Questions tagged [train]
training (or estimation) of statistical models or algorithms.
298
questions
-2
votes
0answers
27 views
The neural network is not learning properly [duplicate]
I already asked a similar question, however, following some tips (simplifying the structure, data scaling) I managed to make some progress.
Now the neural network returns a curve instead of a straight ...
0
votes
1answer
18 views
Split data into test, training and validation when some patients have multiple observations
I have a dataset that has multiple observations for some patients. Patients are labelled by patient IDs and I would like to split the data into testing, training and validating groups as I am trying ...
0
votes
0answers
19 views
How to force a neural network to uniformly decrease MAPE?
I aim for replicating an numerical (non stochastic) algorithm by a neural network. Therefore I have basically an unlimited amount of data and I wish that the network have an almost perfect fit in ...
1
vote
0answers
13 views
Can I set batch size to 1
I am trying to train a T5 (t5_large) transformer model on some data. Since it's out of cuda memory, I was forced to set ...
1
vote
1answer
19 views
Is it possible to train two neural networks combined with same output?
Say I have two kinds of input that I want my neural network to learn a possible from:
one 2D image
some 1D metadata about the image
This case and similar cases seem problematic to me because the two ...
0
votes
0answers
2 views
How are t+n predictions generated?
I'm trying to understand how ARIMA and other models generate or produce a prediction sequence.
I understand how they fit the model with each point with the training set, and then testing with the ...
1
vote
0answers
15 views
If I know specific pair of characters that model confuses in OCR task how can I fixe it?
I train OCR model to recognize cyrillic handwritten text. I know, for example, that it confuses very often 'Б' with '6'. How can I use this information to fine tune the model ?
Just in case, my ...
1
vote
0answers
77 views
Machine Learning Classification of Changes in a Time Series
I would like to classify the changes in the time series depicted in the figure below—representing an angular position history of a robot about the vertical axis—into increment/decrement of roughly (...
1
vote
0answers
10 views
Percentage of positive target values in dataset and the train/test split
Suppose that in my dataset of 100 observations, only 25 have a target variable equal to 1, while the other 75 have target variables equal to 0. Should the portion of target values that are positive ...
0
votes
0answers
13 views
Reporting cross-validated score and test score
Is this the right way to do cross-validation and testing of a machine learning model:
Split my data into train (85%) and test (15%)
Do hyperparameter tuning based on the train set
Do 10-fold cross-...
1
vote
1answer
13 views
Shall I present predictions on the (oversampled) training set as well?
I am dealing with an imbalanced classification problem and used oversampling on my training set to to predict on my testing set.
My PI insists on presenting evaluation metrics of the different trained ...
2
votes
0answers
41 views
Interview question: train/test error and “best” model
I recently had a puzzling interview question and I am wondering whether anybody can tell me the intended answer.
The question shows train and test error for three models plotted against the number of ...
0
votes
0answers
22 views
Two datasets have different imbalanced class. How to split?
I'm new in machine learning. Actually I have two datasets files as my scrapping results from different news webpages. I want to preprocess to just selecting relevant news. So I would like to perform ...
1
vote
1answer
53 views
How LightGBM deal with a new categorical value in the test set
Suppose I have the training data set $(X, y)$ where $X$ is my feature space $(x_1, \dots, x_n)$. Let $x_1$ be a categorical feature column. In my test set, if the feature $x_1$ takes a new categorical ...
1
vote
0answers
52 views
Train and test score - overfitting?
I have hourly time series data with a range of two years. I want to test my model when predicting my target variable (continuous) for a specific week. I'm doing the following:
Splitting my data into ...
6
votes
4answers
346 views
Train and Validation vs. Train, Test, and Validation
I am embarking on a new job that will give me the opportunity to do some cool machine learning stuff. I haven't touched this stuff on a deeper level since graduate school and I wanted to get some ...
0
votes
0answers
7 views
How to evolve an already trained model with new data
How to evolve a model when new data come out?
Suppose I have a model $M_1$ trained on data $D$.
After the model $M_1$ is trained, new data $T$ is available.
How to train a new model $M_2$ from $M_1$ ...
0
votes
1answer
79 views
Random Forest Regression and Overfitting
I did gridsearch with corss-validation on a trainingset to search for best hyperparameters for a Random Forest Regressor. And indeed the best parameterset gives good results in cross-validation (R^2 ~ ...
0
votes
0answers
22 views
How to compute TBATS forecast accuracy without specific test set?
I have used the TBATS model on my data and when I apply the forecast() function, it automatically forecasts two years in the future. I haven't specified any training set or testing set, so how do I ...
0
votes
1answer
24 views
Method for selecting training data and identify predictive genes in a survival model
I am performing a coxph survival analysis and want to split my data into a training and test set. During the training I would like to identify the "best" ...
0
votes
1answer
40 views
How is stratified sampling better than sampling equally from all classes while crossvalidating?
I can see that stratified sampling helps in maintaining the same class distribution in the training set as in the original dataset. However, my understanding is that ideally, the model should be ...
1
vote
2answers
37 views
Training a Neural Network on Music
I'm very picky about the music I like. I can't say I like a genre in particular because my tastes in music boil down to how a specific song sounds. For a while now, I've had a vision of creating a ...
0
votes
0answers
23 views
High precision, but low recall on training data. What could be wrong?
I am training RNN for timeseries binary classification. I observed that network has high precision, but low recall on both training and testing data. I tried multiple architectures, but same problem ...
1
vote
1answer
135 views
Why not use large ($n - [n^{0.75}]$) validation sets in machine learning training loop?
I am seeking feedback in the form of answers to a bold question. Please limit comments to requests for clarification of the question only.
There appears to be an influence on machine learning ...
2
votes
0answers
34 views
Uplift modeling for Train, validation, test data sets
I am wondering when I should tune hyper parameter when we build uplift model.
In a normal machine learning context, data will be split into train, validation, test.
And, we train the model with train ...
0
votes
0answers
49 views
training variational autoencoder - loss is 0 val loss is 0
I am training a VAE in Keras on the colab. The issue that I am encountering is that after the first epoch the loss is 0 and the validation loss is 0 (but the model didn't learn anything).
this is the ...
0
votes
0answers
43 views
Why does switching the architecture result in nans during training?
What are my options to figure out the casue of spontaneous nans/infs during training?
Basically I changed the model from a custom resnet to the vanilla resnet(and also later a vggnet varation) and ...
3
votes
2answers
44 views
R2 on out-sample data set
The conventional definition of $R^2$ is:
$R^2 = 1-SSE/SST$, where SSE denotes sum of squared errors and SST is total sum of squares ($n\times variance$, n being number of sample points in train set).
...
0
votes
0answers
24 views
Small sized training set and results varying based on cross-validation split
I need to try to build a classifier based on around 30 instances. The outcome can also be that the dataset is not large enough for this purpose, however I'm not sure on how can I justify this outcome ...
4
votes
1answer
124 views
Does it make sense to regularize the loss function for binary/multi-class classification?
When discussing linear regression it is well known that you can add regularization terms, such as,
$$\lambda \|w\|^2 \quad \text{(Tikhonov regularization)}$$
to the empirical error/loss function.
...
0
votes
0answers
16 views
Using a poset or directed graph as input for a neural network
I'm not sure if this is the right community to post this in but I would appreciate any help. As the title states, I'm trying to train a neural network using some unconventional input. I'm wondering if ...
0
votes
0answers
10 views
How to train sentiment analysis with neutral
I want to classify sentiment on blog and forum sites. Each post can be positive, negative or neutral. Most of the existing sentiment classification are for two classes, positive and negative. I am ...
1
vote
0answers
60 views
Unit Deviance on test set R
I am evaluating a set of predictions coming from different models against a set of actual values in a regression problem.
I do not want to use the mean squared error as my evaluation metric because my ...
0
votes
1answer
28 views
validation error and test error when limited data is available
In machine learning, to get an unbiased estimate of model performance, we split data 80:20 into train and test set. We use the training set for model training and model selection according to cross-...
1
vote
0answers
8 views
How easily train model with set of hyperparameters with different algorithms?
I have tried hyperprameter optimization and do get top 15 best combination of parameters using gs.cv_result but now I would like to train a model on these 15 combination of hyperprameters seperately....
0
votes
0answers
115 views
Neural network training converges for several epochs, then diverges badly
I have a VGG-like network that I have trained from scratch on a multi-class dataset of my own. The results suggested there were probably some data errors somewhere, so I thought I would train the same ...
1
vote
0answers
38 views
Predicting when and where a conflict is most likely to take place (glm/ train test set error)
I am having difficulty analysing this dataset (I am new to statistics and R). I am trying to model when a conflict is most likely to take place (before, during or after deforestation), and at what ...
10
votes
3answers
685 views
Is it in general helpful to add “external” datasets to the training dataset? [closed]
Several people have already asked "is more data helpful?":
What impact does increasing the training data have on the overall system accuracy?
Can increasing the amount of training data make ...
0
votes
0answers
31 views
How to evaluate xgb.cv and is it okay to use min(test_logloss) even if there is large gab between train_logloss?
Please migrate if this is not correct spot the post this question.
I have a dataset including 70K - 1s and 300K - 0s, a type of classifier. I am using Xgboost to ...
0
votes
0answers
13 views
To what extend can pre-training and training affect the results of a prediction model?
If we pre-train a model forecasting COVID-19 with data of SARS, which had a different transmission pattern, will our model be weakened?
If we train the same model with data of (for instance) the USA, ...
0
votes
1answer
65 views
Training of a deep Artificial Neural Network
I have few doubts related to training a neural network with more parameters (weights and biases) than number of data points. I know there exists discussion (on this platform) related to training such ...
0
votes
1answer
96 views
Why does shuffling in train test split have a big impact with my loss and accuracy?
I used Keras for the train test split.
This is what I get when I shuffle during my train test split:
When I disable shuffle by setting shuffle:False this is what I ...
0
votes
0answers
41 views
svm train score is 1.0 and test score 0.996. How to verify its correct?
My dataset has features CO2, PM2.5, Temperature and Humidity and I am trying to predict the Air Quality Index based on this information. I have approximately 24,000 data points for each of the ...
2
votes
1answer
34 views
Is it possible to overfitting within single epoch
Let me put my question first. For a time-series prediciton, is it possible to overfit even within the first epoch, when training data and validation data should all "new" to model?
Features and ...
0
votes
0answers
48 views
Machine Learning: Why do I have this pattern of train and validation accuracy?
I am trying to understand what would generate this pattern of accuracy in train and validation dataset (second and third plot below).
I am training a network to recognize 6 types of faces (they are ...
0
votes
0answers
55 views
Can the size of the training set of a neural network be smaller compared to other models?
If some of you knows what is the rule thumb one in ten rule this what I want to discuss.
Neural network can, nowadays, go deeper in performances. Did you find, according to your experiences, that you ...
1
vote
2answers
18 views
Issues with training on a sample of training set?
I am training an SVM on highly imbalanced data. I have rectified this issue and my ML pipeline works just fine. I have allocated 70% of my dataset for training, however this takes an infeasible amount ...
0
votes
0answers
21 views
How to improve neural network training against a large data set of points with varying magnitude
I am currently using TensorFlow and have simply been trying to train a neural network directly against a large continuous data set, e.g. $y = [0.014, 1.545, 10.232, 0.948, ...]$. The loss function in ...
3
votes
1answer
239 views
What is the relationship between mean squared error and classification error?
I've trained a network using a genetic algorithm and I have two possible fitness functions for my GA: MSE and CErr.
If I use MSE as my fitness function, over time MSE decreases and classification ...
3
votes
1answer
1k views
Keras: What is the meaning of batch_size for validation?
If I understand correctly that batch size is the number of samples used in the training of a NN before the gradient gets updated, then why do we need a specified batch_size for the validation sample?
...
