# Deep neural networks A deep neural network is a neural network containing more than one hidden layer. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## A study of sequence weighting at scale DevFeed: [A study of sequence weighting at scale]() Original publisher: [Read original article]() Author: Alex Renda Published: 2026-09-14T00:00:00Z Content type: article Language: en Sources: [Jane Street]() Topics: [scaling laws](), [datasets](), [Deep neural networks](), [Chinchilla scaling law]() Tags: [data](), [machine-learning](), [model](), [scaling-laws](), [training]() ### AI overview The article studies how sequence weighting affects language-model training across model scales. It reports non-monotonic behavior: small-to-medium models increasingly learn data-specific patterns in proportion to sequence weights, while large models can learn all patterns in the data more independently of those weights. ### Source excerpt TL;DR: We study the scaling laws of data weighting across in-house and open-weight LMs, finding non-monotonic behavior across scales. We vary the weight assigned to sequences during training and measure how strongly the model's loss reduction on a sequence depends on the sequence's weight. Taken together, our results are consistent with a general trend: as models transition from small to medium scale, they transition from learning general patterns independent of data weight to learning data-specific patterns proportional to the data weights. As models then transition from medium to large scale they are able to learn all patterns present in the data, once again independent of data weight. ## How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules DevFeed: [How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules]() Original publisher: [Read original article]() Published: 2026-09-10T16:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Deep neural networks](), [datasets](), [Code]() Tags: [ai](), [ai-models](), [antibiotics](), [applied-ai](), [chatgpt](), [code](), [codex](), [datasets](), [deep-learning](), [search]() ### AI overview César de la Fuente's lab uses deep-learning models, ChatGPT, and Codex to search genome and protein datasets for antimicrobial candidates that could help fight drug-resistant infections. ### Source excerpt César de la Fuente's lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections. ## Mapping global methane emissions from space with deep learning DevFeed: [Mapping global methane emissions from space with deep learning]() Original publisher: [Read original article]() Published: 2026-09-01T18:40:00Z Content type: article Language: en Sources: [The latest research from Google]() Topics: [Deep neural networks](), [Google]() Tags: [climate-sustainability](), [deep-learning](), [earth-ai](), [framework](), [global](), [google](), [machine-intelligence](), [mapping](), [nasa](), [research](), [space]() ### AI overview The article presents MAPL-EMIT, a deep-learning framework for automating global detection, enhancement prediction, and source estimation of methane plumes from EMIT hyperspectral satellite measurements. ### Source excerpt Climate & Sustainability ## Broadening access to Skala creates a faster path to predictive DFT DevFeed: [Broadening access to Skala creates a faster path to predictive DFT]() Original publisher: [Read original article]() Author: Sebastian Ehlert, Stefano Battaglia, Thijs Vogels, Jan Hermann, Jens Wehner, Giulia Luise, Klaas Giesbertz, Chin-Wei Huang, Aaron Kaplan, Kate Milton, Stephanie Marisa Lanius, Derk Kooi, P. Bernát Sza Published: 2026-08-20T16:00:00Z Content type: article Language: en Sources: [Microsoft Research]() Topics: [Deep neural networks](), [Benchmark](), [Microsoft](), [data](), [Software]() Tags: [benchmark](), [chemistry](), [data](), [deep-learning](), [integration](), [microsoft](), [performance](), [research](), [research-blog](), [software](), [workflows]() ### AI overview Microsoft Research presents Skala 1.1, a deep-learning exchange-correlation functional that improves accuracy for molecular simulations while expanding access through integrations with major electronic-structure software. A living benchmark will track the computational performance of future releases. ### Source excerpt Skala 1.1, the updated deep-learning exchange-correlation functional from Microsoft Research, provides greater accuracy, expanded accessibility across the computational chemistry ecosystem, and a living benchmark to track computational performance. The post Broadening access to Skala creates a faster path to predictive DFT appeared first on Microsoft Research. ## Seeing beyond BMI: Estimating cardiometabolic risk with smartphone imagery DevFeed: [Seeing beyond BMI: Estimating cardiometabolic risk with smartphone imagery]() Original publisher: [Read original article]() Published: 2026-08-17T10:34:00Z Content type: article Language: en Sources: [The latest research from Google]() Topics: [Deep neural networks](), [Deep learning](), [Machine learning](), [Google]() Tags: [deep-learning](), [general-science](), [google](), [health](), [health-bioscience](), [machine-intelligence](), [machine-learning](), [monitoring](), [multimodal](), [science]() ### AI overview Google Research presents PhotoScan, a deep learning approach that estimates body composition from smartphone photos and predicts insulin resistance with accuracy comparable to DXA scans in a clinical research setting. The article explains how body-composition measures such as fat distribution and visceral fat may complement wearable data for earlier cardiometabolic risk assessment. ### Source excerpt General Science ## 🗓 This Week In AI Research (17-24 July 26) DevFeed: [🗓 This Week In AI Research (17-24 July 26)]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-07-30T10:44:06Z Content type: article Language: en Sources: [Into AI]() Topics: [AI Research](), [Artificial Intelligence](), [releases](), [Large Language Model](), [Deep neural networks](), [parquet]() Tags: [ai](), [ai-research](), [anthropic](), [claude](), [compression](), [llms](), [neural-networks](), [releases]() ### AI overview A weekly roundup of AI research papers and releases, including Claude Opus 5, the HOPE framework for analyzing knowledge in deep neural networks through compression, and research on how large language models track evolving user intent across conversations. ### Source excerpt The top 10 AI research papers and releases this week (Claude Opus 5, Laguna S 2.1, Loopie, Nanbeige 4.2, Fugu-Cyber, and more) ## Why intent prediction needs more than an LLM DevFeed: [Why intent prediction needs more than an LLM]() Original publisher: [Read original article]() Author: Phoebe Sajor Published: 2026-06-30T07:40:00Z Content type: article Language: en Sources: [Stack Overflow Blog]() Topics: [foundation-models](), [personalization](), [Graphs](), [Large Language Model](), [data](), [Deep neural networks]() Tags: [agentic-ai](), [ai](), [data](), [foundation-models](), [llm](), [llms](), [math](), [personalization](), [podcast](), [se-stackoverflow](), [se-tech](), [transformers]() ### AI overview Ryan Donovan interviews Frank Portman, CTO at Yobi, about why large language models are not the right inductive bias for predicting human intent and behavior. The discussion covers Yobi's behavioral foundation models, which use transformers and graph neural networks to support large-scale personalization while keeping consumer data private. ### Source excerpt Ryan sits down with Frank Portman, CTO at Yobi, to talk about why next-token prediction, though great for language, isn't the right inductive bias for forecasting human behavior. They discuss how Yobi builds a "foundation model of behavior" using transformers and graph neural networks instead of chat-style LLMs, and what it takes to run millions of personalization decisions per second while keeping consumer data private. ## Neural Networks, As Simple As They Can Get DevFeed: [Neural Networks, As Simple As They Can Get]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-05-13T14:43:34Z Content type: tutorial Language: en Sources: [Dr. Ashish Bamania]() Topics: [Deep neural networks](), [Artificial Intelligence]() Tags: [artificial-intelligence](), [beginner](), [coding](), [data-science](), [machine-learning](), [neural](), [neural-networks](), [programming](), [technology]() ### AI overview A beginner-friendly lesson explaining what neural networks are, how neurons or perceptrons connect into networks, and how these networks are trained for tasks such as text generation and object recognition. ### Source excerpt A beginner's lesson on what Neural networks are and how they are trained to do amazing things. Continue reading on Level Up Coding " ## Coding Challenge #119 - AI Pong Player DevFeed: [Coding Challenge #119 - AI Pong Player]() Original publisher: [Read original article]() Author: John Crickett Published: 2026-05-09T08:01:53Z Content type: tutorial Language: en Sources: [Coding Challenges]() Topics: [Code Challenge](), [Reinforcement learning](), [coding](), [Deep neural networks](), [NumPy]() Tags: [ai](), [coding](), [games](), [neural-networks](), [python](), [reinforcement-learning]() ### AI overview A coding challenge guides readers through building a reinforcement learning agent that learns to play Atari Pong from raw screen pixels. It introduces policy gradients, the REINFORCE algorithm, returns, stochastic policies, and variance reduction, with Python and NumPy as prerequisites. ### Source excerpt This challenge is to build your own AI pong player with reinforcement learning. ## Understanding neural networks through sparse circuits DevFeed: [Understanding neural networks through sparse circuits]() Original publisher: [Read original article]() Published: 2025-11-13T10:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Deep neural networks](), [Artificial Intelligence](), [OpenAI](), [Chain-of-thought]() Tags: [ai](), [analysis](), [chain-of-thought](), [models](), [neural-networks](), [openai](), [reasoning](), [research](), [safety](), [training]() ### AI overview OpenAI explores mechanistic interpretability through sparse circuits: training neural networks to use simpler, more traceable computations. The approach aims to make model behavior easier to understand and support safer, more reliable AI systems. ### Source excerpt OpenAI is exploring mechanistic interpretability to understand how neural networks reason. Our new sparse model approach could make AI systems more transparent and support safer, more reliable behavior. ## Four ways to run ONNX models on a GPU with CUDA in Clojure DevFeed: [Four ways to run ONNX models on a GPU with CUDA in Clojure]() Original publisher: [Read original article]() Published: 2025-11-09T17:49:00Z Content type: tutorial Language: en Sources: [Dragan Djuric]() Topics: [CUDA](), [GPU](), [onnx](), [Clojure](), [Artificial Intelligence](), [Deep neural networks]() Tags: [ai](), [clojure](), [code](), [cuda](), [deep](), [diamond](), [examples](), [gpu](), [image-recognition](), [model](), [neural-networks](), [onnx](), [tensors]() ### AI overview A tutorial presents four ways to run ONNX models on a GPU with CUDA using Clojure libraries including Diamond ONNX RT, Deep Diamond, and ClojureCUDA. It covers GPU tensor backends and a configuration that keeps input and output tensors in main memory while executing the model on the GPU. ### Source excerpt Two weeks ago, I announced a new Clojure ML library, Diamond ONNX RT, which integrates ONNX Runtime into Deep Diamond. In that post, we explored the classic Hello World example of Neural Networks, MNIST handwritten image recognition, step-by-step. We run that example on the CPU, from main memory. The next logical step is to execute this stuff on the GPU. You'll see that with a little help of ClojureCUDA and Deep Diamond built-in CUDA machinery, this is both easy and simple, requiring almost no effort from a curious Clojure programmer. But don't just trust me; let's fire up your REPL, and we can continue together. Here's how you can evaluate this directly in your REPL (you can use the Hello World that is provided in the ./examples sub-folder of Diamond ONNX RT as a springboard). Require Diamond's namespaces First things first, we refer functions that we're going to use. (require '[uncomplicate.commons.core :refer [with-release]] '[uncomplicate.neanderthal.core :refer [transfer! iamax native]] '[uncomplicate.diamond [tensor :refer [tensor with-diamond]] [dnn :refer [network]] [onnxrt :refer [onnx]]] '[uncomplicate.diamond.internal.dnnl.factory :refer [dnnl-factory]] '[uncomplicate.diamond.internal.cudnn.factory :refer [cudnn-factory]] '[hello-world.native :refer [input-desc input-tz mnist-onnx]]) None of the following ways to run CUDA models has preference, you use the one that best suits your needs. Way one One of the ways to run ONNX models on your GPU is to simply use Deep Diamond's cuDNN factory as the backend for your tensors. Then, the machinery recognizes what you need and proceeds doing everything on the GPU, using the right stream for tensors, Deep Diamond operations, and ONNX Runtime operations. This looks exactly the same as any other Deep Diamond example from this blog or the DLFP book. (with-diamond cudnn-factory [] (with-release [cuda-input-tz (tensor input-desc) mnist (network cuda-input-tz [mnist-onnx]) classify! (mnist cuda-input-tz)] (transfer! input ## Using AI to identify genetic variants in tumors with DeepSomatic DevFeed: [Using AI to identify genetic variants in tumors with DeepSomatic]() Original publisher: [Read original article]() Published: 2025-10-16T16:33:00Z Content type: article Language: en Sources: [The latest research from Google]() Topics: [Artificial Intelligence](), [Google](), [Deep neural networks](), [datasets](), [Machine Learning & Artificial Intelligence]() Tags: [2025](), [ai](), [general-science](), [google](), [health-bioscience](), [machine-learning](), [neural-networks](), [open-source-models-datasets](), [research](), [training]() ### AI overview Google Research presents DeepSomatic, an AI-powered tool that uses convolutional neural networks to identify cancer-related genetic variants in tumor sequencing data. The tool supports major sequencing platforms and sample-processing methods, and its software and training dataset are openly available to researchers. ### Source excerpt General Science ## How to Generate New Art From Existing Images Using Deep Learning? DevFeed: [How to Generate New Art From Existing Images Using Deep Learning?]() Original publisher: [Read original article]() Author: Damilare Keshinro Published: 2025-07-24T04:56:41Z Content type: tutorial Language: en Sources: [Baeldung - CS]() Topics: [Deep neural networks](), [Tensorflow](), [Tutorial](), [Image](), [Computer science]() Tags: [building](), [computer-science](), [computer-vision](), [computer-vision-image-processing](), [deep-learning](), [generate](), [how-to](), [image-processing](), [images](), [model](), [techniques](), [tensorflow]() ### AI overview A tutorial on neural style transfer using TensorFlow and a pretrained VGG16 model. It explains how to combine the content of an owl image with the artistic style of Edvard Munch's "Scream" by extracting features, defining losses, preprocessing images, and optimizing the result. ### Source excerpt A walkthrough of building an art style transfer program using deep learning techniques and tools like TensorFlow and the VGG16 model. The post How to Generate New Art From Existing Images Using Deep Learning? first appeared on Baeldung on Computer Science. ## How AI Medical Imaging Is Powering Precision Healthcare DevFeed: [How AI Medical Imaging Is Powering Precision Healthcare]() Original publisher: [Read original article]() Author: MARTIN ELIAS COSTA, AI ENGINEER @ TOPTAL Published: 2025-07-18T05:00:00Z Content type: article Language: en Sources: [Toptal Blog]() Topics: [Artificial Intelligence](), [Medical imaging](), [Computer vision](), [Deep neural networks](), [data](), [Development]() Tags: [ai](), [ai-engineer](), [artificial-intelligence](), [computer-vision](), [data](), [development](), [diagnostics](), [healthcare](), [images](), [machine-learning](), [medical-imaging](), [neural-networks]() ### AI overview This article explains how artificial intelligence is transforming medical imaging from image acquisition through diagnosis. It focuses on computer vision, neural networks, medical imaging data, and GPUs, and describes an AI system that analyzes brain MRI scans, identifies demyelinating lesions, measures brain-region volumes, classifies atrophy patterns, and integrates results into electronic health records. ### Source excerpt Artificial intelligence is revolutionizing how medical images are acquired, analyzed, and interpreted. The transformation ushers in a new era of data-driven diagnostics and faster, more personalized patient care. ## Visualize and understand GPU memory in PyTorch DevFeed: [Visualize and understand GPU memory in PyTorch]() Original publisher: [Read original article]() Author: Quentin Gallouédec Published: 2024-12-24T00:00:00Z Content type: tutorial Language: en Sources: [Hugging Face - Blog]() Topics: [PyTorch](), [GPU](), [Deep neural networks]() Tags: [code](), [gpu](), [graph](), [memory](), [neural-networks](), [pytorch](), [tensors](), [train](), [training](), [tutorial]() ### AI overview A tutorial on visualizing and understanding GPU memory usage in PyTorch during training. It explains how model creation, input and output tensors, activations, forward passes, backpropagation, and variable reassignment affect memory, and discusses estimating requirements and optimizing usage. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Visualizing piecewise linear neural networks DevFeed: [Visualizing piecewise linear neural networks]() Original publisher: [Read original article]() Author: Ricson Cheng Published: 2024-07-22T00:00:00Z Content type: article Language: en Sources: [Jane Street]() Topics: [Deep neural networks](), [Algorithm]() Tags: [algorithm](), [neural](), [neural-networks]() ### AI overview The article explains how piecewise-linearity in ReLU neural networks can be described and visualized. It examines activation patterns, polygonal regions, polyhedral complexes, and how adding layers changes these structures. ### Source excerpt Neural networks are often thought of as opaque, black-box function approximators, but theoretical tools let us describe and visualize their behavior. In particular, let's study piecewise-linearity, a property many neural networks share. This property has been studied before, but we'll try to visualize it in more detail than has been previously done. ## Is your work style written in the stars? DevFeed: [Is your work style written in the stars?]() Original publisher: [Read original article]() Author: Herbert Lui Published: 2023-08-04T00:00:00Z Content type: article Language: en Sources: [Figma Blog]() Topics: [Figma](), [Deep neural networks]() Tags: [collaboration](), [developer](), [figma](), [machine-learning]() ### AI overview Product Research Intern Jacky Huang and Software Engineer Willy Wu discuss FigStrology, a collaborative FigJam widget that uses machine learning to generate an astrological persona and horoscope from sticky notes and the current cosmic conditions of FigJam files. The article also describes Figma community projects that combine astrology, tarot, and collaborative experiences. ### Source excerpt Product Research Intern Jacky Huang and Software Engineer Willy Wu discuss the resurgent interest in astrology and how they're putting it into practice at work with their new collaborative FigJam widget, FigStrology. ## Maximizing Solution Visibility with Machine Learning-Powered App Recommendations DevFeed: [Maximizing Solution Visibility with Machine Learning-Powered App Recommendations]() Original publisher: [Read original article]() Author: Steven Kim Published: 2023-06-13T08:00:00Z Content type: release Language: en Sources: [Square Corner Blog RSS Feed]() Topics: [recommendation systems](), [Machine learning](), [Deep neural networks](), [App]() Tags: [app](), [developer](), [engineering](), [machine-learning](), [recommendations]() ### AI overview Square announces machine learning-powered App Recommendations that personalize third-party solution suggestions in the Seller Dashboard and App Marketplace. The system uses seller attributes and adoption data to rank apps, with ongoing retraining and monitoring for potential bias. ### Source excerpt Amplify solution visibility and drive conversion ## ESP32-S3 Edge-AI｜Human Activity Recognition Using Accelerometer Data and ESP-DL DevFeed: [ESP32-S3 Edge-AI｜Human Activity Recognition Using Accelerometer Data and ESP-DL]() Original publisher: [Read original article]() Author: John Lee Published: 2023-06-06T00:00:00Z Content type: tutorial Language: en Sources: [Blog on Developer Portal]() Topics: [ESP32-S3](), [On-device AI](), [Deep neural networks](), [ESP-IDF](), [Espressif]() Tags: [ai](), [blog](), [edge-ai](), [edge-computing](), [esp](), [esp-dl](), [esp-idf](), [esp32-s3](), [espressif](), [neural-networks](), [recognition]() ### AI overview This tutorial explains how to read accelerometer data and deploy a deep-learning model with ESP-DL on an ESP32-S3 for human activity recognition. It covers prerequisites, ESP-IDF project structure, model definition, layer initialization, and layer construction. ### Source excerpt Edge computing is a distributed computing paradigm that brings computation and data storage closer to the device's location. Edge Artificial Intelligence (edge-AI) is an exciting development within edge computing because it allows traditional technologies to run more efficiently, with higher performance and less power. Trained neural networks are used to make inferences on small devices. ## Hello Deep Learning: Reading handwritten digits DevFeed: [Hello Deep Learning: Reading handwritten digits]() Original publisher: [Read original article]() Published: 2023-03-30T10:00:04Z Content type: tutorial Language: en Sources: [Bert Hubert's writings]() Topics: [Deep learning](), [Neural Network](), [Deep neural networks](), [Image]() Tags: [blog-posts](), [deep-learning](), [images](), [layer](), [matrices](), [neural-network](), [pixels](), [spatial]() ### AI overview A tutorial in the Hello Deep Learning series explains how to build a neural network that recognizes all ten handwritten digits. It describes flattening 28x28 images, applying matrix multiplications and ReLU layers, and selecting the highest-scoring output. ### Source excerpt This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In the previous chapter we described how automatic differentiation of the result of neural networks works. In the first and second chapters we designed and trained a one-layer neural network that could distinguish images of the digit 3 and the digit 7, and the network did so very well. ## A GPU-Free, From-Scratch Introduction to Modern Machine Learning DevFeed: [A GPU-Free, From-Scratch Introduction to Modern Machine Learning]() Original publisher: [Read original article]() Published: 2023-03-30T09:59:00Z Content type: tutorial Language: en Sources: [Bert Hubert's writings]() Topics: [Machine learning](), [Deep learning](), [Deep neural networks](), [Code](), [data augmentation](), [Transformers]() Tags: [code](), [data-augmentation](), [deep-learning](), [gpu](), [machine-learning](), [neural-networks](), [transformers](), [tutorial]() ### AI overview A GPU-free tutorial that explains modern machine learning and deep learning from the ground up, developing neural networks and a practical demo from scratch. ### Source excerpt A from scratch GPU-free introduction to modern machine learning. Many tutorials exist already of course, but this one aims to really explain what is going on, from the ground up. Also, we'll develop the demo until it is actually useful on real life data which you can supply yourself. Other documents start out from the (very impressive) PyTorch environment, or they attempt to math it up from first principles. Trying to understand deep learning via PyTorch is like trying to learn aerodynamics from flying an Airbus A380. ## Gradient descent: back to basics with F# and C# DevFeed: [Gradient descent: back to basics with F# and C#]() Original publisher: [Read original article]() Author: Andrew Khmylov Published: 2023-02-02T00:00:00Z Content type: tutorial Language: en Sources: [Despite the odds]() Topics: [C#](), [F#](), [implementation](), [Artificial Intelligence](), [Deep neural networks]() Tags: [ai](), [c-sharp](), [code](), [f-sharp](), [implementation](), [neural-networks]() ### AI overview An overview of building a gradient descent engine from scratch, with discussion of using F# and C# and a simplified explanation of neural network training. ### Source excerpt After dabbling with high-level AI tools recently, I wanted to revisit the basics and look under the hood. As they say, the best way to learn something is to build it. I've been following wonderful "from scratch" videos from Andrej Karpathy, where he starts with the essence of all modern AI architectures: gradient descent engine. So, I decided to keep along and build my own engine at https://github.com/khmylov/dumb-gradient, and this article is an overview of its implementation, and some thoughts on using F# and C# for a task like that. Side note: look back at university I completely forgot that one of my university projects back in 2010 was actually about building an image recognition neural network! It was funny to look back at the source code from that era. Turns out, the core ideas are still relevant, it's just computational advances, new modelling discoveries and infrastructure around them that allowed so much progress in the last decade. I'm not going to recite all the details here, the main (much simplified) ideas to comprehend are as follows (I recommend Deep Learning with PyTorch, fast.ai course, and Huggingface course to get a much deeper dive into this) We can represent most real-world data, like text, images, sounds, as vast arrays of numbers Turns out, we can model a lot of "intelligence" tasks (natural language processing, image recognition, etc.) with sophisticated mathematical functions operating on our numeric representation of real-world data Universal approximation theorem was a break-through postulating that even most sophisticated functions can be actually modeled by combining only primitive linear f(x) = a * x + b and basic non-linear functions such as tanh into some inter-connected network (here come the inevitable neuron models and neural networks), we just need a huge number of them to do that, and some computation power to figure out the arguments for those functions "Training" such networks is figuring out how to tweak randomly initialized ## Understanding Deep Learning Algorithms that Leverage Unlabeled Data, Part 1: Self-training DevFeed: [Understanding Deep Learning Algorithms that Leverage Unlabeled Data, Part 1: Self-training]() Original publisher: [Read original article]() Author: A Href; Colin Wei; Jeff Z HaoChen; Tengyu Ma Published: 2022-02-24T08:00:00Z Content type: article Language: en Sources: [The Stanford AI Lab Blog]() Topics: [Deep neural networks](), [Algorithms, Complexity](), [datasets](), [Representation learning](), [ImageNet]() Tags: [algorithms](), [analysis](), [data](), [datasets](), [deep-learning](), [image](), [learning](), [model](), [predictions](), [research](), [training]() ### AI overview This article explains self-training, a semi-supervised learning method that uses a classifier to generate pseudo-labels for unlabeled data and then retrains a new classifier on those labels. It introduces the method and presents a theoretical analysis of why retraining can improve accuracy over the original pseudo-labeler. ### Source excerpt Deep models require a lot of training examples, but labeled data is difficult to obtain. This motivates an important line of research on leveraging unlabeled data, which is often more readily available. For example, large quantities of unlabeled image data can be obtained by crawling the web, whereas labeled datasets such as ImageNet require expensive labeling procedures. In recent empirical developments, models trained with unlabeled data have begun to approach fully-supervised performance (e.g., Chen et al., 2020, Sohn et al., 2020). This series of blog posts will discuss our theoretical work which seeks to analyze recent empirical methods which use unlabeled data. In this first post, we'll analyze self-training, which is a very impactful algorithmic paradigm for semi-supervised learning and domain adaptation. In Part 2, we will use related theoretical ideas to analyze self-supervised contrastive learning algorithms, which have been very effective for unsupervised representation learning. Background: self-training We will first provide a basic overview of self-training algorithms, which are the main focus of this blog post. The core idea is to use some pre-existing classifier \(F_{pl}\) (referred to as the "pseudo-labeler") to make predictions (referred to as "pseudo-labels") on a large unlabeled dataset, and then retrain a new model with the pseudo-labels. For example, in semi-supervised learning, the pseudo-labeler is obtained from training on a small labeled dataset, and is then used to predict pseudo-labels on a larger unlabeled dataset. A new classifier \(F\) is then retrained from scratch to fit the pseudo-labels, using additional regularization. In practice, \(F\) will often be more accurate than the original pseudo-labeler \(F_{pl}\) (Lee 2013). The self-training procedure is depicted below. It is quite surprising that self-training can work so well in practice, given that we retrain on our own predictions, i.e. the pseudo-labels, but not the true labels. ## KotlinDL 0.2: Functional API, зоопарк моделей c ResNet и MobileNet, DSL для обработки изображений DevFeed: [KotlinDL 0.2: Functional API, зоопарк моделей c ResNet и MobileNet, DSL для обработки изображений]() Original publisher: [Read original article]() Author: zaleslaw (JetBrains) Published: 2021-05-25T08:54:42Z Content type: release Language: ru Sources: [JetBrains RU]() Topics: [Kotlin](), [Deep neural networks](), [Maven Central](), [Tensorflow](), [Keras](), [ImageNet]() Tags: [dataset](), [deep-learning](), [image-processing](), [keras](), [kotlin](), [machine-learning](), [maven-central](), [neural-networks](), [tag-00486ebd8d87](), [tag-9d8cf70dc46c](), [tensorflow]() ### AI overview This article presents KotlinDL 0.2, a deep-learning library written in Kotlin with TensorFlow used for low-level computations. The release adds a Functional API for graph-based neural networks, an image-preprocessing DSL, new dataset types, and a model zoo containing ResNet, MobileNet, and VGG models. KotlinDL 0.2 is available on Maven Central. ### Source excerpt Представляем вам версию 0.2 библиотеки KotlinDL. Это библиотека глубокого обучения, где для низкоуровневых вычислений используется TensorFlow, но с высокоуровневым API и логикой, написанными на Kotlin. KotlinDL 0.2 теперь доступен на Maven Central (до этого он лежал на bintray, но закатилось солнышко земли опенсорсной). Появилось столько всего нового: новые слои, специальный DSL для препроцессинга изображений, новые типы датасетов, зоопарк моделей с несколькими моделями из семейства ResNet, MobileNet и старой доброй моделью VGG (рабочая лошадка, впрочем). В этой статье мы коснемся самых главных изменений релиза 0.2. Полный список изменений доступен по ссылке. Читать далее [Next page]()