted
Does the ANN's training data include the proper output for every neuron?
Does the ANN's training data include the proper output for every neuron?
The short answer is: no (not usually or directly).
The long answer is that you can train neural networks in different ways. ...
- 39.1k
2
votes
Why does linearly decreasing batch sizes result in exponentially increasing training times?
This has a very simple hardware explanation.
GPUs have several thousand cores. If you are not making use of enough of these cores, eventually the cost of moving data over to the gpu comes to dominate ...
- 2,225
2
votes
Accepted
Can people use neural networks without providing the set of training data?
Neural networks are trained by using pairs of example input/output vectors that they learn to associate and can generalise from. In that sense, they always need training data.
For supervised learning, ...
- 29k
1
vote
Accepted
Creating a Dataset from Time Series Data
What you are trying to accomplish is called sequence prediction. It takes in a sequence of data, and spits out a score (target variable).
Time-series data is usually formatted as 3D arrays with shape <...
- 1,898
1
vote
How many unique angles of an object do you need in your image training set in order to correctly classify it?
[I wanted it to be a comment but it's too long :)]
I don't think it's a good approach to split point of views into a group of 12 angles. The main purpose of using neural net is to have model that is ...
- 232
1
vote
Batching together similar length sequences to avoid padding and packing
You can read that this was done in "Attention is All You Need", for Transformers: "Sentence pairs were batched together by approximate sequence length.".
But, with RNN you don't ...
- 11
1
vote
What kind of neural network and GPU should I use to classify images into > 10 000 classes?
You could look for papers that trained models on the Open Image Dataset, which contains around 6k classes, so pretty close to your final use case.
Regarding the dataset size, most datasets include at ...
- 5,183
1
vote
Accepted
Storing training dataset in a platform like mlflow
What you are looking for is a DVC (Data version control) tool. One such tool is available at dvc.org
You can go through their documentation to get started.
https://dvc.org/doc/start
- 76
1
vote
Accepted
Should I include overlapping (input) Data in my training data
In general, both methods are valid to train temporal models. The only thing you need to check is that validation and test-set don't overlap with any of your training samples.
Using the overlapping ...
- 1,511
1
vote
Given a dataset of people with and without cancer, should I split it into training and test datasets such that the same person is not in both?
we are recognizing the disease, not the person.
If you're training a computer vision model with only images and no auxiliary information then a randomized sampling should be enough to prevent the ...
- 5,183
1
vote
Is binary classification using CNN possible if the training data only consists of one class?
While it won't work as you've possibly imagined it, you might find that implementing it as an autoencoder will allow you to train on one class and then identify things that are "not that."
...
- 787
1
vote
Is creating dataset only by augmentation a bad practice?
Data augmentation is usually rotating, cropping and translating images. And this makes sense if your network could meet these kind of images.
If I take a digit recognition like LeNet, it is useless to ...
- 211
1
vote
Why not make the training set and validation set one if their roles are similar?
I think this is best explained using an analogy. Also you seen to have the misconception that you don't tune hyper-parameters for training data. You want to increase the accuracy of the training set ...
- 1,346
1
vote
Why not make the training set and validation set one if their roles are similar?
Idea is to optimize with regards to unseen data in each step in order to avoid overfitting and data leakage so that the final network will be most generalizable to novel data.
First, you initialize ...
- 296
1
vote
Can people use neural networks without providing the set of training data?
In case the question is if NNs can be trained without data, as pointed by others, the answer is negative - any training by definition involves the use of data in some way - supervised, semi-supervised,...
- 2,494
Only top scored, non community-wiki answers of a minimum length are eligible