... Repräsentationen wieder in Tokens überführt. Die Vektoren selbst sind für Menschen nicht unmittelbar interpretierbar. Genau hier setzen Verfahren wie Sparse Autoencoder an.
4/4
4/4
November 4, 2026 at 11:30 PM
... Repräsentationen wieder in Tokens überführt. Die Vektoren selbst sind für Menschen nicht unmittelbar interpretierbar. Genau hier setzen Verfahren wie Sparse Autoencoder an.
4/4
4/4
ASCENSION: Autoencoder-Based Latent Space Class Expansion for Time Series Data Augmentation
Dorian Joubaud, Matthieu OLEKHNOVITCH, Adrien Bolling et al.
Action editor: Markus Lange-Hegermann
https://openreview.net/forum?id=7KFXjf7K2o
#autoencoders #generative #autoencoder
Dorian Joubaud, Matthieu OLEKHNOVITCH, Adrien Bolling et al.
Action editor: Markus Lange-Hegermann
https://openreview.net/forum?id=7KFXjf7K2o
#autoencoders #generative #autoencoder
November 4, 2026 at 9:21 PM
ASCENSION: Autoencoder-Based Latent Space Class Expansion for Time Series Data Augmentation
Dorian Joubaud, Matthieu OLEKHNOVITCH, Adrien Bolling et al.
Action editor: Markus Lange-Hegermann
https://openreview.net/forum?id=7KFXjf7K2o
#autoencoders #generative #autoencoder
Dorian Joubaud, Matthieu OLEKHNOVITCH, Adrien Bolling et al.
Action editor: Markus Lange-Hegermann
https://openreview.net/forum?id=7KFXjf7K2o
#autoencoders #generative #autoencoder
... durch verteilte und überlagerte Features (Merkmale) repräsentieren und wie Sparse Autoencoder (SAEs) versuchen, diese internen Repräsentationen in einzelne Features zu zerlegen, deren Bedeutung sich anschließend mit menschlicher Sprache und Begriffen beschreiben lässt...
2/...
2/...
November 4, 2026 at 11:30 PM
... durch verteilte und überlagerte Features (Merkmale) repräsentieren und wie Sparse Autoencoder (SAEs) versuchen, diese internen Repräsentationen in einzelne Features zu zerlegen, deren Bedeutung sich anschließend mit menschlicher Sprache und Begriffen beschreiben lässt...
2/...
2/...
Etwas trocken, aber für mein Thema wichtig. Eigentlich ist es generell relevant, da darauf viele weitere Forschungsansätze aufbauen. Die Arbeit veranschaulicht, wie neuronale Netzwerke Informationen in hochdimensionalen Aktivierungsräumen ...
1/...
transformer-circuits.pub/2023/monosem...
1/...
transformer-circuits.pub/2023/monosem...
November 4, 2026 at 11:30 PM
Etwas trocken, aber für mein Thema wichtig. Eigentlich ist es generell relevant, da darauf viele weitere Forschungsansätze aufbauen. Die Arbeit veranschaulicht, wie neuronale Netzwerke Informationen in hochdimensionalen Aktivierungsräumen ...
1/...
transformer-circuits.pub/2023/monosem...
1/...
transformer-circuits.pub/2023/monosem...
New paper in Imaging Neuroscience by Anton Orlichenko, Yize Zhao, et al:
Intervention on modifiable lifestyle and physiological factors via variational autoencoder reveals changes in functional connectivity-mediated risk for Alzheimer’s disease
doi.org/10.1162/IMAG...
Intervention on modifiable lifestyle and physiological factors via variational autoencoder reveals changes in functional connectivity-mediated risk for Alzheimer’s disease
doi.org/10.1162/IMAG...
November 5, 2026 at 10:26 PM
New paper in Imaging Neuroscience by Anton Orlichenko, Yize Zhao, et al:
Intervention on modifiable lifestyle and physiological factors via variational autoencoder reveals changes in functional connectivity-mediated risk for Alzheimer’s disease
doi.org/10.1162/IMAG...
Intervention on modifiable lifestyle and physiological factors via variational autoencoder reveals changes in functional connectivity-mediated risk for Alzheimer’s disease
doi.org/10.1162/IMAG...
Athanasios Babouras, Paul A. Martineau, Thomas Fevens: Evaluating Anterior Cruciate Ligament Risk of Injury Using a Graph Convolutional Autoencoder https://arxiv.org/abs/2611.09724 https://arxiv.org/pdf/2611.09724 https://arxiv.org/html/2611.09724
November 9, 2026 at 6:41 AM
Athanasios Babouras, Paul A. Martineau, Thomas Fevens: Evaluating Anterior Cruciate Ligament Risk of Injury Using a Graph Convolutional Autoencoder https://arxiv.org/abs/2611.09724 https://arxiv.org/pdf/2611.09724 https://arxiv.org/html/2611.09724
Der Verdienst dieser Arbeit war es, von den Aktivierungen zu getunten Linsen überzugehen. Übrigens, die Sparse Autoencoder (SAE) aus einer früheren Vorlesung sind ein exzellentes Werkzeug der Mechanistic Interpretability, allerdings skalieren sie nicht gut. 12/14
November 9, 2026 at 6:05 PM
Der Verdienst dieser Arbeit war es, von den Aktivierungen zu getunten Linsen überzugehen. Übrigens, die Sparse Autoencoder (SAE) aus einer früheren Vorlesung sind ein exzellentes Werkzeug der Mechanistic Interpretability, allerdings skalieren sie nicht gut. 12/14
Happy to introduce PULSE, our generative AI method for chemically disordered materials. Without any pre-existing dataset, an inverse variational autoencoder learns to estimate the partition function, the key to a material’s thermodynamic properties. 1/3
November 7, 2026 at 8:42 AM
Happy to introduce PULSE, our generative AI method for chemically disordered materials. Without any pre-existing dataset, an inverse variational autoencoder learns to estimate the partition function, the key to a material’s thermodynamic properties. 1/3
Robust detection of Chagas disease via spatio-temporal masked autoencoder pubmed.ncbi.nlm.nih.gov/42970825/
November 3, 2026 at 10:38 PM
Robust detection of Chagas disease via spatio-temporal masked autoencoder pubmed.ncbi.nlm.nih.gov/42970825/
#10 weiter zur Geometrie der neuronalen Aktivierungen von LLM und insbesondere zum Brot-und-Butter-Tool der Mechanistic Interpretability: Sparse Autoencoder (SAE). Im Wesentlichen geht es um die mathematische Frage, ... 1/19
November 3, 2026 at 6:04 PM
#10 weiter zur Geometrie der neuronalen Aktivierungen von LLM und insbesondere zum Brot-und-Butter-Tool der Mechanistic Interpretability: Sparse Autoencoder (SAE). Im Wesentlichen geht es um die mathematische Frage, ... 1/19
Xiaoling Han, Bin Lin, Nan Wu, Tony Q. S. Quek: Design of a Speech Enhancement-Based Deep Semantic Autoencoder for Wireless Communications https://arxiv.org/abs/2611.07361 https://arxiv.org/pdf/2611.07361 https://arxiv.org/html/2611.07361
November 6, 2026 at 6:44 AM
Xiaoling Han, Bin Lin, Nan Wu, Tony Q. S. Quek: Design of a Speech Enhancement-Based Deep Semantic Autoencoder for Wireless Communications https://arxiv.org/abs/2611.07361 https://arxiv.org/pdf/2611.07361 https://arxiv.org/html/2611.07361
4/ It works. Eight generated examples fire a held out feature from a sparse autoencoder 89 percent of the time, and beat the best corpus example on 30 percent of features. Trained only on English, it still beats corpus search in 12 of 21 non English domains.
November 8, 2026 at 1:58 AM
4/ It works. Eight generated examples fire a held out feature from a sparse autoencoder 89 percent of the time, and beat the best corpus example on 30 percent of features. Trained only on English, it still beats corpus search in 12 of 21 non English domains.
Thank you to Irene Chang (Allen Lab) and Kevin Hoffer-Hawlik (Azizi/Figueroa Labs) for presenting at this month’s research meeting! They presented:
📉 MEDAL: Manifold Distillation via AutoEncoder Learning
🧬 CORGI (COmbinatorial Response and Gene dependency Inference) in Triple-Negative Breast Cancer
📉 MEDAL: Manifold Distillation via AutoEncoder Learning
🧬 CORGI (COmbinatorial Response and Gene dependency Inference) in Triple-Negative Breast Cancer
November 5, 2026 at 2:21 PM
Thank you to Irene Chang (Allen Lab) and Kevin Hoffer-Hawlik (Azizi/Figueroa Labs) for presenting at this month’s research meeting! They presented:
📉 MEDAL: Manifold Distillation via AutoEncoder Learning
🧬 CORGI (COmbinatorial Response and Gene dependency Inference) in Triple-Negative Breast Cancer
📉 MEDAL: Manifold Distillation via AutoEncoder Learning
🧬 CORGI (COmbinatorial Response and Gene dependency Inference) in Triple-Negative Breast Cancer
A brain chart is only as good as the space it measures distances in. In autoencoders, that space is often a bit lumpy.
At MICCAI 2026, Mariam Zabihi (@mariam-zabihi.bsky.social) presents SIGMA-VAE, which irons it out.
Paper & code: https://papers.miccai.org/miccai-2026/0960-Paper3072.html
At MICCAI 2026, Mariam Zabihi (@mariam-zabihi.bsky.social) presents SIGMA-VAE, which irons it out.
Paper & code: https://papers.miccai.org/miccai-2026/0960-Paper3072.html
November 6, 2026 at 6:23 AM
A brain chart is only as good as the space it measures distances in. In autoencoders, that space is often a bit lumpy.
At MICCAI 2026, Mariam Zabihi (@mariam-zabihi.bsky.social) presents SIGMA-VAE, which irons it out.
Paper & code: https://papers.miccai.org/miccai-2026/0960-Paper3072.html
At MICCAI 2026, Mariam Zabihi (@mariam-zabihi.bsky.social) presents SIGMA-VAE, which irons it out.
Paper & code: https://papers.miccai.org/miccai-2026/0960-Paper3072.html
Efficient and Scalable Multi-View Masked Autoencoder for Heterogeneous Graph Self-Supervised Learning
👥 Hao Geng, Deqing Wang, Lihao Ge et al.
#MetaPathViewMasking #MaskedViewReconstruction #TransformerViewFusion #SparseDenseAggregation
🔗 trendtoknow.com/papers/2611....
👥 Hao Geng, Deqing Wang, Lihao Ge et al.
#MetaPathViewMasking #MaskedViewReconstruction #TransformerViewFusion #SparseDenseAggregation
🔗 trendtoknow.com/papers/2611....
November 4, 2026 at 9:01 AM
Efficient and Scalable Multi-View Masked Autoencoder for Heterogeneous Graph Self-Supervised Learning
👥 Hao Geng, Deqing Wang, Lihao Ge et al.
#MetaPathViewMasking #MaskedViewReconstruction #TransformerViewFusion #SparseDenseAggregation
🔗 trendtoknow.com/papers/2611....
👥 Hao Geng, Deqing Wang, Lihao Ge et al.
#MetaPathViewMasking #MaskedViewReconstruction #TransformerViewFusion #SparseDenseAggregation
🔗 trendtoknow.com/papers/2611....
The paper introduces Multi-View Masked Auto Encoder, which masks entire meta-path views and reconstructs their representations from remaining views.
Source: arXiv cs.LG
Source: arXiv cs.LG
Efficient and Scalable Multi-View Masked Autoencoder for Heterogeneous Graph Self-Supervised Learning
Abstract page for arXiv paper 2611.00511: Efficient and Scalable Multi-View Masked Autoencoder for Heterogeneous Graph Self-Supervised Learning
arxiv.org
November 3, 2026 at 11:10 AM
The paper introduces Multi-View Masked Auto Encoder, which masks entire meta-path views and reconstructs their representations from remaining views.
Source: arXiv cs.LG
Source: arXiv cs.LG
New on CRAN: scCAN (1.0.6). View at https://CRAN.R-project.org/package=scCAN
scCAN: Single-Cell Clustering using Autoencoder and Network Fusion
A single-cell Clustering method using 'Autoencoder' and Network fusion ('scCAN') Bang Tran (2022) <<a href="https://doi.org/10.1038%2Fs41598-022-14218-6" target="_top">doi:10.1038/s41598-022-14218-6</a>> for segregating the cells from the high-dimensional 'scRNA-Seq' data. The software automatically determines the optimal number of clusters and then partitions the cells in a way such that the results are robust to noise and dropouts. 'scCAN' is fast and it supports Windows, Linux, and Mac OS.
CRAN.R-project.org
November 4, 2026 at 10:20 PM
New on CRAN: scCAN (1.0.6). View at https://CRAN.R-project.org/package=scCAN
PCAE, a hybrid quantum-classical architecture, reconstructs images using measured Pauli correlators on 6-qubit quantum processors. Achieves 25.8–41.6 dB PSNR on public datasets and successfully runs on real IBM Quantum hardware without retraining.
Pauli-Correlator Autoencoder for Quantum Image Reconstruction
iq.fp2.dev
November 4, 2026 at 5:24 AM
PCAE, a hybrid quantum-classical architecture, reconstructs images using measured Pauli correlators on 6-qubit quantum processors. Achieves 25.8–41.6 dB PSNR on public datasets and successfully runs on real IBM Quantum hardware without retraining.
PCAE ist eine hybrid quantenklassische Architektur, die Bilder unter Verwendung gemessener Pauli-Korrelatoren auf 6-Qubit-Quantenprozessoren rekonstruiert. Sie erreicht 25,8–41,6 dB PSNR auf öffentlichen Datensätzen und wird erfolgreich auf echter IBM Quantum Hardware ohne ern...
Pauli-Korrelator-Autoencoder zur Quantenbildrekonstruktion
iq.fp2.dev
November 4, 2026 at 5:25 AM
PCAE ist eine hybrid quantenklassische Architektur, die Bilder unter Verwendung gemessener Pauli-Korrelatoren auf 6-Qubit-Quantenprozessoren rekonstruiert. Sie erreicht 25,8–41,6 dB PSNR auf öffentlichen Datensätzen und wird erfolgreich auf echter IBM Quantum Hardware ohne ern...
PCAE, una arquitectura híbrida cuántico-clásica, reconstruye imágenes utilizando correladores de Pauli medidos en procesadores cuánticos de 6 qubits. Logra 25.8–41.6 dB PSNR en conjuntos de datos públicos y se ejecuta exitosamente en hardware IBM Quantum real sin reentrenamiento.
Autoencoder de Correlador de Pauli para Reconstrucción de Imágenes Cuánticas
arxiv.org
November 4, 2026 at 5:24 AM
PCAE, una arquitectura híbrida cuántico-clásica, reconstruye imágenes utilizando correladores de Pauli medidos en procesadores cuánticos de 6 qubits. Logra 25.8–41.6 dB PSNR en conjuntos de datos públicos y se ejecuta exitosamente en hardware IBM Quantum real sin reentrenamiento.
Autoencoder infers Hamiltonians from quantum dot chains
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Autoencoder Infers Hamiltonians From Quantum Dot Chains
This research presents an autoencoder framework that infers Hamiltonians for quantum systems directly from data, bypassing labeled parameters.
quantumzeitgeist.com
October 31, 2026 at 1:16 AM
Autoencoder infers Hamiltonians from quantum dot chains
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Autoencoder infers Hamiltonians from quantum dot chains
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Autoencoder Infers Hamiltonians From Quantum Dot Chains
This research presents an autoencoder framework that infers Hamiltonians for quantum systems directly from data, bypassing labeled parameters.
quantumzeitgeist.com
October 31, 2026 at 1:16 AM
Autoencoder infers Hamiltonians from quantum dot chains
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
Read more:
https://quantumzeitgeist.com/kitaev-autoencoder-hamiltonians-quantum-infers-dot/
PCAE, một kiến trúc lai lượng tử-cổ điển, tái thiết lập hình ảnh bằng các tương quan Pauli được đo lường trên các bộ xử lý lượng tử 6-qubit. Đạt được 25.8–41.6 dB PSNR trên các tập dữ liệu công khai và chạy thành công trên phần cứng IBM Quantum thực tế mà không cần huấn luyện ...
Autoencoder Tương Quan Pauli cho Tái Thiết Lập Hình Ảnh Lượng Tử
arxiv.org
November 4, 2026 at 5:26 AM
PCAE, một kiến trúc lai lượng tử-cổ điển, tái thiết lập hình ảnh bằng các tương quan Pauli được đo lường trên các bộ xử lý lượng tử 6-qubit. Đạt được 25.8–41.6 dB PSNR trên các tập dữ liệu công khai và chạy thành công trên phần cứng IBM Quantum thực tế mà không cần huấn luyện ...
Autoencoder infers Hamiltonians from quantum dot chains
Autoencoder infers Hamiltonians from quantum dot chains
Quantum Zeitgeist
quantumzeitgeist.com
October 30, 2026 at 8:23 PM
Autoencoder infers Hamiltonians from quantum dot chains
Detecting Anomalies in Karst Aquifer Recharge: A Temporal Convolutional AutoEncoder Framework With High‐Frequency Temperature‐Conductivity Signals
Zhu, Niu, Huang et al.
Water Resources Research
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025wr042001
#stableisotopes
Zhu, Niu, Huang et al.
Water Resources Research
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025wr042001
#stableisotopes
Detecting Anomalies in Karst Aquifer Recharge: A Temporal Convolutional AutoEncoder Framework With High‐Frequency Temperature‐Conductivity Signals
A temporal convolutional autoencoder captures recharge anomalies in karst aquifers using high-frequency temperature and conductivity data Event-scale responses are classified by combining anomaly...
agupubs.onlinelibrary.wiley.com
October 30, 2026 at 7:52 PM
Detecting Anomalies in Karst Aquifer Recharge: A Temporal Convolutional AutoEncoder Framework With High‐Frequency Temperature‐Conductivity Signals
Zhu, Niu, Huang et al.
Water Resources Research
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025wr042001
#stableisotopes
Zhu, Niu, Huang et al.
Water Resources Research
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025wr042001
#stableisotopes