# Predictions Published articles for Predictions. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Creating an AI Platform for classic ML online inference DevFeed: [Creating an AI Platform for classic ML online inference]() Original publisher: [Read original article]() Author: Rohith Leeladharan Published: 2026-09-10T07:26:46Z Content type: tutorial Language: en Sources: [Amex GBT Technology]() Topics: [AI Platform](), [Inference](), [MLOps](), [Machine learning](), [model-deployment]() Tags: [ai](), [ai-platform](), [ai-platform-engineering](), [deploy](), [experiments](), [feature-store](), [inference](), [machine-learning](), [ml](), [mlops](), [monitoring](), [predictions]() ### AI overview This article describes how American Express Global Business Travel built an AI platform for deploying classic machine-learning systems and supporting online inference. It explains the platform's requirements--simplicity, self-service, experimentation, and continuous improvement--and details the pre-process, predict, post-process pattern used by inference engines. ### Source excerpt Introduction In 2021, we were given the mission to have AI Systems running in production. The team, instead of just following a classical MLOps process, that involves transforming a Jupyter notebook into a product running in production, decided to go further by creating a platform to deploy AI systems in production. The team decided the platform should respect these requirements: Simplicity: The code powering AI systems should be simple, readable, and easy to maintain -- less intricacy means fewer bugs in production and greater reliability. Self-service: Anyone should be able to build and deploy AI systems autonomously, without depending on a central team. Experimentation: The platform should make it easy to run and iterate on experiments. Continuous improvement: Data related to events and interactions within AI systems must be captured, enabling monitoring and continuous improvement over time. In this article, we will walk through the work done to build a platform that fulfills these four requirements. Background At American Express Global Business Travel, we use machine learning (ML) models for a variety of user experiences like ranking hotel and flight search results. Our ML models are wrapped in inference engines that handle both pre-processing of input data before we run a prediction with the model, and post-processing of output data before returning the output to the caller. The overall flow looks something like this: Figure 1: Handling an inference request A client service that would like the ML model's predictions provides necessary context about the request like which user the request is for. Then, optionally, the inference engine fetches any necessary features for inference from our feature store [part 1][part 2]. Finally, it pre-processes the data, runs the predictions using the trained ML model, and does any necessary post-processing of the model output before returning the response to the caller. We call this the pre-process, predict, post-process patter ## Quiz: Python AI: How to Build a Neural Network & Make Predictions DevFeed: [Quiz: Python AI: How to Build a Neural Network & Make Predictions]() Original publisher: [Read original article]() Author: Real Python Published: 2026-09-03T12:00:00Z Content type: article Language: en Sources: [Real Python]() Topics: [Python](), [Neural Network](), [Artificial Intelligence](), [math](), [dataset]() Tags: [ai](), [math](), [neural-networks](), [predictions](), [python](), [train](), [training]() ### AI overview An interactive 13-question quiz testing understanding of how to build a neural network and make predictions with Python AI. It covers input vectors, layers, weights, bias, dot products, sigmoid activation, mean squared error, and backpropagation. ### Source excerpt Check your grasp of how neural networks make predictions in Python, from dot products and activation functions to gradient descent and backpropagation. ## Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions DevFeed: [Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions]() Original publisher: [Read original article]() Author: Ryan Peterman Published: 2026-08-24T13:05:33Z Content type: opinion Language: en Sources: [The Developing Dev]() Topics: [Robotics](), [Humanoid Robots](), [Physical AI](), [Machine learning](), [Large Language Model]() Tags: [china](), [future](), [humanoid-robots](), [llms](), [machine-learning](), [physical-ai](), [predictions](), [robotics]() ### AI overview An interview with robotics researcher Sergey Levine examines the current state of humanoid robotics, China's robotics ecosystem, possible competitors, future timelines, and how humanoid robots may be deployed. Levine discusses the importance of scaling the right technology, data, and model size. ### Source excerpt Sergey Levine is one of the world's top robotics researchers and co-founder of Physical Intelligence. ## How Fyxer built an AI executive assistant people trust DevFeed: [How Fyxer built an AI executive assistant people trust]() Original publisher: [Read original article]() Published: 2026-08-13T12:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Artificial Intelligence](), [OpenAI](), [Fine-tuning](), [text-generation]() Tags: [ai](), [email](), [fine-tuning](), [memory](), [models](), [openai](), [predictions](), [startup](), [workflow]() ### AI overview Fyxer built an AI executive assistant that uses OpenAI models, fine-tuning, memory, and user feedback to organize inboxes and draft emails in each user's voice. Its system divides email tasks among dozens of specialized models that use context to predict whether a reply is needed and generate suitable responses. ### Source excerpt Fyxer uses OpenAI models, fine-tuning, memory, and real user feedback to organize inboxes and draft emails in each user's voice. ## 10 Platform engineering predictions for 2026 DevFeed: [10 Platform engineering predictions for 2026]() Original publisher: [Read original article]() Author: Luca Galante Published: 2026-07-23T05:40:01Z Content type: article Language: en Sources: [Platform Engineering Blog]() Topics: [Platform Engineering](), [Artificial Intelligence](), [Security](), [Architecture & Design](), [Vibe coding](), [Kubernetes](), [Terraform]() Tags: [2026](), [agentic](), [agentic-ai](), [ai](), [ai-agents](), [architecture](), [code](), [code-generation](), [coding](), [finops](), [iam](), [infrastructure](), [kubernetes](), [models](), [platform](), [platform-engineering](), [predictions](), [security](), [terraform](), [vulnerabilities]() ### AI overview The article presents ten predictions for platform engineering in 2026, focusing on the convergence of AI, security-by-design, developer experience, observability, and FinOps. It argues that platform teams will increasingly govern autonomous AI agents, secure AI-generated infrastructure code, and use AI-driven optimization to re-architect systems for cost and latency targets. ### Source excerpt Explore the 10 platform engineering predictions for 2026. See how AI, FinOps, and new security models will fundamentally redefine the platform team's role and architecture. ## Devavrat Shah's research and Ikigai Labs use tabular data for real-time forecasting and decision-making DevFeed: [Devavrat Shah's research and Ikigai Labs use tabular data for real-time forecasting and decision-making]() Original publisher: [Read original article]() Author: David Chandler | Laboratory for Information and Decision Systems Published: 2026-07-14T20:25:00Z Content type: article Language: en Sources: [MIT AI News]() Topics: [Artificial Intelligence](), [decision-making](), [Structured-data](), [AI Models](), [data](), [Time Series](), [Computer science](), [Electrical engineering and computer science (EECS)](), [Temporal data](), [systems]() Tags: [ai](), [ai-and-decision-making](), [ai-in-business-planning](), [ai-models](), [algorithms](), [artificial-intelligence](), [business-and-management](), [business-modeling](), [celonis-context-model](), [computer-science-and-technology](), [data](), [data-systems](), [decision-making](), [devavrat-shah](), [electrical-engineering-and-computer-science-eecs](), [faculty](), [forecasting](), [idss](), [ikigailabs](), [laboratory-for-information-and-decision-systems-lids](), [mit-eecs](), [mit-idss](), [mit-intellectual-property](), [mit-lids](), [mit-schwarzman-college-of-computing](), [predictions](), [profile](), [real-time](), [research](), [school-of-engineering](), [startups](), [structured](), [tabular-data](), [time-series-data]() ### AI overview MIT Professor Devavrat Shah's research led to a foundation model for tabular and time-series enterprise data. Developed with Ikigai Labs, the system continuously tests predictions against real outcomes to support large-scale forecasting and decision-making. ### Source excerpt Through research and entrepreneurship, Professor Devavrat Shah is helping to design methods that can handle constant decision-making using limited computational resources. ## A better way to model the behavior of metal alloys DevFeed: [A better way to model the behavior of metal alloys]() Original publisher: [Read original article]() Author: Zach Winn | MIT News Published: 2026-06-19T18:00:00Z Content type: news Language: en Sources: [MIT AI News]() Topics: [Simulation](), [Machine learning](), [datasets]() Tags: [artificial-intelligence](), [chemical-engineering](), [chemistry](), [computer-modeling](), [computer-science-and-technology](), [daniel-xiao](), [datasets](), [dmse](), [killian-sheriff](), [lewis-r-owen](), [machine-learning](), [materials-science-and-engineering](), [mit-materials-science-and-engineering](), [models](), [phase-diagrams](), [predicting-new-materials](), [predictions](), [research](), [rodrigo-freitas](), [school-of-engineering](), [simulation](), [yifan-cao]() ### AI overview MIT researchers developed a machine-learning approach that uses diverse training datasets to model chemically complex metal alloys and predict their material properties more accurately across different conditions. ### Source excerpt MIT researchers' approach captures subtle atomic patterns, improving predictions of material properties. ## 5 predictions about agentic AI and analytics in 2026 DevFeed: [5 predictions about agentic AI and analytics in 2026]() Original publisher: [Read original article]() Author: Kristin Crosier Published: 2026-05-12T00:00:00Z Content type: opinion Language: en Sources: [Redpanda]() Topics: [Artificial Intelligence](), [data](), [Security]() Tags: [2026](), [agentic](), [agentic-ai](), [ai](), [analytics](), [aws](), [collaboration](), [cto](), [executive](), [gartner](), [governance](), [predictions](), [thought-leadership]() ### AI overview The article presents five predictions about agentic AI and analytics in 2026, drawing on Gartner's research and Peter Corless's industry observations. It emphasizes the opportunities and governance risks of agentic systems, along with the need for executive accountability, cross-team collaboration, and new skills and best practices. ### Source excerpt Learn about what's top of mind among enterprises planning for agentic systems, and top AI predictions for 2026 and beyond. ## Zeta2.1: 3x Fewer Tokens, 50ms Faster DevFeed: [Zeta2.1: 3x Fewer Tokens, 50ms Faster]() Original publisher: [Read original article]() Author: Ben Kunkle, Oleksiy Syvokon Published: 2026-05-08T00:00:00Z Content type: release Language: en Sources: [Zed Industries - Blog]() Topics: [model-deployment](), [hugging face](), [Open Source](), [Code](), [D]() Tags: [efficiency](), [hugging-face](), [model](), [open-source](), [predictions](), [production](), [pypi](), [release](), [running](), [rust](), [tokens]() ### AI overview Zed announces Zeta2.1, an open-weight edit prediction model that uses a new Multi-Region prompt format. Compared with Zeta2, it emits fewer output tokens, delivers faster predictions, and requires fewer servers for the same traffic. The release also supports local use and includes Rust bindings published to PyPI. ### Source excerpt Zeta2.1 is out and it's more efficient than Zeta2. ## How We Developed Zeta2 DevFeed: [How We Developed Zeta2]() Original publisher: [Read original article]() Author: Oleksiy Syvokon, Ben Kunkle Published: 2026-04-07T00:00:00Z Content type: article Language: en Sources: [Zed Industries - Blog]() Topics: [Development](), [Code](), [Large Language Model](), [datasets](), [Testing](), [Training AI Models](), [GitHub]() Tags: [code](), [context](), [development](), [evaluation](), [examples](), [github](), [llm](), [model](), [predictions](), [synthetic](), [testing](), [train](), [training]() ### AI overview Zed describes how it developed Zeta2, a faster edit-prediction model trained through knowledge distillation. The process used richer code context, ethically collected starting states, teacher-prompt evaluation, and roughly 100,000 synthetic examples derived from GitHub commits. ### Source excerpt A deep dive into how we improved our Zeta edit predictions model, Zeta2. ## 5 Key Predictions for the Data Industry in 2026 DevFeed: [5 Key Predictions for the Data Industry in 2026]() Original publisher: [Read original article]() Author: SeattleDataGuy Published: 2026-01-31T19:29:24Z Content type: opinion Language: en Sources: [SeattleDataGuy's Newsletter]() Topics: [data](), [Microsoft](), [Artificial Intelligence](), [export](), [Azure]() Tags: [2026](), [ai](), [azure](), [data](), [export](), [microsoft](), [predictions]() ### AI overview A commentary piece offers predictions about the data industry over the next year or two, including a possible Microsoft Fabric rebrand and the continued gap between demand for AI and companies' reliance on traditional data workflows such as ERP exports to Excel, SFTP, and APIs. ### Source excerpt Hype Cycles, Rebrands, and the Messy Reality of Data ## Why AI Could Increase Demand for Skilled Developers DevFeed: [Why AI Could Increase Demand for Skilled Developers]() Original publisher: [Read original article]() Published: 2026-01-11T00:00:00Z Content type: opinion Language: en Sources: [Kt. Academy]() Topics: [Development](), [Artificial Intelligence]() Tags: [ai](), [business](), [developers](), [development](), [predictions](), [workshop-learning-programming]() ### AI overview This opinion article argues that AI tools will increase, rather than reduce, demand for skilled developers. It attributes current weakness in the developer market to broader business and industry conditions, including cost-cutting and reduced development budgets, while treating the Twitter example as suggestive rather than conclusive. ### Source excerpt Why the rise of AI means an increased demand for skilled developers. ## 2026 Cybersecurity Predictions by Teleport CEO Ev Kontsevoy DevFeed: [2026 Cybersecurity Predictions by Teleport CEO Ev Kontsevoy]() Original publisher: [Read original article]() Author: ev@goteleport.com (Ev Kontsevoy) Published: 2025-12-18T00:00:00Z Content type: opinion Language: en Sources: [Teleport]() Topics: [Cybersecurity](), [IAM](), [Security & Privacy](), [MFA]() Tags: [2026](), [ai](), [cybersecurity](), [identity](), [identity-management](), [non-human-identity](), [predictions](), [security]() ### AI overview Teleport CEO Ev Kontsevoy presents predictions for 2026 cybersecurity, arguing that AI identities will drive a unified approach to identity management, greater engineering involvement in security, and more granular definitions of agentic AI. ### Source excerpt AI has broken identity security--2026 is when the cracks become impossible to ignore. Teleport CEO Ev Kontsevoy shares predictions and what leaders should do. ## AlphaFold: Five years of impact DevFeed: [AlphaFold: Five years of impact]() Original publisher: [Read original article]() Author: Demis Hassabis; John Jumper; Pushmeet Kohli; Anna Koivuniemi; AlphaFold team Published: 2025-11-25T16:00:12Z Content type: article Language: en Sources: [Google DeepMind News]() Topics: [Artificial Intelligence](), [Bioinformatics]() Tags: [ai](), [drug-discovery](), [explore](), [predictions](), [research](), [science]() ### AI overview The article examines five years of AlphaFold's impact on biological research, describing faster discovery, more novel protein structures, and increased influence on clinical research and patents. It also presents AlphaFold 3, Isomorphic Labs, and the AlphaFold Server as developments advancing AI-assisted drug discovery and digital biology. ### Source excerpt Explore how AlphaFold has accelerated science and fueled a global wave of biological discovery. ## Beyond classification: How AI agents are evolving Shopify's product taxonomy at scale DevFeed: [Beyond classification: How AI agents are evolving Shopify's product taxonomy at scale]() Original publisher: [Read original article]() Author: Kshetrajna Raghavan Published: 2025-10-09T05:18:20Z Content type: article Language: en Sources: [Shopify Engineering](), [Shopify Engineering - Shopify Engineering]() Topics: [Shopify](), [Artificial Intelligence](), [vlm]() Tags: [agents](), [ai](), [ai-agents](), [classification](), [predictions](), [products](), [shopify]() ### AI overview Shopify describes an AI multi-agent system that evolves its product taxonomy alongside changing commerce. The system is intended to improve more than product classification by updating taxonomy labels across more than 10,000 categories and 2,000 attributes, addressing scale, domain expertise, and consistency challenges. ### Source excerpt Last year, over 875 million people bought items from Shopify merchants. Building on our prior Vision Language Model-based product classification, this post explores how AI agents are evolving the taxonomy itself. ## Introduction to ML monitoring DevFeed: [Introduction to ML monitoring]() Original publisher: [Read original article]() Author: Başak Tuğçe Eskili Published: 2025-08-05T21:56:29Z Content type: tutorial Language: en Sources: [MarvelousMLOps]() Topics: [databricks](), [MLOps](), [Monitoring](), [data]() Tags: [article](), [course](), [data](), [databricks](), [errors](), [introduction](), [latency](), [ml](), [mlops](), [model](), [monitoring](), [performance](), [predictions](), [production](), [workflows]() ### AI overview An introductory lecture in the MLOps with Databricks course explains why machine learning systems require monitoring beyond system health, errors, latency, KPIs, and infrastructure costs. It introduces data drift and concept drift as causes of model performance degradation, even when code and infrastructure remain unchanged. ### Source excerpt Lecture 9 of MLOps with Databricks course ## Evolution of Product Classification at Shopify: From Categories to Comprehensive Product Understanding DevFeed: [Evolution of Product Classification at Shopify: From Categories to Comprehensive Product Understanding]() Original publisher: [Read original article]() Author: Kshetrajna Raghavan Published: 2025-05-08T04:00:00Z Content type: article Language: en Sources: [Shopify Engineering](), [Shopify Engineering - Shopify Engineering]() Topics: [Shopify](), [Artificial Intelligence](), [AI Development](), [Machine learning](), [data]() Tags: [ai](), [business](), [framework](), [machine-learning](), [models](), [predictions](), [product](), [search](), [shopify]() ### AI overview Shopify describes the evolution of its product classification system from a 2018 logistic-regression model using TF-IDF to a multimodal system and, by 2023, a product-understanding system built on Vision Language Models and the Shopify Product Taxonomy. The system classifies products, extracts category-specific attributes and metadata, and supports search, discovery, recommendations, safety, and trust. ### Source excerpt Shopify's product classification system has evolved from basic categorization to an AI-driven framework using Vision Language Models and the Shopify Product Taxonomy. It accurately classifies products, extracts attributes, and processes over 30 million predictions daily, improving search, discovery, trust, and efficiency for merchants and buyers. ## Zed now predicts your next edit with Zeta, our new open model DevFeed: [Zed now predicts your next edit with Zeta, our new open model]() Original publisher: [Read original article]() Author: Antonio Scandurra, Nathan Sobo Published: 2025-02-13T00:00:00Z Content type: release Language: en Sources: [Zed Industries - Blog]() Topics: [Language models](), [Open Source](), [datasets](), [Open Data]() Tags: [data](), [language](), [model](), [open-source](), [predictions](), [speed](), [tool]() ### AI overview Zed introduces edit prediction, a feature powered by Zeta, an open-source language model derived from Qwen2.5-Coder-7B. During the public beta, users can accept predicted edits with keyboard shortcuts, and Zed describes how predictions interact with language-server completions on macOS and Linux. Zeta includes an open dataset, with submitted data reviewed for safety and privacy before publication. ### Source excerpt A tool that predicts your next move. Powered by Zeta, our new open-source, open-data language model. ## Tech predictions for 2025 DevFeed: [Tech predictions for 2025]() Original publisher: [Read original article]() Author: Mads Brodt Published: 2025-01-03T13:43:36Z Content type: opinion Language: en Sources: [Front-End Focus]() Topics: [Front end](), [Frameworks](), [React](), [Astro](), [Angular](), [Vite](), [Svelte](), [Vue.js](), [Artificial Intelligence](), [Job](), [Qwik]() Tags: [2025](), [adoption](), [ai](), [angular](), [astro](), [build-tool](), [community](), [components](), [developers](), [job](), [predictions](), [qwik](), [react](), [release](), [svelte](), [tailwind](), [tech](), [ui](), [vite](), [vue]() ### AI overview An opinion piece forecasting front-end development trends for 2025. It discusses the continued prominence of React-based production stacks, the potential standardization of React, NextJS, Tailwind, and ShadCN UI, the effects of framework competition and AI on developers, and expected growth for Astro and Vite/Rolldown. ### Source excerpt Happy New Year! ## 2024 Predictions from Teleport CEO Ev Kontsevoy DevFeed: [2024 Predictions from Teleport CEO Ev Kontsevoy]() Original publisher: [Read original article]() Author: ev@goteleport.com (Ev Kontsevoy) Published: 2024-02-07T00:00:00Z Content type: opinion Language: en Sources: [Teleport]() Topics: [Cybersecurity](), [Security & Privacy](), [Authentication](), [Architecture & Design]() Tags: [api-keys](), [authentication](), [browser](), [cloud-computing](), [cybersecurity](), [identity](), [identity-and-access](), [predictions](), [secrets](), [security](), [social-engineering](), [tokens]() ### AI overview Teleport CEO Ev Kontsevoy's 2024 predictions describe rising identity-focused attacks, closer collaboration between engineering and security teams, and broader adoption of secretless authentication to reduce risks from exposed credentials and human error. ### Source excerpt Teleport CEO forecasts 2024 cybersecurity shifts: secretless access, evolving security roles, M&A consolidation, and least privilege strategies. ## Evaluating Binary Classification Metrics on Imbalanced Datasets DevFeed: [Evaluating Binary Classification Metrics on Imbalanced Datasets]() Original publisher: [Read original article]() Author: Angel Igareta Published: 2023-11-09T09:22:53Z Content type: tutorial Language: en Sources: [Klarna Engineering]() Topics: [Data Science](), [Machine Learning & Artificial Intelligence](), [data]() Tags: [classification](), [data-science](), [datasets](), [evaluation](), [false-positives](), [imbalanced-data](), [machine-learning](), [model](), [predictions]() ### AI overview This tutorial examines GINI and ROC_AUC for evaluating binary classification models, explaining why their effectiveness can be compromised by imbalanced datasets and presenting the Precision-Recall curve as a potentially more robust alternative. ### Source excerpt Discover how the Precision-Recall curve can provide a more robust metric for binary classification in data science and machine learning. Imagine stepping into the complex world of binary classification problems. As a Senior Data Scientist at Klarna, this is my day-to-day reality. Binary classification is a cornerstone of data science, with applications touching everything from credit default predictions to medical diagnoses and spam detection. Yet, these problems come with their own unique set of challenges. Metrics such as the GINI coefficient and ROC_AUC often serve as our compass in this maze. They are widely trusted and used for evaluating models. But here's the catch: they might not always point us in the right direction. Can we rely on them blindly, or do we need to dig deeper? The path gets even more challenging when we encounter imbalanced datasets. In such cases, the effectiveness of our trusted metrics can be seriously compromised. In this post, I invite you to join me on a journey to explore these metrics in greater depth. We will question their effectiveness, understand their limitations, and reveal alternatives that could prove to be more reliable navigational tools in the world of binary classification problems. Understanding Model Predictions and Metrics To truly grasp the nuances of model evaluation, let's start by setting the stage with a real-world scenario that we often encounter at Klarna. Imagine we're tasked with predicting customer loan defaults. We have two categories to consider -- paid or default. However, in our scenario, the default rate is a mere 2%. This is a classic case of data imbalance, and it's exactly the kind of challenge we're up against. To evaluate our model's performance in this scenario, we need to understand its predictions. We break these down into four distinct outcomes, also known as the confusion matrix: True Positives (TP): These are the customers who our model correctly identifies as defaulters. False Positives (FP): T ## On-device fetal ultrasound assessment with TensorFlow Lite DevFeed: [On-device fetal ultrasound assessment with TensorFlow Lite]() Original publisher: [Read original article]() Author: TensorFlow Blog (noreply@blogger.com) Published: 2023-06-20T16:00:00Z Content type: article Language: en Sources: [The TensorFlow Blog]() Topics: [TensorFlow Lite](), [Google](), [Artificial Intelligence](), [Inference](), [Mobile](), [Protocol (disambiguation)]() Tags: [ai](), [devices](), [google](), [health](), [healthcare](), [mobile](), [on-device](), [open-source](), [predictions](), [protocol](), [smartphones](), [tensorflow-lite]() ### AI overview This article explains how TensorFlow Lite and Google Research AI models are being used with portable, smartphone-integrated ultrasound devices to help health workers assess gestational age and fetal presentation in under-resourced settings. The models analyze ultrasound video collected through a blind sweep protocol and are intended to expand access to clinically useful prenatal screening. ### Source excerpt Posted by Angelica Willis and Akib Uddin, Health AI Team, Google Research How researchers at Google are working to expand global access to maternal healthcare with the help of AI TensorFlow Lite* is an open-source framework to run machine learning models on mobile and edge devices. It's popular for use cases ranging from image classification, object detection, speech recognition, natural language tasks, and more. From helping parents of deaf children learn sign language, to predicting air quality, projects using TensorFlow Lite are demonstrating how on-device ML could directly and positively impact lives by making these socially beneficial applications of AI more accessible, globally. In this post, we describe how TensorFlow Lite is being used to help develop ultrasound tools in under-resourced settings. Motivation According to the WHO, complications from pregnancy and childbirth contribute to roughly 287,000 maternal deaths and 2.4 million neonatal deaths worldwide each year. As many as 95% of these deaths occur in under-resourced settings and many are preventable if detected early. Obstetric diagnostics, such as determining gestational age and fetal presentation, are important indicators in planning prenatal care, monitoring the health of the birthing parent and fetus, and determining when intervention is required. Many of these factors are traditionally determined by ultrasound. Advancements in sensor technology have made ultrasound devices more affordable and portable, integrating directly with smartphones. However, ultrasound requires years of training and experience, and, in many rural or underserved regions, there is a shortage of trained ultrasonography experts, making it difficult for people to access care. Due to this global lack of availability, it has been estimated that as many as two-thirds of pregnant people in these settings do not receive ultrasound screening during pregnancy. Expanding access by enabling non-experts Google Research is building AI m ## Visualizing and interpreting decision trees DevFeed: [Visualizing and interpreting decision trees]() Original publisher: [Read original article]() Author: TensorFlow Blog (noreply@blogger.com) Published: 2023-06-06T17:00:00Z Content type: tutorial Language: en Sources: [The TensorFlow Blog]() Topics: [Machine Learning & Artificial Intelligence](), [Model Development](), [Google]() Tags: [article](), [google](), [how-to](), [learn](), [learning](), [machine-learning](), [model](), [predictions](), [tensorflow](), [tensorflow-decision-forests](), [training](), [visualization]() ### AI overview A tutorial on visualizing and interpreting decision trees with the dtreeviz library and TensorFlow Decision Forests. It explains how decision trees use feature-based splits and leaf nodes to make regression or classification predictions, then demonstrates how visualization can clarify those predictions and the distribution of training instances. ### Source excerpt Posted by Terence Parr, Google Decision trees are the fundamental building block of Gradient Boosted Trees and Random Forests, the two most popular machine learning models for tabular data. To learn how decision trees work and how to interpret your models, visualization is essential. TensorFlow recently published a new tutorial that shows how to use dtreeviz, a state-of-the-art visualization library, to visualize and interpret TensorFlow Decision Forest Trees. The dtreeviz library, first released in 2018, is now the most popular visualization library for decision trees. The library is constantly being updated and improved, and there is a large community of users who can provide support and answer questions. There is a helpful YouTube video and article on the design of dtreeviz. Let's demonstrate how to use dtreeviz to interpret decision tree predictions. At a basic level, a decision tree is a machine learning model that learns the relationship between observations and target values by examining and condensing training data into a binary tree. Each leaf in the decision tree is responsible for making a specific prediction. For regression trees, the prediction is a value, such as price. For classifier trees, the prediction is a target category, such as cancer or not-cancer. Any path from the root of the decision tree to a specific leaf predictor passes through a series of (internal) decision nodes. Each decision node compares a single feature's value with a specific split point value learned during training. Making a prediction means walking from the root down the tree, comparing feature values, until we reach a leaf. Consider the following simple decision tree that tries to classify animals based upon two features, the number of legs and the number of eyes. Let's say that our test animal has four legs and two eyes. To classify the test animal, we start at the root of the tree and compare our test animal's number of legs to four. Since the number of legs is equal to fo ## 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. [Next page]()