Questions tagged [machine-learning]

Machine learning algorithms build a model of the training data. The term "machine learning" is vaguely defined; it includes what is also called statistical learning, reinforcement learning, unsupervised learning, etc. ALWAYS ADD A MORE SPECIFIC TAG.

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

A better linear model has less precision(relative to the worse model) at a larger threshold

I trained two models using the same algorithm - logistic regression (LogisticRegression(max_iter=180, C=1.05) for ~27 features and ~330K observations). I used the ...
0 votes
0 answers
9 views

I want to compare the results from two methods. Which test will be better for that? [closed]

For different set of data, I have experimental results by using two different methods. I want to compare the results obtained by the methods. Which test will be better for that?
  • 1
0 votes
0 answers
9 views

Regression Model for Two Non-Independent Groups

I am rather new to machine learning and data formatting, so apologies in advance for any vagueness or significant errors. I am attempting to use a supervised regression ML program to analyze 3-on-3 ...
1 vote
1 answer
8 views

Create dataset by sampling from near-boundary of binary classifier to improve accuracy

Say I have some binary classifier $f: X \to [0, 1]$. I think the following bi-stage training method is straightforward to reduce error. Step1. Sample uniformly from $X$ and create dataset with ...
0 votes
0 answers
26 views

How to compare Predictive Power of Models, why will some models work better than others?

for an assignment, I am doing a regression problem on this dataset: https://archive.ics.uci.edu/ml/datasets/Online+News+Popularity, where I need to run Multiple linear regression, Forward stepwise, ...
0 votes
0 answers
18 views

Bounds of Shapley values for variable importance

Imagine you have a very good predictive model $f$ for a response $y$. Is it possible to bound the "true" Shapley values of $y$ in terms of the Shapley values of $f$ and the prediction error? ...
  • 2,187
1 vote
0 answers
8 views

How to tell feature vectors are sufficient for a graph neural network?

Not sure the right terminology, but I am working on a node classification problem, and I am only able to obtain a training accuracy of 50% for my GNN (test accuracy is 40%). It seems I should be able ...
  • 193
1 vote
0 answers
9 views

Model Architecture Multi-Agents

My question is based on a previews post: How to train LSTM model on multiple time series data? . The task is basically the same. However, my question builds on that: What does the model architecture ...
0 votes
0 answers
18 views

What's a good predictive modeling approach for a very small dataset (50-100 rows) [closed]

I want to model the performance of specific athletes, but instead of taking the big data approach and letting the model generalize (something I've already done), I want to create a high bias model ...
0 votes
0 answers
6 views

Should the order of magnitude of response affect machine learning model performance

Does the range for the response variable affect the modeling performance of supervised machine learning models (such as ANN, SVR,...). Many test functions (e.g. perm, zakaharov, etc.) used to create ...
  • 307
0 votes
0 answers
13 views

Question on the proof of convergence of K-Means

First, the question has answers here and here, however I am still slightly confused. Let's state the problem formally below, extracted from Bishop's Pattern Recognition and Machine Learning book, ...
  • 699
0 votes
0 answers
15 views

Guarantees for optimizer convergence?

I'm trying to invent a custom multi-objective optimizer. For this I evaluate five different starting points shown with the large dot below: The image shows the comparison for loss one and loss two in ...
1 vote
0 answers
7 views

Predicting discount % based on past failure rate and product types

I have 2 datasets with details about discount percentages and failure rates. The discount rates are given for each model manufactured by a certain brand. The failure rates are given at the brand level....
  • 11
3 votes
1 answer
55 views

Is it normal for ROC curve thresholds to be Inf or -Inf?

I am running a binary classification with a random forest via the ranger package in R, and am using the ...
1 vote
0 answers
7 views

Does scikit-learn support plotting calibration curve for multiclass classifier?

I have trained a multi class classifier and calibrated the classified using scikit-learn library's CalibratedClassifierCV class. To check how well the probabilities are calibrated, I tried plotting ...
0 votes
0 answers
18 views

Two-Tier ML Model to Predict Data Relevance

I'm dealing with a video classification ML problem where there are ~10k total scenes, each labeled with a category I need to predict. The problem is that each scene consists of ~100 separate videos (...
1 vote
1 answer
38 views

how to include minority class in all folds

I am working on an imbalance dataset with a 98:2 ratio (1M record in the majority class and 20K in the minority class) I am planning to run my model for 30 folds, I tried with stratified K folds but ...
  • 121
-1 votes
0 answers
20 views

Is my model good? or prone to overfit due to syntax error? [closed]

I am training an echo state network on a .csv file containing a 29x1000 grid (rows 1000 digits long), these rows are necessarily long. My output looks like this: Epoch: 0 Train Loss: 0....
0 votes
0 answers
25 views

Low Pearson coefficient with high R2 [closed]

I have a deep learning regression model which outputs the values of N variables for each sample in the data set (the variables represents particle displacement of N particle system, each sample ...
0 votes
0 answers
15 views

Multi-class Discovery and Classification From Noisy Dataset

I have a set of multiple audio recordings of someone reciting 10 word sentences. There are 3 different sentences, and ~50 recordings of recitations per each different sentence. Each individual ...
  • 101
0 votes
0 answers
12 views

Updating models with new data: how much is needed to keep a model accurate?

How do you update a model when the implementation of your model eliminates new data? I have created a boosted trees classification model that predicts whether or not the amount of money requested (<...
  • 41
0 votes
0 answers
13 views

Unsupervised learning: How to identify differences between clusters?

I'm learning about unsupervised learning and I tried to use KMeans, AgglomerativeClustering and DBSCAN on the same datase. The result was ok, they seems to work fine according silhouette_score() ...
0 votes
0 answers
9 views

Handling different image sizes in a VAE

I am working on segmentation and classification of cells based on their shape. After segmentation using CNN my images have different sizes. Next I want to run these images through a VAE for ...
  • 1
0 votes
0 answers
18 views

Classification of images of different sizes in a Variational Autoencoder

Context: We are using deep learning in image analysis of cells for segmentation and classification in a research setting. So far, we can successfully segment the cells using the U-Net/CellPose ...
  • 1
3 votes
1 answer
143 views

should RMSE compared to the magnitude of response variable?

My response variable has large magnitudes. The mean is 30000ish in the unit of dollars. Then I can only get the RMSE to ~3 which does not feel too bad to me given the magnitude of the response ...
0 votes
0 answers
11 views

Bagging with correlated prediction models

Bagging works best for Ensembles that have a low bias and high variance, such as deep and complex decision trees, when the errors of the models in the ensemble are uncorrelated. I have a formal proof ...
  • 301
1 vote
1 answer
12 views

How can I draw the decision boundary for a simple competitive network?

I know how to train a simple competitive network. Let's say I have three inputs $x_1, x_2, x_3$ and learning coefficient $\eta=0.5.$ Let's say I have two neurons $w_1, w_2$. For each input I will ...
1 vote
2 answers
84 views

What is the proof that non-linearly separable data can't become linearly separable with the results of PCA?

Give a non-linearly separable dataset $X,$ I want to proof that after performing PCA on it, the resulting dataset is guaranteed to be still non-linearly separable. I think we could argue that we still ...
0 votes
0 answers
25 views

How to interpret the loss function $\mathcal{L}$ as i.i.d. random variables?

$\mathcal{D}$ is the fixed but unknown distribution of the data. Usually, this refers to the joint distribution of the input and the label, $$ \begin{aligned} \mathcal{D} &= \mathbb{P}(\...
  • 699
0 votes
0 answers
14 views

Neural Network performs well on Training, Validation, and Testing but cannot do well on completely, unseen new data

Recently built a VGG16 within Tensorflow and I am working on a small dataset. My original dataset is about 2000 images before any sort of splitting or image augmentation. The way that I have tackled ...
-1 votes
0 answers
21 views

Why won't my neural network learn? The accuracy doesn't improve, and the loss doesn't really go down [duplicate]

I am relatively dnew to machine learning, and I have decided to train my own handwritten digit raecognizer. I ahve it training on ceertain images, and it just won't learn. The accuracy stays at 0, ...
  • 9
1 vote
1 answer
52 views

How to generate new data with a VAE?

I have built the following function which takes as input some data and runs a VAE on them: ...
  • 421
0 votes
1 answer
29 views

How to conduct an A/B test on time series

Assuming I have time series for different for some consumers visiting a website. The time series would look something like this: ...
  • 421
0 votes
0 answers
7 views

Best approach for prediction with many (1k) short (12 time points) time series measuring multiple variables

I was wondering what is the best approach to make predictions in a scenario like this: A large set of S (e.g. 1k) short time series Each having T (e.g. 12) time points At each time point t, we have ...
0 votes
0 answers
12 views

In order to use the model on the most recent data, do they have to be cleaned in the same way as the data used to train the model?

I built Machine Learning model based on for example 100 variables which have been previously cleaned. Then I saved my ML model in pickle. Now, I would like to use my ML model to score my clients. And ...
  • 141
0 votes
0 answers
7 views

In a binary classifier, does it make sense to calculate the f-measure for each class separately?

I'm trying to evaluate a binary classifier. I'm not sure if it makes sense to calculate the f1-measure for each class separately or if it would be better to calculate the overall f1-measure. Can you ...
  • 646
1 vote
1 answer
41 views

Why maximize expected reward instead of sum of rewards?

In a typical regression set up, we want to maximize the expected reward(or minimize the expected loss). Empirically, we are maximizing the average return over all samples. However, in reality wouldn't ...
0 votes
1 answer
39 views

Extract the functional mapping between input and output from a machine learning model

A lot of ML models, such as neural networks, are Universal Function Approximators. But when evaluating ML models, we use usually metrics, such as MSE or accuracy, to assess the performance of a ML ...
0 votes
0 answers
14 views

Why is variable importance for boosted trees a squared value?

I am currently trying to understand the tree variable importance calculation as proposed by Breiman; I am coming from this question and will be using its links: Relative variable importance for ...
2 votes
0 answers
31 views

xgboost feature importance vs shap values ranking interpretation

If we have two features, A and B. Feature A has a higher gain than feature B when analyzing feature importance in xgboost with gain. However, when we plot the shap ...
  • 465
0 votes
0 answers
21 views

What are the advantages of model drift vs concept drift in online learning?

Let's say I have a simple linear predictor model and I want to update my model to adapt to the changes that have happened in the environment. I mainly have two tools to detect dynamic changes: Model ...
  • 123
2 votes
2 answers
233 views

Reinforcement learning needs dataset?

Sorry for this dummy question based on the number of contents in the field that I am asking about but It seems that there are tons of texts and videos explaining what ...
  • 127
1 vote
1 answer
13 views

Difference between predcting multiple outputs and single output with random forest

I am trying to predict certain output features (6 in total) with random forest with the input features always being the same. I noticed that my random forest model always fits better when I am trying ...
2 votes
1 answer
27 views

Is bias nothing but perceptron threshold value?

I was revisiting neural network basics from this post. The perceptron follows below equation: $$\begin{align} y & = 1 & \text{if } \sum_{i=1}^n w_i\times x_i \geq \theta \\ & = 0 & \...
  • 151
0 votes
0 answers
31 views

Is there a way to fit a machine learning model to MICE imputed datasets and pool the results?

I have a medical dataset that has a lot of missing values. I imputed five datasets using MICE in R. I want to fit a classification machine learning model to the dataset. I want to identify the most ...
0 votes
0 answers
17 views

Find a projection that minizimizes residuals on multivariate regression

The problem I'm working on is from neuroscience. We've got multiple electrodes with weakly correlated noise all sampling the same system. What are using these to do autoregression on a latent 1D ...
  • 1
0 votes
0 answers
11 views

Why does different bagging fractions lead to the same result?

I am using R to subset a few data samples with a random seed beforehand. However, by trying different values of bagging fraction, the results somehow are the same. The lgbm model (lgbm.mod) is ...
  • 31
1 vote
1 answer
38 views

Should you scale the dataset (normalization or standardization) for a simple multiple logistic regression model?

I have read a lot of conflicting literature about scaling the dataset (using methods such as normalization or standardization) for a multiple logistic regression model, and I am wondering if scaling ...
0 votes
0 answers
11 views

How do the "C" step and robust weighting work in the FASTMCD algorithm for robust covariance estimation?

I am reading up on the FastMCD algorithm [https://arxiv.org/pdf/1709.07045.pdf#page=2] hoping to better understand its implemenation. I think I follow the high level concept for MCD Its basically the ...
2 votes
1 answer
76 views

Justification of the fixed variational distribution in diffusion models

Diffusion models can be regarded as latent variable models (Ho et al., 2020; Section 2), with the latents being an hierarchical chain of random variables $z_T → \dots → z_t → z_{t-1} → \dots → z_1$ (...

1
2 3 4 5
385