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
1answer
24 views

Ensemble Model using Stacking

I learned that building an ensemble model using stacking is done by training a meta-model on the predictions of $n$ other models in order to combine the predictions and try to enhance the performance. ...
2
votes
1answer
63 views

How can I reduce the propagation of errors in multi-step time series forecasting?

I have a multi-step forecasting task where I am predicting values $H$ hours in the future. Supposing that the forecast issue is done at time t, I will produce predictions for the next $H$ hours: $\{\...
2
votes
2answers
27 views

Which predictive models output the posterior distribution?

In a supervised learning context, the posterior distribution of the target given the predictors is often discussed in foundational treatments of the subject. One way this comes up is in decision ...
2
votes
0answers
34 views

Find key nodes in Graph Neural Netwroks

Given a graph dataset, in which links of graphs are the same while features of each node may be varied, how can we locate those critical nodes in this graph structure that contribute the most to ...
0
votes
0answers
29 views

A method that uses machine operation data to predict how many days until this machine will fail

Suppose there is a machine that operates every day, and time series data of its operating conditions can be obtained for each day. In this case, I want to use machine learning or etc. to detect when ...
0
votes
1answer
20 views

How to train on one dataset and deploy on another without loss of performance?

This could be a general question but something that concerns ml4health more. When you train your ML model using one dataset collected from one centre and would like to deploy your model on the data ...
1
vote
1answer
138 views

Proof of upper bound on the leave-one-out-cross-validation error of linear SVM

I have to show that for a two-class SVM classifier trained on a linearly separable dataset $(x_i, y_i)_{i =1}^n$ the following upper bound on the leave-one-out-cross-validation error holds true: $$ ...
2
votes
1answer
18 views

Is early stopping a reasonable method to prevent overfitting in Machine Learning algorithms?

I'm new to Machine learning and have just come across early stopping criteria. I understand it uses a validation set to measure how the accuracy (or any score) improves over iteration (while training)....
1
vote
0answers
26 views

Does always gradients in mini-batch SGD have to be unbiased in order to prove convergence?

I am currently reading this paper [1] and [2]. The authors state that: Our analytical results include almost all of the unbiased compression techniques. And also: (i) gradient compression must be ...
0
votes
0answers
39 views

thresholding prior to model evaluation

Methodology question. The ML textbook approach is this: perform model fit - optimisation assess fit with Cross-Validation tune decision rule by thresholding on the prediction probability (...
0
votes
0answers
15 views

Correlation Feature Selection

I have a huge dataset with a lot of features (approx. 1050) and a target variable. The features can be broken up into groups (7 groups to be exact). My initial approach to feature selection was to ...
0
votes
0answers
11 views

Making predictions from a trained ML model based on incomplete input variables from user

I have a regression-based (xgboost and randomforest) model that is trained using a set of input variables. The model specification looks something like this: ...
1
vote
1answer
46 views

Does a larger sample size increase multi-collinearity between predictors, after imputation of missing data?

I have two datasets that have exactly the same 1701 predictors, but one has 936 subjects and the other has 547 subjects. (The initial rationale for creating these two different datasets was to see ...
1
vote
1answer
20 views

Soft Target in Knowledge Distillation

I am currently reading the paper Distilling the Knowledge in a Neural Network and in the introduction I came across the following sentence - When the soft targets have high entropy, they provide much ...
0
votes
0answers
10 views

Testing a churn model trained on fix period data

A Churn model that needs to predict the customers more likely to churn after 9 months (from the executon time),the model was trained on customer's (retail bank) transactions/ loans 3 years data ...
0
votes
0answers
22 views

Are Federated Learning and Model-Agnostic Meta-Learning the Same Thing?

I am currently reading this paper [1] and [2]. The author makes the claim that their Federated Learning scheme is similar to Model-Agnostic Meta-Learning? They state: Interestingly, FFL is similar to ...
0
votes
0answers
9 views

Learning/Validation/Test sets for parameter tuning

You have $N$ observations of a function $F$ and you want to get an approximation $G$ of the function $F$. The type of model $G$ that you selected depends on a hyperparameter $l$. To determine the ...
0
votes
1answer
9 views

Characterizations of uniformly learnable function classes in the distribution-specific setting

Let $X$ be some input domain (a measurable space). Then let $D$ be some class of probability distributions on $X\times\{0,1\}$. We will call such distributions learning tasks. We say that $D$ is ...
2
votes
0answers
26 views

Intuition of using p(x) (true distribution probability) in KL Divergence definition

We all know that $D(p||q) = \sum_x p(x)log\frac{p(x)}{q(x)}$ and it is used to quantify the difference between the true distribution p and the observed distribution q. However, I do not get the ...
2
votes
0answers
23 views

Measure-theoretically rigorous treatment of statistical learning theory

My main source on statistical learning theory has been Shwartz/Ben-David. This is a good book but it's a little vague from a measure-theoretic point of view. For example, in the definition of PAC ...
0
votes
0answers
17 views

Why are latent spaces able to learn representations - autoencoder?

As the title states, why are latent spaces even able to intelligently learn representations? There's no guarantee that we learn the most important features since it's all done automatically in ...
0
votes
1answer
15 views

actor update in DDPG algorithm (and in general actor-critic algorithms)

The update equations for the parameters of the actor and the critic are: $$ \delta_t = r_t+\gamma Q^\omega (x_{t+1},a_{t+1})-Q^\omega(x_t,a_t)$$ $$ \omega_{t+1} = \omega_t+\alpha_\omega \delta_t ...
2
votes
1answer
33 views

Stochastic Gradient Descent Code Check for Least Squares

I have an R-based implementation of the gradient descent and am trying to also get it to work as SGD. The function matches R's lm function when using the entire data set. But, when I sample from the ...
1
vote
1answer
81 views

Target population for power analysis of ML model A/B test

We are working on an ML model that predicts a numeric result (call it $\hat{x}$). Eventually, we will perform an A/B test, where the metric is a function that takes $\hat{x}$ as an input (call it $f(\...
0
votes
0answers
8 views

Which hypthesis Test for Recall and Jaccard Index on same model

I cross validated a semantic segmentation network on a medical data set to segment and diagnose cells. Now I want to see if my model has better performance if I just evaluate the segmentation of the ...
0
votes
0answers
18 views

Encoding hierarchical relations between samples for machine learning

I have a set of samples with a corresponding set of continuous features which I am using to make predictions about a particular property of the samples. However, the samples are organized in a ...
2
votes
1answer
44 views

Can a classifier be used to improve its own training data?

Introduction: Consider a classification problem of $\mathbb{R}^n$ into $\mathbb{R}^2$. Let $\mathcal{U}$ be a set of instances whose class is unknown, but can be discovered paying a cost $\gamma$ for ...
1
vote
1answer
10 views

What is the best way to validate whether a set of given scores are a good predictor of a success metric?

For example, I have a table of items each with the given scores (no unit) and success metric (quantity value). The success metric is the real-life performance indicating number of items sold, etc. ...
0
votes
0answers
10 views

Does Guo's "On Calibration of Modern Neural Networks" Miss the Probabilities of the Non-Dominant Class?

The gist behind Harrell's rms::calibrate function makes sense to me. While I have yet to understand the magic that lets us calculate the "true" ...
0
votes
0answers
20 views

In Probabilistic Machine Learning, how that the parameters are identifiable works on the asymptotic normality of the sampling distribution

How that the parameters are identifiable works on the asymptotic normality of the sampling distribution
0
votes
0answers
11 views

When do we use stratified splitting?

I have a problem with my data. I only have 59 recordings and that is not optimal for a machine learning regression pipeline. Anyway, I have multiple independent variables, most of them are in this ...
0
votes
0answers
33 views

Applying SMOTE multiple times?

More of a curiosity, but I'm currently learning how to deal with imbalanced datasets and came across the SMOTE method to bias the minority class. The images below show before and after SMOTE was ...
0
votes
0answers
8 views

Display inverted ROC plot

my anomaly detection algorithm gave me an array of predictions where all the values greater than 0 should be of the positive class (= 0) and all the other should be classified as anomalies (= 1). I ...
0
votes
0answers
33 views

Combining bootstrapping and cross validation for predicting sensitivity to drugs in Barretina et al., 2012

Also posted on https://discourse.datamethods.org/t/combining-bootstrapping-and-cross-validation-for-predicting-sensitivity-to-drugs-in-barretina-et-al-2012/5093. Was curious for any additional inputs/...
1
vote
0answers
26 views

Resources on on-line machine learning

I am wondering if there are any books/articles/tutorials about "on-line machine learning"? For example, this website has nice lecture notes (from lec16) on some of the aspects: https://web....
2
votes
1answer
18 views

Boosting reduces bias when compared to what algorithm?

I am reading on bagging and boosting, and I understand how they both work (at least I think I do). I would like to talk in the context of decision tree ensembles as I think (not sure if correct) that ...
0
votes
0answers
13 views

Time steps in RNN and LSTM

I am quite new to recurrent neural networks and how to use them for sequence classification. I was wondering if anyone could shed some light on how RNNs (specifically LSTMs) capture time. That is, can ...
0
votes
1answer
27 views

When is Non-Max Suppression used in Object Detection

Is non-max suppression for bounding boxes obtained from a Region Proposal Network performed during training? From what I gather, NMS is not differentiable-- in which case, it can't be performed during ...
1
vote
2answers
28 views

What would be a reason to use the Root Mean Square Error (RMSE) to combine data?

In this machine learning tutorial by Google they use the Root Mean Square to create a similarity measure between two shoes. They first calculate the difference between the size of two shoes and then ...
0
votes
0answers
19 views

Are graphical neural networks the right approach for isomorphic graphs?

I have a set of $N$ observations ($N>100,000$): each observation takes the form of a homogeneous, undirected graph $G_n=(V,E)$ all graphs $G_n$ have the same nodes and edges - around 5,000 nodes ...
1
vote
1answer
14 views

Random state value changes the results of rmse and R2

I want to know why everytime I run my algorithm (XGBoost regressor) with a different random state (applied to train/test split part) I get different values for R2 and RMSE. For example : Random state ...
0
votes
0answers
18 views

Constructing a User Profile for Music Taste

My goal is to construct User profiles based on positive (and maybe also negative) interactions with songs. A User has the option to like a song. This would give me a list of likes for each user. With ...
0
votes
0answers
16 views

How to improve the PMI (Pointwise Mutual Information) Quality for document based PMI

Generating word embeddings from the PMI is well understood and known to be equivalent to SGNS (skipgram negative-sampling) under certain conditions. I was able to get good quality word embedding using ...
0
votes
1answer
14 views

How to use test sample weights for prediction in XGBoost regression

I have an highly imbalanced dataset where very few y values are 'out of norm'. I want to predict as close as possible to these 'out of norm' values for those observations. For this I am trying to ...
0
votes
0answers
10 views

ROSE acceptable dispersion/shrinkage

To solve imbalanced data, I used oversampling strategy using ROSE algorithm in Python. As you may know, ROSE is a smoothed bootstrapping method and we can control the dispersion of the augmented data. ...
0
votes
0answers
49 views

Predictor With Lower Mean Absolute Error Ends Up Worse

I have been recently working on a problem to estimate the ETAs of vehicles using ensemble techniques such as LightGBM. As expected, the distance taken by the vehicle's route to its destination is a ...
0
votes
0answers
8 views

Building a ranking model, using linear regression with manually updated inputs by end users

I am trying to solve a ranking problem and starting from a linear regression here. As a dependent variable I currently have the score of different authors in academic literature and want to convert ...
1
vote
1answer
28 views

700,000 data points are true values while 1,300,000 data points are false values, is it an imbalanaced dataset? [closed]

I am trying to build a decision tree model and I have 700,000 true values while 1,300,000 data points are false values, in total, I have 2,000,000 data points including duplicates. I am wondering if ...
0
votes
0answers
10 views

Look ahead bias by standardization of a time series in predictive models?

I'm using some machine learning models to predict future values of a time series (stock returns). In the data preprocessing step I'm standardizing all variables (incl. the target variable) using ...
1
vote
0answers
71 views

Can Missing Data ever be considered as "Informative"?

I have always had the following question: Can Missing Data ever be considered as "Informative"? Often we view missing data as problematic, and either discard missing data or try to use ...