#ResNet
relevant:

arxiv.org/abs/1911.13299 (a random Wide ResNet-50 contains a subnet that matches the performance of a trained ResNet-34, and such subnet can be efficiently found)

arxiv.org/abs/2006.07990 (with log overparameterization, a random ReLU net contains a subnet close to any target network)
What's Hidden in a Randomly Weighted Neural Network?
Training a neural network is synonymous with learning the values of the weights. By contrast, we demonstrate that randomly weighted neural networks contain subnetworks which achieve impressive perform...
arxiv.org
November 13, 2026 at 4:00 AM
I did find something Combos would find unfortunate... Contact for more info - keep this on RESNET+ (TS/REL 7EYE)
November 13, 2026 at 10:09 PM
この手法の大きな特徴として、ResNetの出力に単純に線形補間で引き伸ばした画像を加算している。これによってResNetは線形補間で引き伸ばした事で失われた画像の高周波成分をでっち上げる事だけに集中できる。実験ではLPIPSを使って品質を評価し、NTCと同水準の品質が達成出来ているとしている。
使い勝手も品質も良いイケてるニューラルテクスチャ圧縮なだけに、気になるのは「そのでかいニューラルネットワークは本当にサンプリング時に走らせて良い性能で動くのか」だが、性能の評価はfuture workとされている
November 17, 2026 at 6:50 AM
2段組PDFをコピペすると左右が混ざるのは、抽出順が版面と一致しないから。文章は壊れていない。範囲を囲めば見えている順のまま日本語に。
www.zipp-mac.com?utm_source=b...
November 15, 2026 at 6:22 AM
Depth becomes time in a ResNet. Explore adjoint training, noncommuting linear flows, and a loss with no bad local minima—but increasingly flat saddles. https://www.gpeyre.com/blog/2026/11/10/training-very-deep-resnets/
November 10, 2026 at 6:13 AM
Signal or noise? > AI, Vol. 7, Pages 479: MS-DA-ResNet: A Multiscale
Dual-Attention Residual Network for Remaining Useful Life
Prediction of Aero-Engines Under Multiple Operating
Conditions >> Comment below! #mhealth #healthtech #IoT #industry40 #AI
AI, Vol. 7, Pages 479: MS-DA-ResNet: A Multiscale Dual-Attention Residual Network for Remaining Useful Life Prediction of Aero-Engines Under Multiple Operating Conditions
Accurate remaining useful life (RUL) prediction of aero-engines under varying operating conditions is challenging because condition-induced sensor variations can obscure degradation signatures. This study evaluates a multiscale dual-attention residual network (MS-DA-ResNet) under a leakage-controlled protocol on the multiple-condition FD002 and FD004 subsets. The 125-cycle cap is applied only to training and validation targets, whereas official NASA test endpoint labels remain uncapped. All models use the same engine-level split, fixed 14-sensor input, causal preprocessing, and best-validation checkpoint, as well as three identical matched seeds. Under the corrected official-label evaluation, MS-DA-ResNet provides the best aggregate results on both datasets: FD002 RMSE 25.14 ± 0.41, MAE 16.71 ± 0.21, R2 0.781 ± 0.007, and NASA Score 5597 ± 389; FD004 RMSE 24.58 ± 0.22, MAE 17.35 ± 0.07, R2 0.797 ± 0.004, and NASA Score 3430 ± 77. MS-DA-ResNet remains competitive but does not achieve universal superiority. Its error-direction behavior is interpreted separately, using Bias and the overestimation ratio. Condition-wise normalization with causal smoothing improves FD004 relative to the tested global-normalization and no-smoothing alternatives. A corrected five-seed L = 30 sensitivity run does not improve the primary L = 15 configuration.
dlvr.it
November 12, 2026 at 3:24 PM
These jobs are all across the country, and the numbers are growing as more cities & states require HERS ratings for code compliance. They pay well and are almost all small businesses. They're closer to trades jobs than office jobs, w/ technical training.

RESNET has a "how to become a Rater" (link)
How to Become a Certified HERS® Rater - RESNET
www.resnet.us
November 10, 2026 at 8:14 PM
Patch-Based Generalized Neural Texture Upscalingは隠れ層のレイヤー数256、4段のResNetにサンプリングしたい位置とその周囲の3x3のピクセルを食べさせて、2x2のピクセルを出力させる。NTCと比べてモデルは大きいが、単一のモデルであらゆる画像を高解像度化できるように学習を行う。
入力は従来の低解像度テクスチャをサンプリングした物がそのまま使える為、既存の実装に追加で仕込みやすい。学習はアルベド、法線、マテリアルのテクスチャをごちゃまぜにして行われており、どの種類のテクスチャを食べさせても同じモデルで高解像度化できるようになっている
November 17, 2026 at 6:50 AM
Most people don't connect green building to jobs. But most above-code programs ( #ENERGYSTAR, #LEED) and many local codes require energy "ratings" to evaluate the #energy-efficiency of homes.

According to the industry group RESNET, there are 6,700+ people employed as HERS raters.
HERS Raters do on-site testing and diagnostics, as well as some basic computer modeling of energy performance. There are HERS Raters across the country - mostly working on new home / building compliance with energy codes and above-code programs (e.g. #ENERGYSTAR, #LEED).
November 10, 2026 at 8:14 PM
Differential-Neural Cryptanalysis of ChaCha and Related ARX Stream Ciphers (Nimai Parsa, Nitin Kumar Sharma) ia.cr/2026/2382
November 9, 2026 at 1:35 AM
Tool Duel: DeepVariant (Google) vs Clair3 (Ruibang Luo's lab, HKU). One reads every candidate site as a pileup image; the other runs a fast Bi-LSTM on read counts and re-calls only the lowest-quality 30% with a small ResNet. Details at toolome.bio #Genomics #Bioinformatics
November 8, 2026 at 4:14 PM
RESNET HERS Certified Energy Rating Home

Need a HERS rating Florida homes can rely on? Save this for later.

https://randolphscottbell.com/consulting/hers-energy-rating/?utm_source=bluesky&utm_medium=social&utm_campaign=bec-consulting-20261106&utm_content=pc254208e
November 6, 2026 at 7:00 PM
A 3-Step Data Quality Pre-Flight Check Before Training Vision Models
A new state-of-the-art vision architecture or backbone checkpoint seems to appear every few days. Teams rush to benchmark the newest transformer variant, convolutional hybrid, or multi-scale neck against their internal tasks, expecting architecture alone to solve stubborn operational accuracy problems. Yet when training completes, the performance gap between a three-year-old ResNet and the newest flagship model is frequently negligible. The reason is simple: modern deep neural networks are expressive enough to fit whatever distribution they receive. If the underlying data is structurally inconsistent, riddled with annotation noise, or severely skewed, no architectural advancement will compensate. High-performing engineering teams treat dataset preparation with the same rigor as pre-flight aerospace checklists. Before launching your next training run, execute this three-step data quality pre-flight audit. +-----------------------------------------------------------------+ | 3-STEP DATA QUALITY PRE-FLIGHT CHECK | +-----------------------------------------------------------------+ [Step 1: Resolution & Geometry] --> Verify native resolutions, aspect ratios & resize artifacts [Step 2: Annotation Verification] --> Audit label consistency, tight bounds & edge disagreements [Step 3: Class Distribution & Coverage] --> Measure effective sample counts, tail ratios & rare states +-----------------------------------------------------------------+ ## Step 1: Examine Image Resolution and Geometry Consistency Real-world computer vision datasets rarely come from uniform cameras. They aggregate images from mobile devices, surveillance streams, industrial webcams, and drone payloads. Before feeding images into a standardized training pipeline, audit your geometric distribution: * **Aspect ratio variance:** Plot the ratio of width to height across all images. If aspect ratios range from 1:2 to 16:9, standard square resizing ($512 \times 512$ or $640 \times 640$) will either crush horizontal features or require massive letterboxing that wastes spatial tokens. * **Native resolution floor:** Identify samples whose native capture resolution is substantially lower than your model's input size. Upsampling a $120 \times 120$ thumbnail to $640 \times 640$ introduces interpolation blur, teaching the model blurry edge representations. * **Bounding box scale distribution:** Calculate relative bounding box areas ($A_{box} / A_{image}$). If 80% of target objects occupy less than 0.5% of total image area, standard feature pyramids without high-resolution crops will fail to detect them. import os from PIL import Image import numpy as np def audit_image_geometry(image_paths): aspect_ratios = [] resolutions = [] for path in image_paths: with Image.open(path) as img: w, h = img.size aspect_ratios.append(w / h) resolutions.append((w, h)) ar_array = np.array(aspect_ratios) print(f"Aspect Ratio Mean: {ar_array.mean():.2f}, Std: {ar_array.std():.2f}") print(f"Aspect Ratio Min: {ar_array.min():.2f}, Max: {ar_array.max():.2f}") anomalies = [p for p, ar in zip(image_paths, aspect_ratios) if ar < 0.5 or ar > 2.5] print(f"Found {len(anomalies)} severe aspect ratio anomalies.") return anomalies ## Step 2: Verify Labeling Accuracy with Expert Consensus Review Annotation errors are the single most destructive form of training noise because gradient updates actively penalize correct feature representations. Run a targeted label audit on a stratified 5% sample of your dataset: * **Check boundary tightness:** Are bounding boxes drawn tightly around object hulls, or do they include excessive background margins? Loose bounding boxes pollute foreground feature embeddings with background clutter. * **Audit extreme occlusion rules:** Does your labeling team annotate objects that are 90% occluded? If one annotator marks an occluded vehicle bumper while another ignores it, the loss function receives contradictory guidance. * **Measure Inter-Annotator Agreement (IAA):** Have multiple annotators label the same 100 validation images. Compute Cohen's Kappa for categorical attributes and mean IoU for bounding boxes. A mean IoU below 0.75 indicates that your labeling guidelines are ambiguous and require immediate clarification. ## Step 3: Assess Class Distribution Balance and Tail Coverage Class imbalance is a primary source of silent production failure. While aggregate mAP might look strong, performance on tail classes can easily collapse to zero. Evaluate your class distributions across two dimensions: 1. **Raw Class Frequencies:** Calculate the imbalance ratio ($N_{max} / N_{min}$). Ratios exceeding 50:1 require explicit remediation, such as focal loss, repeat-factor sampling, or synthetic injection. 2. **Co-occurrence and Scenario Diversity:** Measure how frequently classes appear across distinct backgrounds and lighting conditions. If a forklift class appears in 500 images but all 500 were captured inside a single warehouse facility on the same morning, your effective sample count for that class is closer to 5 than 500. Class Metric | Target Standard | Warning Threshold | Remediation Strategy ---|---|---|--- **Imbalance Ratio** | $< 10:1$ | $> 50:1$ | Active learning targeted retrieval or repeat-factor sampling **Minimum Instances per Class** | $\ge 200$ verified | $< 50$ instances | Targeted domain collection **Environmental Diversity** | $\ge 5$ distinct venues | $1-2$ venues only | Data collection in alternate environments **Label Agreement (mIoU)** | $\ge 0.82$ | $< 0.75$ | Refine annotation guidelines with edge examples ## Operationalizing Pre-Flight Audits Integrating these three checks into your dataset intake pipeline transforms training predictability. Instead of spending weeks diagnosing poor convergence after a multi-GPU training run, you eliminate structural data bugs before spending a single dollar on compute. For related insights on data quality and evaluation engineering, read our guide on building an agent evaluation CI/CD pipeline and robot data quality scoring. To gain practical experience auditing and curating production visual datasets, audit Harpreet Sahota's Coursera course: Hands-on Data Centric Visual AI. ## Frequently asked questions ### What is the purpose of a pre-flight data quality check? A pre-flight check validates dataset consistency, label integrity, and class distribution before spending GPU compute on training runs. Catching structural data errors beforehand prevents silent convergence failures and expensive debugging cycles. ### How do aspect ratio disparities degrade vision model training? When source images vary wildly in aspect ratio, standard resize or square-padding transforms either distort physical object geometry or introduce excessive zero-padding pixels, degrading convolution and attention feature extractors. ### What threshold of inter-annotator disagreement indicates a defective label specification? In multi-annotator audits, a Cohen's Kappa below 0.70 or an IoU overlap below 0.75 across bounding box labels strongly indicates ambiguous labeling instructions rather than human carelessness. ### Where can teams learn end-to-end visual data auditing practices? Complete methodologies for dataset auditing and pre-flight quality verification are covered in Harpreet Sahota's Coursera course, Hands-on Data Centric Visual AI. _This post first appeared as the original article. I'm Harpreet Sahota, Hacker-in-Residence at The Employable Data Scientist, where we help AI and developer-tool teams ship prototypes, evaluations, and technical content. I also host the Artists of Intelligence podcast. For new field notes by email, subscribe on Substack._
dev.to
November 7, 2026 at 2:37 PM
The Minnehaha Creek Watershed District has unveiled a groundbreaking flood model that combines real-time weather data with advanced mapping to revolutionize flood planning for 29 cities across 178 square miles.

Learn more here

#HennepinCounty #MN #CitizenPortal #FloodPlanning
Watershed district unveils watershed‑scale flood model to guide county and city planning
James Whisker of the Minnehaha Creek Watershed District told Hennepin County commissioners the district has built a first‑of‑its‑kind, watershed‑scale integrated model that links 7‑day forecasts, county mesonet data and a 23‑station USGS ResNet to improve short‑term response and long‑term planning across 178 square miles and 29 cities.
citizenportal.ai
November 2, 2026 at 7:19 AM
No metrics, no preprocessing, no conversion scripts, target logged as 'other' — the card is bare. Still, Apache-2.0 and a 25.6M-parameter backbone make it the obvious Axera smoke test. Hard to see anyone validating a toolchain with it until the glue ships.
Axera's ResNet-50 Lands on Modelers.cn: A Tiny Apache-2.0 Edge Classifier — and a Toolchain Test Bench
Axera, the Chinese edge-AI chip maker, has published a packaged ResNet-50 image classifier on Modelers.cn — the Huawei/Ascend-ecosystem model hub — under an Apache-2.0 licence. On paper that sounds…
aichina.news
November 1, 2026 at 2:21 PM
Yang Li, Gang He, Ruichen Zhang, Guangyuan Liu, Yinqiu Liu
Pacifo: An Efficient Fully Homomorphic Encryption-based Lightweight ResNet for Trustworthy Machine Learning as a Service
https://arxiv.org/abs/2610.33525
October 28, 2026 at 10:44 AM
Real-world use: Chatbots, translation, sentiment analysis, and document summarization.

6. Computer Vision Models — Understand Images

These models extract useful information from images and videos.

Examples: ResNet, YOLO, Vision Transformers, and segmentation models.
October 28, 2026 at 8:35 AM
Yang Li, Gang He, Ruichen Zhang, Guangyuan Liu, Yinqiu Liu: Pacifo: An Efficient Fully Homomorphic Encryption-based Lightweight ResNet for Trustworthy Machine Learning as a Service https://arxiv.org/abs/2610.33525 https://arxiv.org/pdf/2610.33525 https://arxiv.org/html/2610.33525
October 27, 2026 at 6:39 AM
CNN+Transformer系列を洗い直してみた。SwinはTransformerをCNNに寄せ、ConvNeXtはAttentionなしでResNetを作り直してSwinを上回った。ImageNetのtop-1は系統を問わず90〜91%で頭打ち。大規模になるとアーキテクチャの差は縮み、効くのはデータと計算量。timmの重みも、名前より事前学習データを見たい。
yurudeep.com/posts/deeple...
画像のバックボーン、CNNとTransformerはどっちが強い?──Swin・ConvNeXt・CoAtNetを整理したら、差はデータと計算で縮んでいた - ゆるディープ
Swin Transformer、ConvNeXt、CoAtNet、MaxViTなど、2021年以降の画像バックボーンを「CNNとTransformerがどう歩み寄ったか」で3系統に整理した。大規模になるとアーキテクチャの差は縮み、効くのはデータと計算量になる、という話まで。
yurudeep.com
October 27, 2026 at 2:44 AM
Yep. 'Large' language models are new. While we had language models before, they weren't large. If you tried to make prior techniques large, they wouldn't learn.

I would argue 'deep' networks also did not exist before resnet and highway networks for the same reason, although the term did exist.
October 18, 2026 at 2:17 PM
mensen, laten we alsjeblieft resnet niet AI gaan noemen.

Machine Learning. Als je toch een Engelse term gebruikt, gebruik dan deze.

Wat zou een goed Nederlands woord ervoor zijn?

Aanleeralgoritme? Flexfunctie?
Grootste 2D-kaart van het heelal en AI helpen sterrenkundigen bij het ontdekken van nieuwe zwaartekrachtlenzen
www.astroblogs.nl/2026/10/15/g...
October 15, 2026 at 12:02 PM
In our MED-VAE architecture a common space arises by using a pretrained ResNet-50 as a scaffold onto which subjects brain activity can align to. Our method does not require common inputs across individuals and after training, the Resnet is no longer needed. 2/6
October 14, 2026 at 4:08 PM
The shared space clusters by image category, inherited from the ResNet-50 scaffold.

This is not a property of the RN-50 architecture alone:
Untrained ResNet-50: retrieval drops from 46.2% to 6.6%.
Shuffled neural data–model feature 1:1 correspondence: 1.0% (chance 0.8%). 4/6
October 14, 2026 at 4:08 PM
DeepLabV3+ — the segmentation baseline since 2018 — is now in Tsinghua's Jittor model zoo. But an example proves the ops run, not training at scale. Hard to see CV teams leaving torchvision until Mask2Former-class ports train on domestic chips.

*(244 graphemes. Same three beats, same reader —...
Example 4: DeepLabV3+ for Semantic Segmentation — Jittor
A Jittor framework tutorial walking through a full implementation of the DeepLabV3+ semantic segmentation model, from VOC2012 data loading and the ResNet/ASPP/Decoder architecture through to training, validation and mIoU evaluation.
aichina.news
October 14, 2026 at 10:05 AM
was applied to understand the visual differences between the presynaptic sites. To make it work for 2D single section EM images a specialised 2D ResNet-1 8 classifier was trained. QuAC primarily altered the VESICLE morphology like converting a pale A-delta HTMR vesicle int a dense core/19
October 12, 2026 at 7:00 PM