Happy to join this community. Thank you in advance for your kind help! :) Intro I have a physical device characterized by its internal parameters, of which I know the nominal values. I also have the theoretical model of the device, that differs from the physical device because fabrication tolerances change the internal parameters. I would like to extract the internal parameters of the device by fitting the model onto it. The device also has additional inputs that alter its …

Say we have a trained image classification model. Theoretically, is it possible to update the model with only a sample without retraining? If not, is there any kinds of active image classification DL model that can allow incremental update?

what about difference between the meta learning and semi-supervised learning and self-supervised learning and active learning and federated learning and few-shot learning? in application and in definition? Pros and cons?

I have the task of coming up with a model of 95% accuracy for a classification problem. I have training data and a hold-out data set. I have the opportunity to request data of a particular class with desired characteristics to achieve this objective. What method shall I use to plan the data acquisition through another team? I am currently at 86% accuracy. I use LightGBM for the model development. Would consider parameter tuning and ensemble with XGBoost and TabNet. …

During my research in the active learning field, I found a similar concept that has the same idea that is the optimal experimental design (OED) for machine learning which is based on finding new data points to do experiments on in order to improve the performance of our model. This made me wonder if OED is a subfield from active learning or it is completely different. Any information will be useful and appreciated. Thank you.

I am trying to apply active learning to a model to improve its performance. however the oracle, cannot label the samples based on the input space's features, instead, it uses the output sample(label) to do the experiment and get the data point. I am wondering if it would be possible to make the query strategy ask for points by the label and query its corresponding input values?

I have known Active Learning for a while and I am wondering what is the difference between the well known AL, and Continuous Learning? Is this a new buzzword for the old approach or is there anything new to it on top of AL?

This video https://youtu.be/60Sk-mq3Cr8, from 0 to 2:00 minutes, mentioned an object detection model that can train on 10 samples and improved over time, what kinds of the latest model out there that can achieve this?

Case 1: I would apply active learning to query a small chunk of samples gradually to label them and my model is being trained during this process. After a certain number of iterations, I have a training dataset with specific performance of the model. Case 2: I re-train the model from scratch with the training dataset in case 1. Question 1: do you think the performance of the model will be the same in both cases? why, please? Question 2: …

How can I make sure that the initial model trained on a small dataset will not suffer from overfitting before applying the active learning sampling techniques? because I will use this model to select new unlabeled samples.

With active learning I hope to keep the annotation effort to a minimum, yet building still a good classifier. My initial starting point is that I have about 20k images which can belong to ten different classes, and have 0 labeled images at the moment. After each active learning iteration, I hope to get the labels of e.g. 100 images. If it matters, unfortunately, the data is very likely imbalanced which means that five classes are probably very rare. So …

I'm doing some active learning with uncertainty sampling on a self-attention model implemented in PyTorch. The algorithm works as follows (steps 3-7 are repeated for 14 iterations): 1. Take 10% of the data as training set, L 2. Train the model on L 3. Either rank the remaining samples U by a certain informativeness measure and pick a batch B of the top n samples, or randomly pick a batch B 4. Add B to L 5. Remove B from …

Suppose I have a dataset of people's phone numbers and heights, and I'm interested in learning the parameters $p_{girl}$, $p_{boy}=1-p_{girl}$, $\mu_{boy}$, $\mu_{girl}$, and overall $\sigma$ governing the distribution of peoples' heights. I don't have labels for boys or girls yet, but if I really want to, I can call the phone number and ask if the person is a boy or girl. Procedure: Fit a Gaussian mixture model to heights via EM. Assign the greater of the $\mu$s to be …

In a lot of cases unlabelled data needs to be transformed to labelled data. The best solution is to use (multiple) human classifiers. However, going to all the data by hand (i.e. in text-mining or image-processing) is often a daunting task. Is there software that can combine human classifiers and machine-learning techniques in real time? I am especially interested in python packages. To illustrate, classifying images from video streams is very repetitive. After 100 images (from different streams) a machine-learning …

I read the book "Human-in-the-Loop Machine Learning" by Robert (Munro) Monarch about Active Learning. I don't understand the following approach to get a diverse set of items for humans to label: Take each item in the unlabeled data and count the average number of word matches it has with items already in the training data Rank the items by their average match Sample the item with the lowest average number of matches Add that item to the ‘labeled’ data and …

I am dabbling in active learning and was wondering how to combine this in seeking out the best architecture for the network. In my understanding, active learning uses a heuristic for selecting the best instances to label in order to learn as quickly as possible. However, the way these instances are chosen are dependent on the model itself. Is there a way to handle this model dependency? It seems to me that models architecture is dependent on the train size, …

From Wikipedia: Active learning is a special case of machine learning in which a learning algorithm can interactively query a user (or some other information source) to label new data points with the desired outputs. Reinforcement learning (RL) is an area of machine learning concerned with how software agents ought to take actions in an environment in order to maximize the notion of cumulative reward. How to distinguish them? What are the exact differences?

I have data where companies ask users to score a bunch of questions but some items may be randomly chosen while others are personalized. Users score higher in personalized questions on average. I have a user ID, question ID, corresponding score of the question by the user, and whether the question is random or personalized. I want to build a recommendation system that incorporates the feature of a question being random or personalized. I assume that for a personalized item …

Let's say we have a set of data points that need to be labelled for a classification task. In the pool-based active learning, if we go with the uncertainty measure, is the AL approach able to detect challenging cases? By challenging cases I mean samples that receive a high prediction score for $\hat{y}$ (e.g. >90%) but, most probably, $\neg\hat{y}$ is the correct prediction. The rationale behind my question is: does adding more samples to the training set always improve the …

I would like to label character data with BIO tags as part of an active learning process on unlabelled data. I am assuming there are open source GUI tools available which I can use to make this easier - i.e. present the string to be labeled and some way of tagging characters from a predefined set of tags (and probably allow new tags to be added). I have not been able to find anything though - ideally cross-platform (Linux and …