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.
534
questions
0
votes
0
answers
10
views
Problem when merge two probability distribution [Pointer Network]
I'm trying to re-implement Pointer Gen Net from this paper. Ya but you don't really need to read the paper.
To sum up briefly,
I have a vector probability distribution over vocabulary called p_vocab. ...
- 101
1
vote
1
answer
37
views
Does using different optimizer change the loss landscape
I plot the landscape using this code, and I notice the landscape shape has changed a lot. My understanding is that the optimizer does not change the loss landscape. But now I'm confused if its just ...
- 127
0
votes
0
answers
14
views
How to combine a classificiation dataset with a pair-wise comparison dataset
Let's say I'm trying to train a neural network that predicts a single output [0.0, 1.0] value that correlates to photo realism which I can use either in a classification setting or for ranking. I have ...
- 191
1
vote
1
answer
39
views
How to comment on goodness of loss functions?
I have two loss functions $\mathcal{L}_1$ and $\mathcal{L}_2$ to train my model. The model is predominantly a classification model. Both $\mathcal{L}_1$ and $\mathcal{L}_2$ takes are two variants of ...
- 123
0
votes
0
answers
8
views
Add a custom loss in a public github repository on distillation
I’m trying to implement a custom loss in a public repository regarding knowledge distillation. The link to the repository is the following:
" https://github.com/DefangChen/SimKD "
The main ...
- 1
0
votes
1
answer
33
views
My custom neural network is converging but keras model not
in most cases it is probably the other way round but...
I have implemented a basic MLP neural network structure with backpropagation. My data is just a shifted quadratic function with 100 samples. I ...
1
vote
1
answer
58
views
Use of Gradient with respect to feature instead of model parameters
Generally, for any machine learning/deep learning system, we compute a loss, $L = l(x, \theta, y)$ which is a function of the input feature vector $x$ (after activation), model parameters $\theta$ (...
- 61
0
votes
0
answers
17
views
Custom loss function for collinearity of 3 embeddings
I am trying to implement a loss function that takes as input 3 embeddings and output a value that is proportional to the collinearity of the embeddings.
This is to shape the latent space of a ...
- 1
0
votes
0
answers
10
views
Cross-Entropy Loss for Prediction Distribution
I'm trying to adapt an example from the Neural Tangents Cookbook which uses a variant of Mean Squared Error (MSE) loss suited for Regression prediction distributions given by Mean and Variance as ...
- 1
0
votes
0
answers
9
views
How to split the data for speaker verification using AAM Softmax loss?
There are several models for speaker verificaiton task (wavlm-ecapa / xvector / ...).
Some of those model where trained with AAM Softmax loss, which gets the number of labels as input.
When training ...
- 383
1
vote
0
answers
20
views
How to force my model heads to learn different things?
I have an Seq2Seq model that has 2 generative LM heads. I want the two heads to focus on different features/styles while decoding. The approach that I was thinking of is adding a distance cost to the ...
0
votes
0
answers
63
views
Custom loss function for multi label classification in catboost?
I have a data frame which I want to use for multi class classification problem. There are total 6 classes (say a, b, c, d, e, f). I want to improve the precision for three classes (say a, b, c) i.e. ...
- 111
0
votes
0
answers
40
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
36
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
1
answer
62
views
Is my model overfitting based on my accuracy/loss curves?
Do those results indicate that my model is overfitting?
- 17
0
votes
0
answers
55
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
104
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
49
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
14
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
100
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
18
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
87
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
21
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
12
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
94
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
17
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
89
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
13
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
130
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
31
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
3
votes
1
answer
94
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 ...
- 103
1
vote
0
answers
70
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
49
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
47
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
67
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, ...
- 101
0
votes
0
answers
89
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
57
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
62
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
408
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
48
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
33
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
54
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
13
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
342
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
38
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
578
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
53
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