#classifier — Public Fediverse posts

@phpmacher #RSS nutzen: Sicher. Immer noch populär. Ob als Eingangsformat für andere Medien oder direkte Nutzung durch Menschen.
Gibt diverse Systeme die lokal aggregieren und filtern, auch self-hostable. Teilweise über die eingebetteten #Metadaten, teilweise mit #Regex gegen andere Inhaltsfelder.
Der Schritt von da zu einer Art #Bayesfilter oder den NN wie sie #rspamd zur Klassifizierung nutzt oder sogar einem SLM/kleineren #LLM auf #Ollama, ggf. mit #TPU / #APU Support sollte überschaubar groß sein.
Eins der Probleme wird aber die zunehmend schlechte Feedqualität was Tags/Metadaten und der notorische (wenn gleich sehr verständliche) Hang zur Nicht-Auslieferung des Volltext im Feed sein. Da müsste man ggf auf Verdacht das Original fetchen, ggf. als zweite Stufe.

Und natürlich ist ein #Recommender-System was deutlich anderes als ein mehr oder minder ausgefeilter #Classifier. Aber lassen wir das...

'Consistent Multiclass Algorithms for Complex Metrics and Constraints', by Harikrishna Narasimhan et al.

http://jmlr.org/papers/v25/22-1137.html

#multiclass #classifier #classification

'Non-splitting Neyman-Pearson Classifiers', by Jingming Wang, Lucy Xia, Zhigang Bao, Xin Tong.

http://jmlr.org/papers/v25/22-0795.html

#classifiers #classifier #classification

'Regimes of No Gain in Multi-class Active Learning', by Gan Yuan, Yunfan Zhao, Samory Kpotufe.

http://jmlr.org/papers/v25/23-0234.html

#classifier #classification #classes

How would you go about creating a filter that blocks posts about things that people hate?

I've thought I could build a text classifier, but it could be hard to train since I'd need to guess whether or not the author hates the thing they are posting about.

I wouldn't want it to become a filter for all current events news, but I suspect that's what it would become.

#fediverse #mastodon #machineLearning #tfidf #classification #socialMedia #classifier #textAnalysis #programming #tech #technology

'Multi-class Probabilistic Bounds for Majority Vote Classifiers with Partially Labeled Data', by Vasilii Feofanov, Emilie Devijver, Massih-Reza Amini.

http://jmlr.org/papers/v25/23-0121.html

#classifiers #classifier #labeling

'On the Learnability of Out-of-distribution Detection', by Zhen Fang, Yixuan Li, Feng Liu, Bo Han, Jie Lu.

http://jmlr.org/papers/v25/23-1257.html

#learnability #classifier #detection

'A Multilabel Classification Framework for Approximate Nearest Neighbor Search', by Ville Hyvönen, Elias Jääsaari, Teemu Roos.

http://jmlr.org/papers/v25/23-0286.html

#classification #classifiers #classifier

Except for naming the shapes I figure something like following might work?

- Render each Unicode script in several fonts.

- Training per-script classifiers (OCRs) on those.

- Run the per-script classifiers on each others training data to identify homographs.

There's a tricky bits, like creating non-confusable out-of-script scribbles to include in training data, etc.

Still... this doesn't seem like all that impossible?

What am I missing?

#unicode #homography #ml #classifier

'On Truthing Issues in Supervised Classification', by Jonathan K. Su.

http://jmlr.org/papers/v25/19-301.html

#classification #classifier #supervised

Делаем intent classifier для службы поддержки без доменного датасета на русском

В этой статье я продемонстрирую, как без собственного датасета сделать классификатор намерений пользователя для службы поддержки в сфере e-commerce. И более того, я расскажу, как у меня получилось сделать классификатор для русского языка без датасета на русском языке. Меня зовут Елизавета Колмакова, я Data Scientist в компании, которая разрабатывает айти-решения для крупного ритейла.

https://habr.com/ru/articles/792542/

#intent_recognition #intent #encoder #служба_поддержки #no_dataset #classifier #telegrambot #nlp #bot

Делаем intent classifier для службы поддержки без доменного датасета на русском

В этой статье я продемонстрирую, как без собственного датасета сделать классификатор намерений пользователя для службы поддержки в сфере e-commerce. И более того, я расскажу, как у меня получилось сделать классификатор для русского языка без датасета на русском языке. Меня зовут Елизавета Колмакова, я Data Scientist в компании, которая разрабатывает айти-решения для крупного ритейла.

https://habr.com/ru/articles/792542/

#intent_recognition #intent #encoder #служба_поддержки #no_dataset #classifier #telegrambot #nlp #bot

'Set-valued Classification with Out-of-distribution Detection for Many Classes', by Zhou Wang, Xingye Qiao.

http://jmlr.org/papers/v24/23-0712.html

#classifier #classification #classes

'MAUVE Scores for Generative Models: Theory and Practice', by Krishna Pillutla et al.

http://jmlr.org/papers/v24/23-0023.html

#generative #classifier #divergences

'MAUVE Scores for Generative Models: Theory and Practice', by Krishna Pillutla et al.

http://jmlr.org/papers/v24/23-0023.html

#generative #classifier #divergences

'MAUVE Scores for Generative Models: Theory and Practice', by Krishna Pillutla et al.

http://jmlr.org/papers/v24/23-0023.html

#generative #classifier #divergences

'MAUVE Scores for Generative Models: Theory and Practice', by Krishna Pillutla et al.

http://jmlr.org/papers/v24/23-0023.html

#generative #classifier #divergences

'MAUVE Scores for Generative Models: Theory and Practice', by Krishna Pillutla et al.

http://jmlr.org/papers/v24/23-0023.html

#generative #classifier #divergences

I already have a Mastodon client (https://schizo.social) so I'm already on my way!

I want to pull the user's favorites and run them through a #machineLearning #classifier of some sort.

Then when I pull their Timeline I can compare each post to their favorites and determine a similarity score.

I envision a sort of slider that lets them set a threshold of "favoriteness" to filter by; posts that are unlikely to be favorites will be hidden.

'Erratum: Risk Bounds for the Majority Vote: From a PAC-Bayesian Analysis to a Learning Algorithm', by Louis-Philippe Vignault, Audrey Durand, Pascal Germain.

http://jmlr.org/papers/v24/22-0747.html

#bound #classifier #majority

I think what I'll do is tally up the words in every toot and then use tf-idf to calculate a match score for each toot. Then I can specify a threshold that will show/hide toots in my timeline based on these.

This will require manual labeling to train the classifier, but I think it should work.

https://en.wikipedia.org/wiki/Tf%E2%80%93idf

#ml #classification #classifier #machinelearning

'Leaky Hockey Stick Loss: The First Negatively Divergent Margin-based Loss Function for Classification', by Oh-Ran Kwon, Hui Zou.

http://jmlr.org/papers/v24/22-1104.html

#classifier #classification #svm

'Random Forests for Change Point Detection', by Malte Londschien, Peter Bühlmann, Solt Kovács.

http://jmlr.org/papers/v24/22-0512.html

#changeforest #classifier #classifiers

Mitigating Confirmation Bias in Semi-supervised Learning via Efficient Bayesian Model Averaging

Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj et al.

Action editor: Frederic Sala.

https://openreview.net/forum?id=PRrKOaDQtQ

#supervised #labeling #classifier

I have plenty more achievable goals for https://schizo.social (like multi-account, or #Calckey support) but something I'd love to try is #classifying posts with #machineLearning #tfidf

I'd like to be able to define "labels" and then train it to identify those on the fly. Then either mute or highlight posts that #classify highly.

Not so much an #algorithm, as a #filter.

#ai #ml #webDev #classifier

I have plenty more achievable goals for https://schizo.social (like multi-account, or #Calckey support) but something I'd love to try is #classifying posts with #machineLearning #tfidf

I'd like to be able to define "labels" and then train it to identify those on the fly. Then either mute or highlight posts that #classify highly.

Not so much an #algorithm, as a #filter.

#ai #ml #webDev #classifier

I have plenty more achievable goals for https://schizo.social (like multi-account, or #Calckey support) but something I'd love to try is #classifying posts with #machineLearning #tfidf

I'd like to be able to define "labels" and then train it to identify those on the fly. Then either mute or highlight posts that #classify highly.

Not so much an #algorithm, as a #filter.

#ai #ml #webDev #classifier

I have plenty more achievable goals for https://schizo.social (like multi-account, or #Calckey support) but something I'd love to try is #classifying posts with #machineLearning #tfidf

I'd like to be able to define "labels" and then train it to identify those on the fly. Then either mute or highlight posts that #classify highly.

Not so much an #algorithm, as a #filter.

#ai #ml #webDev #classifier

I have plenty more achievable goals for https://schizo.social (like multi-account, or #Calckey support) but something I'd love to try is #classifying posts with #machineLearning #tfidf

I'd like to be able to define "labels" and then train it to identify those on the fly. Then either mute or highlight posts that #classify highly.

Not so much an #algorithm, as a #filter.

#ai #ml #webDev #classifier

Invariant Feature Coding using Tensor Product Representation

YUSUKE Mukuta, Tatsuya Harada

Action editor: Seungjin Choi.

https://openreview.net/forum?id=uv32JOdQuh

#tensor #representations #classifier

Predicting Out-of-Domain Generalization with Neighborhood Invariance

Nathan Hoyen Ng, Neha Hulkund, Kyunghyun Cho, Marzyeh Ghassemi

Action editor: Vincent Dumoulin.

https://openreview.net/forum?id=jYkWdJzTwn

#generalization #classifier #classification

New #SurveyCertification:

On Averaging ROC Curves

Jack Hogan, Niall M. Adams

https://openreview.net/forum?id=FByH3qL87G

#classifiers #classifier #roc

On Averaging ROC Curves

Jack Hogan, Niall M. Adams

Action editor: Hsuan-Tien Lin.

https://openreview.net/forum?id=FByH3qL87G

#classifiers #classifier #roc

Finding Competence Regions in Domain Generalization

Jens Müller, Stefan T. Radev, Robert Schmier, Felix Draxler, Carsten Rother, Ullrich Koethe

Action editor: Hanwang Zhang.

https://openreview.net/forum?id=TSy0vuwQFN

#classifiers #accuracy #classifier

Assuming Locally Equal Calibration Errors for Non-Parametric Multiclass Calibration

Kaspar Valk, Meelis Kull

Action editor: Aditya Menon.

https://openreview.net/forum?id=na5sHG69rI

#classifiers #classifier #calibration

'Risk Bounds for Positive-Unlabeled Learning Under the Selected At Random Assumption', by Olivier Coudray, Christine Keribin, Pascal Massart, Patrick Pamphile.

http://jmlr.org/papers/v24/22-067.html

#labeled #classification #classifier

Bridging performance gap between minimal and maximal SVM models

Ondrej Such, René Fabricius

https://openreview.net/forum?id=SM1BkjGePI

#svm #svms #classifier

Bridging performance gap between minimal and maximal SVM models

Ondrej Such, René Fabricius

https://openreview.net/forum?id=SM1BkjGePI

#svm #svms #classifier

Bridging performance gap between minimal and maximal SVM models

Ondrej Such, René Fabricius

https://openreview.net/forum?id=SM1BkjGePI

#svm #svms #classifier

Bridging performance gap between minimal and maximal SVM models

Ondrej Such, René Fabricius

https://openreview.net/forum?id=SM1BkjGePI

#svm #svms #classifier

Bridging performance gap between minimal and maximal SVM models

Ondrej Such, René Fabricius

https://openreview.net/forum?id=SM1BkjGePI

#svm #svms #classifier

Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logist...

Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann

https://openreview.net/forum?id=jM8nzUzBWr

#classification #classify #classifier

Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logist...

Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann

https://openreview.net/forum?id=jM8nzUzBWr

#classification #classify #classifier

Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logist...

Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann

https://openreview.net/forum?id=jM8nzUzBWr

#classification #classify #classifier

Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logist...

Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann

https://openreview.net/forum?id=jM8nzUzBWr

#classification #classify #classifier

Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logist...

Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann

https://openreview.net/forum?id=jM8nzUzBWr

#classification #classify #classifier

KRADA: Known-region-aware Domain Alignment for Open-set Domain Adaptation in Semantic Segmentation

Chenhong Zhou, Feng Liu, Chen Gong, Rongfei Zeng, Tongliang Liu, William Cheung, Bo Han

https://openreview.net/forum?id=5II12ypVQo

#classify #labeled #classifier

KRADA: Known-region-aware Domain Alignment for Open-set Domain Adaptation in Semantic Segmentation

Chenhong Zhou, Feng Liu, Chen Gong, Rongfei Zeng, Tongliang Liu, William Cheung, Bo Han

https://openreview.net/forum?id=5II12ypVQo

#classify #labeled #classifier

KRADA: Known-region-aware Domain Alignment for Open-set Domain Adaptation in Semantic Segmentation

Chenhong Zhou, Feng Liu, Chen Gong, Rongfei Zeng, Tongliang Liu, William Cheung, Bo Han

https://openreview.net/forum?id=5II12ypVQo

#classify #labeled #classifier

KRADA: Known-region-aware Domain Alignment for Open-set Domain Adaptation in Semantic Segmentation

Chenhong Zhou, Feng Liu, Chen Gong, Rongfei Zeng, Tongliang Liu, William Cheung, Bo Han

https://openreview.net/forum?id=5II12ypVQo

#classify #labeled #classifier