DEV Community: deeplearning

SAM.MD: Zero-shot medical image segmentation capabilities of the SegmentAnything Model

Paperium — Tue, 30 Jun 2026 11:50:28 +0000

Monocular Depth Estimation using Diffusion Models

Paperium — Tue, 30 Jun 2026 10:40:28 +0000

Dropout: Switch Off Neurons to Stop Overfitting

Devanshu Biswas — Tue, 30 Jun 2026 10:03:20 +0000

Dropout is almost absurdly simple — randomly switch off neurons during training — yet it was one of the biggest anti-overfitting wins in deep learning. Here's why it works, visualized.

🎲 Watch neurons drop (toggle the rate): https://dev48v.infy.uk/dl/day20-dropout.html

What it does

On each training step, each hidden neuron is kept with probability (1−p) and zeroed out with probability p (say p=0.5). A different random subset drops every step. The demo grays out a fresh random set of neurons each pass and cuts their edges.

Why that helps

Neurons can't rely on any specific other neuron being present, so they can't co-adapt into a fragile memorized solution — each must learn a feature that's useful on its own. It's like training a huge ensemble of subnetworks that share weights. Result: a smaller train/val gap (less overfitting) — which the two accuracy curves in the demo show.

Train vs inference (the gotcha)

You drop during training only. At inference, all neurons are on. To keep the expected activations consistent, inverted dropout scales the kept activations by 1/(1−p) during training, so inference needs no change.

Modern note

With batch norm (Day 19) and huge datasets, dropout is needed less in conv nets — but it's still standard in Transformers (attention + feed-forward). It's regularization, alongside L2 (Day 17).

🔨 Built from scratch (mask = rand > p → scale by 1/(1−p) → off at eval) on the page: https://dev48v.infy.uk/dl/day20-dropout.html

Part of DeepLearningFromZero. 🌐 https://dev48v.infy.uk

Project: Cancer Classification Model

Ваграм Катранян — Tue, 30 Jun 2026 09:56:42 +0000

A year ago, I developed a study prototype of a neural network that combines two types of data:

medical images in DICOM format;
clinical tabular data (patient age, tumor size, biopsy results).

The goal of the model is to analyze both images and numerical data simultaneously to classify cancer presence.

Key Features

Multimodality: the model processes both images and tabular features.
Attention mechanism: highlights the most important features to improve accuracy.
GPU/CPU support: training can be performed on a regular computer or on a GPU.
Evaluation metrics: AUC, F1, Precision, Recall — to measure performance objectively.
Engineering design: separate classes for dataset, model, training, and logging.

In summary: this project gave me hands‑on experience with medical data and showed how Python can be applied not only in backend development but also in machine learning tasks.

Проект: Модель для классификации рака

Ваграм Катранян — Tue, 30 Jun 2026 09:53:12 +0000

Год назад я разработал учебный прототип нейросети, которая объединяет два типа данных:

медицинские снимки в формате DICOM;
клинические табличные данные (возраст пациента, размер опухоли, результаты биопсии).

Задача модели — анализировать изображения и цифры одновременно, чтобы классифицировать наличие рака.

Особенности реализации

Мультимодальность: модель работает сразу с изображениями и табличными признаками.
Механизм внимания (attention): помогает выделять наиболее важные признаки и повышает точность.
** Поддержка GPU/CPU**: обучение возможно как на обычном компьютере, так и на графическом процессоре.
Метрики качества: AUC, F1, Precision, Recall — для объективной оценки работы модели.
** Инженерная структура**: отдельные классы для датасета, модели, обучения и логирования.

Вывод

проект дал мне опыт работы с медицинскими данными и показал, как Python можно применять в задачах машинного обучения.

Data Science vs AI: Which Field Has Better Career Growth in 2026?

Subhalaxmi Paikaray — Tue, 30 Jun 2026 09:47:31 +0000

Artificial Intelligence (AI) and Data Science are two of the fastest-growing technology domains today. From startups to Fortune 500 companies, organizations are investing heavily in intelligent systems, predictive analytics, and automation. As a result, students often ask one important question:

Should I choose Data Science or Artificial Intelligence?

The truth is, there isn't a universal answer. Both fields offer exciting career opportunities, competitive salaries, and long-term growth. However, they focus on different skills, solve different business problems, and lead to different career paths.

If you're planning a career in technology, this guide will help you understand the differences and decide which field aligns with your goals.

Understanding Data Science

Data Science focuses on extracting meaningful insights from data. Every day, businesses collect enormous amounts of information—from customer behavior and sales performance to website traffic and financial transactions.

A Data Scientist analyzes this data to answer important business questions and support decision-making.

Typical responsibilities include:

Collecting and cleaning data
Building dashboards
Performing statistical analysis
Identifying business trends
Creating predictive models
Visualizing insights

Popular tools include:

Python
SQL
R
Power BI
Tableau
Excel
Apache Spark

If you enjoy mathematics, statistics, and problem-solving, Data Science can be an excellent career choice.

Understanding Artificial Intelligence

Artificial Intelligence focuses on building systems that can perform tasks requiring human intelligence.

Rather than simply analyzing data, AI enables machines to:

Learn from experience
Understand language
Recognize images
Generate content
Make predictions
Automate complex workflows

AI professionals often work with technologies such as:

Machine Learning
Deep Learning
Natural Language Processing (NLP)
Computer Vision
Large Language Models (LLMs)
Generative AI
AI Agents

The rapid adoption of Agentic AI, AI Copilots, Multimodal AI, and Generative AI is creating exciting opportunities for developers and engineers worldwide.

Key Differences Between Data Science and AI

Although these fields overlap, their primary objectives differ.

Data Science focuses on understanding and interpreting data to drive business decisions.

Artificial Intelligence focuses on building intelligent systems that learn, automate, and interact with users.

Think of it this way:

Data Science answers: "What happened and why?"
AI answers: "How can machines solve this problem automatically?"

Understanding this distinction makes it easier to choose the right learning path.

Which Skills Should You Learn?

For Data Science

Employers often look for professionals who understand:

Statistics
Probability
Python
SQL
Data Visualization
Business Analytics
Machine Learning basics
Dashboard Development

Strong analytical thinking is one of the biggest advantages in this field.

For Artificial Intelligence

AI professionals typically require knowledge of:

Python Programming
Machine Learning
Deep Learning
Neural Networks
Prompt Engineering
TensorFlow
PyTorch
AI Model Deployment
Large Language Models (LLMs)

As AI technologies continue to evolve, continuous learning becomes an essential part of every AI career.

Career Opportunities

Data Science Roles

Popular job profiles include:

Data Scientist
Data Analyst
Business Intelligence Analyst
Analytics Consultant
Data Engineer
Product Analyst

These professionals work across industries such as finance, healthcare, e-commerce, education, logistics, and retail.

AI Career Roles

Artificial Intelligence opens opportunities such as:

AI Engineer
Machine Learning Engineer
NLP Engineer
Computer Vision Engineer
Generative AI Developer
AI Solutions Architect
Robotics Engineer

The increasing demand for AI-powered products means these roles are expected to remain highly valuable over the coming years.

Which Field Has Better Career Growth?

Both careers offer strong growth, but the answer depends on your interests.

Choose Data Science if you enjoy:

Working with data
Finding business insights
Building dashboards
Solving analytical problems
Supporting business strategy

Choose Artificial Intelligence if you enjoy:

Programming
Building intelligent applications
Automation
Robotics
Generative AI
Developing AI-powered software

In reality, many companies now expect professionals to understand both disciplines. A Data Scientist often applies Machine Learning models, while an AI Engineer frequently works with large datasets.

The boundaries between these fields are becoming increasingly interconnected.

Industry Trends to Watch in 2026

Technology is evolving rapidly, and several trends are shaping the future of both AI and Data Science.

Some of the biggest trends include:

Generative AI
AI Agents
AI-Assisted Development
Responsible AI
Explainable AI (XAI)
MLOps
Predictive Analytics
Edge AI
Real-Time Data Processing
AI Automation

Students who stay updated with these technologies will have a competitive advantage in the job market.

Why Practical Learning Matters

Learning theories alone isn't enough to build a successful technology career.

Employers increasingly prefer candidates who can demonstrate practical experience through:

Capstone projects
Hackathons
GitHub portfolios
Open-source contributions
AI model development
Data analytics dashboards
Industry internships

Hands-on learning helps students understand how technologies are applied in real business environments.

Recognizing this industry shift, institutions such as the Regional College of Management (RCM) are integrating project-based learning, internships, AI-focused coursework, Data Science, Full Stack Development, and industry collaborations into their technology programs. This practical approach helps students develop both technical expertise and workplace-ready skills.

Final Thoughts

Choosing between Data Science and Artificial Intelligence isn't about selecting the "better" field—it's about choosing the one that matches your interests and long-term career goals.

Data Science empowers organizations to make smarter decisions through data, while Artificial Intelligence focuses on building systems that can learn, automate, and solve complex problems.

As businesses continue adopting AI-powered technologies, professionals who combine programming, analytics, machine learning, cloud computing, and problem-solving skills will remain in high demand.

No matter which path you choose, keep learning, build real-world projects, contribute to open-source communities, and stay curious. In today's tech landscape, adaptability is one of the most valuable skills you can have.

What would you choose—Data Science or Artificial Intelligence? Share your thoughts and career goals in the comments below!

Meituan Open-Sources 1.6T-Parameter LongCat-2.0 Trained on Domestic Chips

gentic news — Tue, 30 Jun 2026 09:38:15 +0000

Meituan open-sourced 1.6T-parameter LongCat-2.0 trained on 50,000 domestic ASICs, claiming China's first full-process domestic-chip trillion-parameter model.

Meituan open-sourced LongCat-2.0, a 1.6 trillion-parameter LLM trained entirely on domestic chips. The model claims to be China's first trillion-parameter AI fully pre-trained and inferred on a 50,000-card ASIC cluster.

Key facts

1.6 trillion parameters in LongCat-2.0.
1 million-token context window.
50,000-card domestic ASIC cluster used for training.
DeepSeek V4-pro also has 1.6 trillion parameters.
Open-sourced on Tuesday by Meituan.

Food delivery giant Meituan on Tuesday open-sourced LongCat-2.0, a large language model boasting 1.6 trillion parameters and a 1 million-token context window According to SCMP. The Beijing-based company claimed this is the industry's first trillion-parameter model to complete full-process training and inference on a 50,000-card domestic computing power cluster built with AI ASIC superpods.

Beyond Inference

While DeepSeek's V4-pro (1.6 trillion parameters, launched April 2026) relied on home-grown chips only for inference, Meituan says LongCat-2.0 used domestic hardware for both pre-training and inference. Pre-training is far more computationally intensive — it involves digesting massive datasets to learn basic patterns. This marks a significant step for China's push to move domestic chips beyond inference workloads.

The Hardware Question

Meituan did not disclose the specific ASIC vendor or chip performance metrics. The claim of a 50,000-card cluster raises questions about interconnect efficiency and training stability at scale on non-Nvidia hardware. DeepSeek's V4-pro, by contrast, used domestic chips only for inference — a less demanding task — while likely relying on Nvidia or other foreign GPUs for pre-training, though DeepSeek has not confirmed that.

Open-Source and Context

LongCat-2.0 is open-sourced, following Meituan's earlier LongCat-1.0 release. The 1 million-token context window matches frontier models like DeepSeek V4 (which achieved 500K context with FlashMemory optimization in June 2026) and positions LongCat for long-document and enterprise RAG use cases. Meituan has not published benchmark results on standard evaluations like MMLU, HumanEval, or SWE-Bench.

What to watch

Watch for benchmark results from Meituan on standard evaluations like MMLU, HumanEval, and SWE-Bench. Also track whether DeepSeek responds with a fully domestic-chip pre-training claim for its next model, potentially V5.

Source: scmp.com

Originally published on gentic.news

Quantum Tagging: Authenticating Location via Quantum Information andRelativistic Signalling Constraints

Paperium — Tue, 30 Jun 2026 09:30:28 +0000

Directed clustering in weighted networks: a new perspective

Paperium — Tue, 30 Jun 2026 08:20:28 +0000

ChatGPT: Beginning of an End of Manual Linguistic Data Annotation? Use Case ofAutomatic Genre Identification

Paperium — Tue, 30 Jun 2026 07:10:29 +0000

Modeling and Analysis of Uplink Non-Orthogonal Multiple Access (NOMA) inLarge-Scale Cellular Networks Using Poisson Cluster Proc

Paperium — Tue, 30 Jun 2026 06:00:29 +0000

Role of Digital Twin in Optical Communication: Fault Management, HardwareConfiguration, and Transmission Simulation

Paperium — Tue, 30 Jun 2026 04:50:28 +0000