Questions tagged [loss-function]
A function used to quantify the difference between observed data and predicted values according to a model. Minimization of loss functions is a way to estimate the parameters of the model.
532
questions
0
votes
0
answers
33
views
Setting derivative to 0 when minimizing expected loss over possible data samples
Suppose, given $x\in \mathbb{R}^d$, you want $\theta$, which is the solution of
$$
\text{argmin}_{\theta} \mathbb{E}_a[L(\theta;x,a)]
$$
where $a \sim \mathcal{N}(x,\Sigma)$ and $L$ measures a (convex)...
1
vote
0
answers
33
views
Analysis of relationship between accuracy and total loss (or cost) during training with logistic loss function and threshold 0.5
I'm trying to understanding the relationship between training accuracy and training loss in classification tasks, specifically using logistic regression. When using logistic loss as the loss function ...
- 155
0
votes
0
answers
26
views
Ordinal log-loss in a multiclass classification in XGBoost?
I have a multi-class problem that which classes are simultaneously mutually exclusive and have ordering. You can think of the classes as being some score: 0 (Low), 1 (Medium), 2 (High).
What I would ...
- 111
0
votes
0
answers
23
views
Generator loss not decreasing while training GAN
I’ve been attempting to create a basic GAN to generate images using this database of flowers (https://www.robots.ox.ac.uk/~vgg/data/flowers/102/).
I’ve spent a few days on this, and largely based my ...
0
votes
0
answers
38
views
Doubts on a custom loss function for regression problems
From what I read, I know we don't use log loss or cross entropy for regression problems. However, the entire logic behind binary cross entropy(say) is to firstly squeeze the y_hat between 0 and 1 (...
- 23
0
votes
0
answers
11
views
Meaning of mean squared error in multistep prediction
In multistep prediction with LSTM(keras), say we had this kind of result:
target = [[1,2,3] ,[4,5,6] ]
predictions = [[1.1,2.2,3.3] , [4.4,5.5,6.6]]
When we choose mean_squared_error as the loss ...
- 23
0
votes
0
answers
42
views
Custom Loss Function Returns Graph Execution Error: Can not squeeze dim[0], expected a dimension of 1, got 32
I have built a loss function which adds time and frequency weighted averages and variances to the MSE:
...
0
votes
0
answers
15
views
Minimize MAE loss for a target that is sum of two other targets
Working on a regression modelling task where my dataset have some feature columns, two more columns A, B and a target column T. The goal is to predict T, and minimize MAE, that is ...
0
votes
0
answers
49
views
Custom loss and metric functions including additional parameter in Keras
The following example is based on this approach.
Similar to that approach, I am wanting to pass an additional parameter with y_true for my custom metric, as both will be used in the computation of ...
- 1
0
votes
0
answers
19
views
In the GAN objective function, why do we first do we first find the D(x) that maximizes the objective function and then maximise wrt the generator?
The GAN objective function is optimised like this: argmin(argmax(L(G,D))) where the argmax finds the D (Discriminator) that maximises L(G,D). Why is it not the other way around, i.e. argmax(argmin(L(G,...
0
votes
0
answers
17
views
Cost function looks like the real math which is responsible for actually working on out problem statement If I talk on a whole and on a surface level?
Looking at the cost function for say linear regression, apart from changing the weight or the parameters, the cost function does the real job, right? If it is correct, what does cost function do in ...
0
votes
0
answers
9
views
Search recall optimization - what appropriate loss function to use?
I am studying machine learning and wanted to work on a project of my own so that I have better chances after graduating college. I'm studying the application of ML to improve searches using a toy ...
- 167
0
votes
0
answers
33
views
Training loss is much higher than validation loss
I am trying to train a neural network with 2 hidden layers to perform a multi class classification of 3 different classes. There is a huge imbalance to the classes, with the distribution being around ...
1
vote
0
answers
16
views
Avoid overfitting to noise by a noise penalty approach instead of early stopping?
I came across this article on deep learning for computational MRI and found an interesting sentence "However, early stopping has to be performed to not overfit to the noisy measurements." ...
- 11
0
votes
0
answers
38
views
Understanding the desired behavior of the loss function of Variational Autoencoders
So I understood that when training VAE, we need to weight the KL part of the loss with a weight less than 1 so that the reconstruction loss can get a chance to learn (avoiding the posterior collapse).
...
0
votes
0
answers
11
views
Loss function for classifcation rewarding closer guess?
The default loss function in multi class classification is cross_entropy, which treats all wrong guesses equally. If the distance between buckets are meaningful, for example, given the real bucket is ...
- 101
2
votes
1
answer
68
views
What is the benefit of the exponential function inside softmax?
I know that softmax is:
$$ softmax(x) = \frac{e^{x_i}}{\sum_j^n e^{x_j}}$$
This is an $\mathbb{R}^n \implies \mathbb{R}^n$ function, and the elements of the output add up to 1. I understand that the ...
- 123
0
votes
0
answers
20
views
Is there a canonical cross entropy from the confusion matrix?
In
Wu, MT. Confusion matrix and minimum cross-entropy metrics based motion recognition system in the classroom. Sci Rep 12, 3095 (2022). https://doi.org/10.1038/s41598-022-07137-z
the author uses a ...
- 101
2
votes
1
answer
86
views
Representation of Strictly Proper Scoring Rule for Multiclasss Classificaiton
I am working on a classification problem, using features $\mathbf{x}$ to predict a target variable $y \in \mathbb{N}_0$. By a strictly proper scoring rule I mean a loss function $\ell(y,\hat{y})$ for ...
- 93
1
vote
0
answers
43
views
Replication of XGBoost's binary:logistic loss
I am trying to replicate XGBoost's logistic loss function as a first step before implementing my own custom loss functions. Following from here and looking at the original code in git repository, I ...
- 111
0
votes
0
answers
48
views
Train neural network to predict multiple distributions
I aim to train a neural network to predict 2 distributions (10 quantiles, i.e. deciles) at 5 time points. So my y is of shape:
...
- 30
0
votes
0
answers
48
views
Need feedback on idea for new regularization term
I've been working on creating a regularization term that ensures that correlated attributes are given similar weights in a linear model. This helps to avoid some of the inconsistency in the weights of ...
- 3
0
votes
0
answers
45
views
The cost function gets stuck at 120 epochs
I did a neural network in c++ to recognize handwritten digits using the MNIST dataset without any neural network pre-existing libraries. My network has 784 inputs neuron (the pixel of the image), 100 ...
- 1
0
votes
0
answers
55
views
XGBoost Architecture Diagram required
Good Day!
My topic is general and theory related, about XGBoost working. I am searching XGBoost Architecture Diagram. I know it works on principles of Decision Trees, Bagging, Random Forest, Boosting, ...
- 1
0
votes
0
answers
49
views
BCE loss stuck at 0.693 in the beginnng of training and then started to decrease, why?
I'm using a Transformer encoder with a binary cross entropy loss for CTR prediction. The training batch loss is at around 0.693 constantly for the beginning several thousand steps (batches). I'm using ...
- 392
0
votes
1
answer
47
views
How to implement a custom loss with a non-mathematical operation (simulation) that backpropagates with PyTorch?
I am writing a Neural Network, which output is not used directly for the loss-function, but rather as the input for a simulation model. After the simulation ran, I am using the simulated_value and the ...
- 1
0
votes
1
answer
51
views
Transpose of a 3D tensor
I need to transpose a 3-dimensional tensor of the shape (batch_size, N, M) to (batch_size, M, N) in a custom loss function in Keras with tensorflow as the backend. I tried using the following function
...
0
votes
0
answers
198
views
How can I interpret the results of my loss functions?
I use yolov8 for object detection. The results for my training look like this:
As you can see in general my validation losses are quite higher than my training losses. Here the comparison of box_loss ...
0
votes
1
answer
33
views
How to select the validation loss value in this model to be compared with other models?
I'm training an LSTM model. I'm confusing about the validation loss of the model. Which value better represents the validation loss of the model? Is it the last value I obtain in the floowing loop, or ...
0
votes
0
answers
29
views
Which loss function should I use if multiple values are correct?
My task is to create a QA-model. I give it a context and a question that it should answer. The answer is usually one word, so a very simplified input would be e.g.
Context: "Max eats a banana. ...
0
votes
0
answers
29
views
How to implement a custom loss function acting differently on multiple instances with keras?
I want to reproduce the results in "Online Neural Networks for Change-Point Detection" Hushchyn et al., but I'm having trouble implementing their loss function with Keras. The algorithm ...
1
vote
1
answer
31
views
GAN Output Gradient Calculation
Loss function for discriminator, which needs to be maximized: -log(D(x)) + log(1-D(G(z))).
Loss function for generator, which needs to be maximized: log(D(G(z)))
What would the calculation of the loss ...
- 11
0
votes
0
answers
11
views
Learning target of Denoising Diffusion Probabilistic Model
I am trying to understand the learning target of DDPM. Trying to understand $D_{KL}(q(x_{1:T}|x_0)||p_\theta(x_{1:T}|x_0)$ in following line.
$$
-\log p_\theta(x_0) <= -\log p_\theta(x_0) + D_{KL}(...
0
votes
1
answer
188
views
How complex can I make a classifier's loss function in Scikit-Learn?
I want to customize the loss vanilla loss function being used by scikit-learn classifiers like the Logistic Regression classifier, etc.
For example, if the vanilla empirical risk minimization ...
1
vote
0
answers
30
views
What is the l2-norm of a scalar
What is the meaning of the l2-norm when dealing with scalar values? I'm assuming it would be the same thing as taking the absolute value.
For context: I am trying to implement the clustering method ...
1
vote
1
answer
335
views
PyTorch CrossEntropyLoss and Log_SoftMAx + NLLLoss give different results
As per PyTorch documentation CrossEntropyLoss() is a combination of LogSoftMax() and NLLLoss() function. However, calling CrossEntropyLoss() gives different results compared to calling LogSoftMax() ...
- 113
0
votes
0
answers
41
views
CTC loss Expected input_lengths to have value at most 144, but got value 174
Can someone help me with CTC loss. I wrote a conformer model for ASR and to train encoder I need CTC loss. But when I train model I got error "Expected input_lengths to have value at most 144, ...
- 1
0
votes
1
answer
47
views
What is this "F" subscript symbol that shows up in this loss function?
i was reading this https://arxiv.org/pdf/2303.14535v1.pdf paper when i came across this:
What is this F? Initially i assumed this was a standard L2 distance but i'm not so sure anymore.
0
votes
0
answers
32
views
Parametric loss function
I would like to train a network to predict two continuously-valued outputs, x and y, from 64 continuously-valued inputs.
For each set of 64 input values, there isn't a single "correct" or &...
- 101
1
vote
2
answers
86
views
Neural network not learning at all
I am training a MLP on a tabular dataset, the pendigits dataset. Problem is that training loss and accuracy are more or less stable, while validation and test loss and accuracy are completely constant....
- 229
0
votes
0
answers
180
views
Weighting and loss function for multi-dimensional output on ECG neural network in Tensorflow
I am working on a DNN that is training on ecg data with a shape of [None,1,2500] and output shape of [None,12,19] where 19 is a ...
0
votes
0
answers
47
views
Problem of constant shift in prediction for neural network regression model with gradient-domain loss function
I'm training a regression model using neural network which is trained on MSE of both output and spatial gradient of output. With some simplification, the model is:
$$
y = f(\mathbf{x};\theta)
$$
where ...
- 1
1
vote
0
answers
53
views
NANs, Infinities, and very large losses with normalizing flows
I am new to normalizing flows and have been trying to use them with a high-dimensional dataset, and I have been running into very large numbers and errors with sampling that don't occur when I use a ...
- 31
0
votes
1
answer
236
views
Why is cross entropy loss averaged and not used directly as a sum during model training(such as in neural networks)
Why is the cross entropy loss for all training examples(or the training examples in a batch) averaged over size of the training set(or batch size) ?
Why is it not just summed and used ?
0
votes
0
answers
23
views
What is log likelihood/maximum likelihood loss?
Very naively, how might one conceptualize log likelihood/maximum likelihood? What situations is it used in? Why would someone prefer this sort of loss function compared to something like an L2 or MSE ...
- 31
0
votes
0
answers
24
views
Weighting loss functions for multi task learning
I am training a multi-task neural network which is predicting a binary target variable, an 18-class target, and a 17-class target. I am calculating the cross-entropy loss for each task, then summing ...
- 1
0
votes
0
answers
98
views
How does cross-entropy loss change with the number of classes?
How does the value of the cross-entropy loss function vary with the number of classes being predicted?
Formally, if the loss function is
$$
L = - \sum_{x \in X} P^*(x) \log P(x)
$$
where $P^*(\cdot)$ ...
- 1
0
votes
0
answers
7
views
Approaching multiple records for one observation; radiomics of 2D slices of a 3D object
Background
I am trying to create a model that can predict Type 2 diabetes in a patient based on MRI scans of their thigh muscle. Previous literature has shown that fat deposition in the muscle of ...
0
votes
1
answer
26
views
Which is the loss function used for validating a CF Recommender System?
I am developing (from scratch) a memory-based CF Recommender System based on movielens dataset.
My CF RS uses a URM (User Rating Matrix) where r_ij contains the rating the user i gave to movie j (or ...
- 117
1
vote
0
answers
382
views
The val_loss is nan, but loss is printing. Both train and validation losses are nan in model.evaluate(), and the acc improves during training
There is a 2-class classification problem, and my loss function is custom. The labels are categorical, and the final activation function is Softmax. During the training, the loss is printed, but the ...
- 11