ELE Times

Decision Tree Learning is a type of supervised machine learning used in classification as well as regression problems. It tries to mimic real-world decision making by representing decisions and their possible outcomes in the form of a tree. Each internal node in the tree denotes a test on a feature, each branch denotes an outcome of the test, and the leaf node gives the final decision. It is easy to understand, requires no complex data preprocessing, and is visually very informative.

Decision tree learning history:

The concept of decision trees has roots in decision analysis and logic, but their formal application in machine learning began in the 1980s. The ID3 algorithm, developed by Ross Quinlan in 1986, was one of the first major breakthroughs in decision tree learning. It introduced the use of information gain as a criterion for splitting nodes. This was followed by C4.5, an improved version of ID3, and CART (Classification and Regression Trees), developed by Breiman et al, which used the Gini index and supported both classification and regression tasks. These algorithms laid the foundation for modern decision tree models used today.

How does decision tree learning work:

Decision tree learning is a type of algorithm in machine learning where data gets split into smaller subsets and gets organized in the form of a tree. The splitting is based on the value of the data features. At the beginning, with the root node, a feature of the data gets selected. This selection feature tends to be the one that gets deemed most informative by the Gini impurity or entropy criteria. As mentioned earlier, internal nodes get to represent a certain decision rule. This process continues until the data is sufficiently partitioned or a stopping condition is met, resulting in leaf nodes that represent final predictions or classifications. The tree structure makes it easy to interpret and visualize how decisions are made step by step.

Types of Decision Trees:

1. Classification Trees

These are utilized when the dependent variable is categorical. Such trees assist in categorizing the dataset into specific categories (e.g., spam and non-spam). Each split aims to enhance class separation based on certain features.

2. Regression Trees

These trees are used when the dependent variable is continuous. Unlike categorization, these trees aim to provide numerical predictions (e.g., house prices). The split in these trees is done for minimizing prediction error.

Examples of Decision Tree Learning:

• Email Filtering: Marking emails as spam or not using keywords and sender details.
• Loan Approval: Deciding loan approval using income, credit score, and employment status.
• Medical Diagnosis: Identifying a disease with the help of symptoms and test results.
• Weather Prediction: Predicting rain using humidity, temperature, and wind speed.

Applications of Decision Tree Learning:

1. Finance

Decision trees analyze customer data and transaction behavior for credit scoring, fraud detection, and risk management.

2. Healthcare

With the use of medical records and test outcomes, they aid in disease diagnosis, treatment suggestions, and patient outcome predictions.

3. Marketing

Segmenting customers, predicting buying behavior, and optimizing campaign strategies based on demographic and behavioral data.

4. Retail

Forecasting sales, managing inventory, and personalizing product recommendations.

5. Education

Predicting student performance, dropout risk, and tailoring learning paths based on academic data.

Decision Tree Learning Advantages:

Decision Tree learning has numerous benefits, all of which contribute to its widespread use in machine learning. It is simple to grasp and analyze because the structure of the tree is akin to human decision-making and can be easily visualised. It can process both numerical and categorical data without the need for advanced data preprocessing or feature scaling. Decision trees are not affected by outliers or missing data, and they can model non-linear patterns in data. It requires very little in the way of data preparation and is immensely powerful and user-friendly because it inherently takes into account feature combinations through its hierarchical splits.

Conclusion:

Decision Tree Learning is going to mature into a dynamic, real-time intelligence system processing complex data, providing direction to autonomous systems, and enabling accountable decision-making in all sectors. These trees will, in time, become self-optimizing systems that reason, tell stories, and co-exist with human cognition, and they will serve as the ethical and intellectual foundation of future AI.

The post Decision Tree Learning Definition, Types, Examples and Applications appeared first on ELE Times.

New solution empowers engineers to meet HDMI Forum compliance standards while optimizing signal integrity and performance across high-bandwidth video applications

Keysight Technologies, Inc. announced the release of its enhanced physical layer compliance test solution for high-definition multimedia interface (HDMI), delivering robust compliance and performance validation capabilities for transmitter [source] and cable devices. The Keysight Electrical Performance, Validation, and Compliance Test Solution for HDMI addresses the growing complexity, and bandwidth demands of modern HDMI applications, including ultra-high definition (UHD) video, high dynamic range (HDR) content, and immersive audio experiences.

With the rising demand for 8K/12K video, HDR content, and high-speed connectivity, engineers face growing challenges in maintaining signal integrity across HDMI interfaces. The recent release of the HDMI 2.2 test specification by the HDMI Forum introduces more stringent compliance requirements for transmitters and cables, highlighting gaps in traditional test coverage. Without robust validation tools, manufacturers risk costly redesigns and certification delays. As HDMI technology advances, the need for comprehensive, automated test solutions is critical to ensure performance, reliability, and faster time-to-market.

As the HDMI ecosystem evolves to support higher resolutions, faster refresh rates, and increased bandwidth demands, the Keysight Electrical Performance, Validation, and Compliance Test Solution for HDMI offers a fully automated and scalable platform for professionals in design, engineering, and compliance testing to validate device performance with confidence and precision. The new test solution provides a unified platform for automated electrical testing as specified in the HDMI 2.2 test specification, ensuring that device manufacturers can confidently validate product performance at the transmission and cable, while reception testing is introduced at a later stage.

Keysight’s physical layer compliance test solution for HDMI meets the latest technical and procedural demands of the HDMI Forum. Designed for precision and efficiency, the solution integrates high-bandwidth measurement hardware with automated compliance workflows to manage complex test scenarios across transmitters and cables. The modular architecture of the solution supports flexible test configurations, while built-in diagnostics provide deep insight into the root causes of signal degradation. This enables design and validation teams to not only verify compliance but also optimize performance early in the development cycle.

Han Sing Lim, Vice President and General Manager of Keysight’s General Electronic Measurement Division, said: “With the introduction of the Keysight Electrical Performance, Validation, and Compliance Test Solution for HDMI, our customers can accelerate time-to-market for next-gen consumer electronics while ensuring robust integrity and regulatory compliance. By incorporating the latest version of HDMI technology in our solution, we are enabling leading consumer electronics designers and manufacturers to continue to push the boundaries of digital display and multimedia performance.”

Backed by Keysight’s global expertise in compliance testing and proven in high-volume production environments, the new solution delivers a trusted path to certification readiness and superior end-product performance.

The post Keysight Unveils Physical Layer Compliance Test Solution for HDMI to Meet Rising Demands for Ultra-High Resolution and High Dynamic Range appeared first on ELE Times.

Ultra-low-power RL78/L23 MCUs with segment LCD displays & capacitive touch for HMI applications

Renesas Electronics Corporation, a premier supplier of advanced semiconductor solutions, introduced the new 16-bit RL78/L23 microcontroller (MCU) group, expanding its low-power RL78 family. Running at 32MHz, the RL78/L23 MCUs combine industry-leading low-power performance with essential features such as dual-bank flash memory, segment LCD control, and capacitive touch functionality to support smart home appliances, consumer electronics, IoT and metering systems. These compact, cost-effective devices address the performance and power requirements of modern display-based human-machine interface (HMI) applications.

Ultra-Low Power Operation with Optimized LCD Performance

The RL78/L23 is optimized for ultra-low power consumption and ideal for battery-powered applications that spend the majority of time in standby. They offer an active current of just 109μA/MHz and a standby current as low as 0.365μA, along with a fast 1μs wake-up time to help minimize CPU activity. The LCD controller’s new reference mode, VL4, reduces LCD operating current by approximately 30 percent when compared to the existing RL78/L1X group. The MCUs come with SMS (SNOOZE Mode Sequencer), which enables dynamic LCD segment display without CPU intervention. By offloading tasks to the SMS, the devices minimize CPU wake-ups and contribute to system-level power savings. These innovations significantly extend battery life, simplify design and reduce replacement costs, while minimizing environmental impact.

The RL78/L23 offers a wide operating voltage range of 1.6V to 5.5V, which supports direct operation from 5V power supplies commonly used in home appliances and industrial systems. This capability reduces the need for external voltage regulators. The MCUs also integrate key components such as capacitive touch sensing, a temperature sensor, and internal oscillator, reducing BOM cost and PCB size.

Feature-Rich Peripherals for HMI Systems

Designed to meet the dynamic requirements of the HMI market, RL78/L23 integrates a suite of advanced features in a compact, cost-effective package. Its built-in segment LCD controller and capacitive touch realize sleek, responsive user interfaces for products such as induction cooktops and HVAC systems. The IH timer (Timer KB40) enables precise multi-channel heat control, which is essential in smart kitchen appliances such as rice cookers and IH cooktops. The devices include dual-bank flash memory for seamless firmware updates via FOTA (Firmware Over-the-Air), allowing continuous system operation in applications like metering, where downtime must be minimized. The dual-bank architecture allows one memory bank to run the user program, while the other receives updates. This approach keeps the system functional throughout the process for improved reliability.

“The Renesas RL78 Family of 16-bit microcontrollers has been one of the most successful products since its launch more than 10 years ago, particularly in home appliances,” said Daryl Khoo, Vice President of Embedded Processing at Renesas. “I’m pleased to announce the RL78/L23, a new generation of RL78 microcontrollers with rich features, ideally suited for smart home appliances and cost-sensitive IoT solutions. With these devices, we aim to provide a better user experience with our intuitive development environment so that customers can get to production faster with confidence, based on market-proven Renesas technologies.”

Key Features of the RL78/L23

16-bit RL78 microcontroller running at 32MHzBuilt-in segment LCD controller and capacitive touchUp to 512KB of dual-bank flash memory for seamless FOTAUp to 32KB of SRAM and 8KB of data flashSMS for ultra-low power operationIH Timer (KB40) supporting up to 3-channel induction heating controlWide operating voltage range from 1.6V to 5.5VOperating temperature range of -40°C to +105°CMultiple serial interfaces including UART, I2C, CSIIEC60730-compliant self-test library44-100-pin LFQFP, LQFP and HWQFN packages.

Intuitive Development Environment for Faster Time-to-Market

The RL78/L23 comes with an easy-to-use development environment. Developers can leverage robust support tools such as Smart Configurator and QE for Capacitive Touch to streamline system design.

The post Renesas Introduces Ultra-Low-Power RL78/L23 MCUs for Next-Generation Smart Home Appliances appeared first on ELE Times.

Anand Kumar is the Managing Director of STMicroelectronics (ST), India, and has held this position since June 2025. He heads the Global IP and Library Design team, located across France and India, within ST’s Global Technology R&D organization. Kumar’s team takes IPs from concept to production maturity and brings variants over time, across the full portfolio of ST’s proprietary, differentiated technologies and technology platforms.

Anand Kumar began his career in 1999 as an Analog Designer in ST’s Central R&D organization. With over 25 years of experience as a designer, design team lead, and global lead of IP design, he has deep expertise in designing Analog and Mixed-Signal IPs and leading high-performing IP and Library Design teams delivering a wide range of IPs. These include memories, IOs, standard cells, clock-generation IPs, data converters, sensors, power-management IPs, fuses, and other analog and digital IPs across all ST technology nodes.

Kumar holds more than 20 patents and has presented at multiple top-tier international scientific and technical semiconductor conferences.

Anand Kumar was born in Narwana (Haryana state), India, in 1976, and holds a degree in Electronics and Communication Engineering from NSUT (Netaji Subhas University of Technology, formerly Delhi Institute of Technology), Delhi.

The post STMicroelectronics Appoints MD India appeared first on ELE Times.

Nowadays, individuals own an increasing number of devices—such as fitness trackers and smartphones that continuously generate valuable data. At the same time, organizations like banks, hospitals, and enterprises produce vast amounts of sensitive information. However, due to strict privacy regulations, this data cannot be openly shared for centralized processing. In such scenarios, federated learning offers a transformative solution: it enables machine learning models to be trained directly on-device or within institutional boundaries, without transferring raw data. This approach preserves privacy while unlocking powerful, collaborative AI capabilities. As a result, data from diverse sources both personal and institutional can be securely leveraged to extract insights and drive smarter decisions. Below are 10 compelling real-world applications where federated learning is making a significant impact.

1. Telecommunications

The federated model enables telecommunication firms to study patterns of their clients, enhance network performance, and make accurate tele-service projections for their distributed systems. This fosters efficient network systems while safeguarding customer information. In the same context, mobile operators stand to enhance calling services from user data sourced from spatially dispersed systems.

2. Autonomous Vehicles

Self-driving cars and connected vehicles utilize federated learning to collaborate vehicles enhance navigation, obstacle identification, and safety measures. This eliminates the need to consolidate personal driving information. Drivers of self-driven automobiles and fleet operators utilize federated learning to enhance safety, navigation, and object detection with the aid of local sensor data consisting of cameras, LIDAR, and object detection.

3. Finance

Banks and fintech companies use federated learning for detecting fraud, credit scoring, and modeling credit risk. One example is the training of a multi-bank fraud detection model to recognise suspicious transactions while safeguarding user information.

4. Smart Devices & IoT

Smartphones, as well as other wearable devices, use federated learning to enhance voice recognition, keyboard prediction, and health tracking functions. An instance is the Gboard keyboard from Google, which leverages federated learning to upgrade its autocorrect as well as next-word prediction features grounded on users’ typing patterns.

5. Cybersecurity

Federated learning is employed in factories for process optimization, predictive maintenance, and even defect detection. Federated learning enables multiple organizations to collaboratively train intrusion detection models using local network logs. This approach enhances threat detection accuracy while preserving sensitive data and complying with privacy regulations.

6. Manufacturing

Factories use federated learning for predictive maintenance, defect detection, and process optimization. For instance, multiple production lines can train a model to predict equipment failure using local sensor data, reducing downtime.

7. Energy & Utilities

Energy companies and power grids use advanced techniques to forecast demand and anticipate failures in the system by learning from distributed sensor data across substations and smart meters. Use Case includes a national utility company uses federated learning to predict peak electricity usage across cities, helping balance load distribution without accessing individual household data.

8. Retail & E-commerce

Retailers customize product recommendations cross-sell and up-sell suggestions and basket-level cross-product purchase analytics across different store locations without sharing any stepwise item-level purchase data of shoppers. A classic use case is a global fashion retailer who wants to suggest outfit combinations based on current trends of different geographies. The retailer can now use the federated approach, enabling training of the model across all the stores in the regions while protecting shopper and purchase data.

9. Content Platforms

With less risk to user privacy, platforms can better personalize user feeds and automatically moderate content by learning locally from user interactions. Use Case: A video streaming app enhances its recommendation system by locally training on user watch histories stored on devices, ensuring tailored recommendations while refraining from uploading any viewing data to the cloud.

10. Aviation

Carriers and aircraft manufacturers develop models from flight execution and servicing records over different fleets in an attempt to improve safety and cut downtime, with the added benefit of keeping proprietary data private. A use case is offered by the federated model training from different airlines that enables the prediction of an engine’s wear and tear based on flight conditions, which aids in the scheduling of proactive maintenance without the need to share sensitive operational data.

Conclusion:

Federated learning protects privacy while facilitating cooperative model training across dispersed data sources. It lowers the risks associated with data transfers, conforms with data protection laws, and enables businesses to leverage insights without jeopardizing user privacy.

The post Top 10 Federated Learning Applications and Use Cases appeared first on ELE Times.

Federated learning is transforming AI’s potential in India by allowing models to be trained without infringing on the privacy of decentralized data. Federated learning is of critical importance in healthcare, finance, and consumer technology due to the rising needs of industries for AI that is secure, compliant with regulations, and privacy-preserving. Due to India having a flourishing technology ecosystem as well as a strong pool of AI talents, India is emerging as a leader in this technology. This article will discuss the leading 10 companies in India that focus on federated learning.

1. TCS Research

TCS Research as the innovation wing of Tata Consultancy Services, TCS Research collaborates with federated learning for enterprise AI. Their initiatives cover healthcare, banking, and smart city projects, centering on the safe training of models over distributed data silos.

2. Wipro HOLMES

Wipro’s AI platform, uses federated learning to provide intelligent automation and edge AI. Its application in telecommunications, manufacturing, and IT services aids in the development of AI models without eroding data privacy.

3. Infosys Nia

Infosys Nia An all-in-one AI platform, Infosys Nia also offers federated learning for decentralized data modeling, which is especially beneficial in retail, and finance, where data sensitivity is high and compliance is critical.

4. SigTuple

With its headquarters in Bengaluru, SigTuple is a health tech company which employs federated learning to streamline the analysis of medical images and diagnostics, while still maintaining patient data privacy. Their AI solutions not only save time but also improve the decision-making processes of pathologists and radiologists.

5. Qure.ai

With over a decade of specialization in AI-driven radiology, Qure.ai is a clear leader. They are notable examples of the application of federated learning in radiology, not only for advancing diagnostic precision but also for safeguarding critical medical information.

6. Vaidik AI

Vaidik AI marks a new chapter in the federated learning narrative of India. It launched an extensive selection of AI services, including the fine-tuning of LLMs and multilingual AI. Its multidisciplinary expertise in data annotation and the privacy-first approach to AI model development is well known. It provides healthcare, finance, and education sectors with economical and scalable solutions.

7. ActionLabs AI

ActionLabs AI is located in Bengaluru and works with federated learning, edge AI, and generative model creation. Though healthcare and fintech startups appearing to be ActionLabs’ primary areas of focus, the company’s small size allows it to efficiently cater to a wider range of companies.

8. Accenture India

Accenture adapts federated learning to its Responsible AI framework, assisting clients spanning the energy sector to public services in securely training models on decentralized data.

9. Fractal Analytics

Fractal Analytics Fractal applies federated learning to generate consumer insights for retail and CPG. Their solutions enable brands to analyze consumer behavior without pooling sensitive data.

10. Intel India

Intel India, with its offices in Bengaluru and Hyderabad, is pivotal in advancing federated learning as it refines secure hardware platforms such as Trusted Execution Environments (TEEs) and furthers AI research through Intel Labs. It also champions privacy-preserving AI in healthcare, smart cities, and edge computing.

Conclusion:

The federated learning ecosystem in India is evolving rapidly with the presence of global technology leaders such as Intel and the innovative local startups such as ActionLabs AI, Vaidik AI, and SigTuple. These firms not only expand the frontiers of privacy-preserving AI but also position the federated learning ecosystem to thrive on data collaboration devoid of security risks. With growing demand across healthcare, finance, and edge computing, federated learning is becoming a cornerstone of ethical AI development in India.

The post Top 10 Federated Learning Companies in India appeared first on ELE Times.

New Cadence Palladium Dynamic Power Analysis App enables designers of AI/ML chips and systems to create more energy-efficient designs and accelerate time to market

Cadence announced a significant leap forward in the power analysis of pre-silicon designs through its close collaboration with NVIDIA. Leveraging the advanced capabilities of the Cadence Palladium Z3 Enterprise Emulation Platform, utilizing the new Cadence Dynamic Power Analysis (DPA) App, Cadence and NVIDIA have achieved what was previously considered impossible: hardware accelerated dynamic power analysis of billion-gate AI designs, spanning billions of cycles within a few hours with up to 97 percent accuracy. This milestone enables semiconductor and systems developers targeting AI, machine learning (ML) and GPU-accelerated applications to design more energy-efficient systems and accelerate their time to market.

The massive complexity and computational requirements of today’s most advanced semiconductors and systems present a challenge for designers, who have until now been unable to accurately predict their power consumption under realistic conditions. Conventional power analysis tools cannot scale beyond a few hundred thousand cycles without requiring impractical timelines. In close collaboration with NVIDIA, Cadence has overcome these challenges through hardware-assisted power acceleration and parallel processing innovations, enabling previously unattainable precision across billions of cycles in early-stage designs.

“Cadence and NVIDIA are building on our long history of introducing transformative technologies developed through deep collaboration,” said Dhiraj Goswami, corporate vice president and general manager at Cadence. “This project redefined boundaries, processing billions of cycles in as few as two to three hours. This empowers customers to confidently meet aggressive performance and power targets and accelerate their time to silicon.”

“As the era of agentic AI and next-generation AI infrastructure rapidly evolves, engineers need sophisticated tools to design more energy-efficient solutions,” said Narendra Konda, vice president, Hardware Engineering at NVIDIA. “By combining NVIDIA’s accelerated computing expertise with Cadence’s EDA leadership, we’re advancing hardware-accelerated power profiling to enable more precise efficiency in accelerated computing platforms.”

The Palladium Z3 Platform uses the DPA App to accurately estimate power consumption under real-world workloads, allowing functionality, power usage and performance to be verified before tapeout, when the design can still be optimized. Especially useful in AI, ML and GPU-accelerated applications, early power modeling increases energy efficiency while avoiding delays from over- or under-designed semiconductors. Palladium DPA is integrated into the Cadence analysis and implementation solution to allow designers to address power estimation, reduction and signoff throughout the entire design process, resulting in the most efficient silicon and system designs possible.

The post Cadence Accelerates Development of Billion-Gate AI Designs with Innovative Power Analysis Technology Built on NVIDIA appeared first on ELE Times.

Federated Learning (FL) has been termed a revolutionary manner of machine learning because it provides the capability of collaborative model training across devices in a decentralized manner while preserving data privacy. Instead of transferring data to a centralized server for training, devices train locally, and only their model updates are shared. This way, it finds applicability in sensitive areas like healthcare, finance, and mobile applications. As Federated Learning continues to evolve, an increasingly diverse array of algorithms has emerged each designed to enhance communication efficiency, boost model accuracy, and strengthen resilience against data heterogeneity and adversarial challenges. This article will delve into the types, examples, and top 10 Federated Learning Algorithms.

Types of federated learning algorithms:

Federated Learning algorithms get classified by how data is laid out, by the system structure, and by the privacy requirements. Horizontal FL covers clients with the same features but distinct data points. Vertical FL captures the case where features are different but clients overlap. When users and features are both different, we use Federated Transfer Learning. Decentralized FL, as opposed to Centralized FL, doesn’t use a central server and instead allows for peer-to-peer communication. In terms of FL deployment methods, Cross-Silo FL consists of powerful participants like hospitals and banks, while Cross-Device FL focuses on lightweight devices, such as smartphones. In addition, Privacy-Preserving FL protects user data with encryption, differential privacy, and other techniques, and Robust FL attempts to protect the system from malicious, adversarial, or broken clients.

Examples of federated learning algorithms:

Examples of Federated Learning Algorithms: A number of algorithms have been created to overcome challenges specific to Federated Learning problems. The basic approach of Federated Learning is FedAvg, which, in contrast, models client averaging. FedProx, which is designed to work well with data heterogeneity, is a more advanced approach. For personalization, FedPer customizes top layers for each client, and pFedMe applies meta-learning techniques. Communication-efficient algorithms like SCAFFOLD and FedPAQ reduce bandwidth usage and client drift. Robust algorithms such as Krum, Bulyan, and RFA filter out malicious or noisy updates to maintain model integrity. Privacy-focused methods like DP-FedAvg and Secure Aggregation ensure data confidentiality during training. These algorithms are often tailored or combined to suit specific domains like healthcare, finance, and IoT.

Top 10 Federated Learning Algorithms:

1. Federated Averaging (FedAvg):

FedAvg stands as the founding algorithm for Federated Learning. The weight averaging is performed after models are trained locally on each client for updating the global model. Due to its simple design and the ease with which one can scale, it has been widely implemented in practice.

2. FedProx

FedProx improves upon FedAvg by adding a proximal term to the loss function. FedProx builds upon FedAvg by introducing a proximal term in the loss function. By penalizing local updates that diverge too much from the global model, this term helps stabilize training in settings with widely differing client data distributions. It is especially helpful in fields like healthcare and finance, where heterogeneous data is prevalent.

3. FedNova (Federated Normalized Averaging)

To address the drift of the client, FedNova normalizes updates with respect to the number of local steps and learning rates. This ensures each client has an equal contribution to the global model regardless of its computational capabilities or data volume. This further favors convergence and fairness in heterogeneous setups.

4. SCAFFOLD

SCAFFOLD, an abbreviation for Stochastic Controlled Averaging for Federated Learning, employs control variates to make corrections to the client’s updates. This limits the variance that exists owing to non-IID data and speeds the convergence. It is particularly effective in an edge computing environment, where data come from various sources.

5. MOON (Model-Contrastive Federated Learning)

MOON brings contrastive learning into FL by aligning local and global model representations. It enforces consistency of models that are particularly necessary when client data are highly divergent. MOON should often be used for image and text classification tasks for very heterogeneous user bases.

6. FedDyn (Federated Dynamic Regularization)

FedDyn incorporates a dynamic regularization term in the loss function to enable the global model to accommodate client-specific updates better. Because of this, it can withstand situations involving extremely diverse data, such user-specific recommendation systems or personalized healthcare.

7. FedOpt

FedOpt substitutes in place of the vanilla averaging mechanisms with advanced server-side optimizers like Adam, Yogi, and Adagrad. Using these optimizers leads to faster and more stable convergence, which is paramount in deep learning tasks with large neural networks.

8. Per-FedAvg (Personalized Federated Averaging)

Personalized Federated Averaging hopes to balance global generalization with local adaption by allowing clients to fine-tune the global model locally. Because of this, Per-FedAvg is suitable for personalized recommendations, mobile apps, and wearable health monitors.

9. FedMA (Federated Matched Averaging)

The distinguishing feature of this method is the matching of neurons across client models before averaging. This retains the architecture of a deep neural network and hence allows for much more meaningful aggregation, especially for convolutional and recurrent architectures.

10. FedSGD (Federated Stochastic Gradient Descent)

A simpler alternative to FedAvg, FedSGD sends gradients instead of model weights. It’s more communication-intensive but can be useful when frequent updates are needed or when model sizes are small.

Conclusion:

These algorithms represent the cutting edge of federated learning, each tailored to address specific challenges like data heterogeneity, personalization, and communication efficiency. As FL continues to grow in importance especially in privacy-sensitive domains these innovations will be crucial in building robust, scalable, and ethical AI systems.

The post Top 10 Federated Learning Algorithms appeared first on ELE Times.

• Bharat set to welcome delegates from 33 Countries, 50+ CXOs,  350 Exhibitors
• At country’s biggest Semiconductors & Electronics Show in New Delhi from 2-4 September 2025
• Over 50+  Eminent Global Visionary Speakers
• Event To Highlight Robust Local Semiconductor Ecosystem Expansion and Industry Trends

The fourth edition of SEMICON India 2025 will be officially inaugurated by Hon’ble Prime Minister Shri. Narendra Modi on 2nd September 2025 at Yashobhoomi (India International Convention and Expo Centre), New Delhi. Staying true to its legacy of positioning India as a global Semiconductor powerhouse, the fourth edition of SEMICON India 2025 will convene key stakeholders including global leaders, semiconductor industry experts, academia, government officials and students.

Under the Semicon India program, 10 strategic projects have been approved across high-volume fabs, 3D heterogeneous packaging, compound semiconductors (including SiC), and OSATs, marking a significant milestone for the country. Recognizing semiconductors as a foundational technology, over 280 academic institutes and 72 startups have been equipped with state-of-the-art design tools, while 23 startups have already been approved under the DLI scheme. These initiatives are driving innovations in critical applications such as CCTV systems, navigation chips, motor controllers, communication devices, and microprocessors—strengthening India’s journey towards Atmanirbhar Bharat.

Accelerating India’s semiconductor revolution, SEMI, the global industry association prompting the semiconductor industry and India Semiconductor Mission (ISM), Ministry of Electronics and Information Technology (MeitY), announced the programming for SEMICON India 2025 at a press conference held in the national capital.

Under the theme Building the Next Semiconductor Powerhouse, the event will offer valuable insights into innovations and trends in key areas such as Fabs, Advanced packaging, smart manufacturing, AI, supply chain management, sustainability, workforce development, Designs and Start Up’s along with 6 country round tables.

The SEMICON India exhibition will feature nearly 350 exhibitors from across the global semiconductor value chain including 6 county Round Tables, 4 country pavilions,  9 states participations and over 15000 expected visitors providing South Asia’s single largest platform for showcasing the latest advancements in the semiconductor and electronics industries, said Shri S Krishnan, Secretary, MeitY.

“SEMI is bringing the combined expertise and capabilities of our member companies across the global electronics design and manufacturing supply chain to SEMICON India, helping to advance both India’s semiconductor ecosystem expansion and industry supply chain resiliency,” said Ajit Manocha, President and CEO, SEMI. “The event will feature signature SEMICON opportunities for professional networking, business development, and insights into technology and market trends from a star-studded lineup of leading industry experts.”

SEMICON India 2025 is designed to maximize technological advancements in the semiconductor and electronics domain and highlight India’s policies aimed at strengthening its semiconductor ecosystem.

The event is a remarkable convergence of ideas, collaboration and innovation, and provides a unique opportunity to address complex challenges of tomorrow while fostering collaboration across the semiconductor ecosystem. We are looking forward to an astounding number of participations this year, Said Shri Amitesh Kumar Sinha, Additional Secretary , MeitY and CEO ISM.

“India’s semiconductor industry is poised for a breakthrough, with domestic policies and private sector capacity finally aligning to propel the nation to global prominence. As we navigate this transformative landscape, collaboration and ecosystem building will be key to unlocking the next wave of growth and breakthroughs and SEMICON India 2025 plays the catalyst for this.” said Ashok Chandak, President, SEMI India and IESA.

In addition to distinguished government officials, this year’s event will also feature an impressive lineup of industry leaders from top companies including Applied Materials, ASML, IBM, Infineon, KLA, Lam Research, MERCK, Micron, PSMC, Rapidus, Sandisk, Siemens, SK Hynix, TATA Electronics, Tokyo Electron, and many more.

Over the span of three days, the flagship event will feature a diverse range of activities including high profile keynotes, panel discussions, fireside chats, paper presentations, 6 international roundtables and more that will converge to drive the next wave of semiconductor innovation and growth. The event will also include a ‘Workforce Development Pavilion’ to showcase microelectronics career prospects and attract new talent.

SEMICON India is one of eight annual SEMICON expositions worldwide hosted by SEMI that bring together executives and leading experts in the global semiconductor design and manufacturing ecosystem. The upcoming event marks the beginning of an exciting journey into the future of technological innovation, fostering collaboration and sustainability in the global semiconductor ecosystem.

The post Hon’ble PM Shri. Narendra Modi to inaugurate fourth edition of SEMICON India 2025 appeared first on ELE Times.

The new BBA series features higher field strengths for critical test environments up to 18 GHz

Rohde & Schwarz, a leading global supplier of test and measurement equipment and a reliable partner for turnkey EMC solutions, has expanded its broadband amplifier portfolio of the R&SBBA300 family with the two innovative amplifier series R&SBBA300-F for 6 to 13 GHz and R&SBBA300-FG for 6 to 18 GHz with additional power classes such as 90W, 180W and 300W.

Together with the already successfully introduced broadband amplifier series R&SBBA300-CDE for 380 MHz to 6 GHz and R&SBBA300-DE for 1 to 6 GHz, Rohde & Schwarz now offers compact dual-band amplifiers covering the entire frequency range from 380 MHz to 18 GHz in 4HU desktop models only.

The R&SBBA300 family is the new generation of compact, solid-state broadband amplifiers, designed for high availability and a linear output across an ultra-wide frequency range. It supports amplitude, frequency, phase, pulse and complex OFDM modulation modes and is extremely robust under all mismatch conditions, providing reliable test results in all circumstances.

Typical applications include EMC, co-existence and RF component tests during development, compliance test and production. The very wide frequency range makes them ideal for wireless and ultra-wideband testing.

The R&SBBA300-F series is a cost-effective solution for applications between 6 GHz and 13 GHz; the R&SBBA300-FG series covers a continuous frequency band from 6 GHz to 18 GHz. The two amplifier series can be used for ultrawideband applications as well as to address various EMC standards within mobile communications (FCC, ETSI), automotive (ISO), aerospace (DO-160), and military (MIL-STD-461). Both the R&SBBA300-F and the R&SBBA300-FG are now available in the power classes 30 W, 50 W, 90 W, 180 W, 300 W.

The R&SBBA300 broadband amplifier family offers two powerful tools for tailoring the RF output signal to the application: adjusting the amplifier either for excellent linearity or faithful reproduction of pulse signals by shifting the operating point between class A and class AB, and setting the amplifier for maximum tolerance to output mismatch or for maximum RF output power to utilize the power reserves for the application.

This allows users like developers, test engineers, integrators, or operators to optimize the output signal and react flexibly to a wide variety of requirements. Both parameters can be changed during amplifier operation.

“In addition to high linearity and excellent harmonic properties, our users also need extremely wide, continuous frequency bands at high RF output power,” said Michael Hempel, product manager for amplifier systems at Rohde & Schwarz. “The BBA300 series is our direct response to these requirements, offering outstanding bandwidth with high output power.”

Rohde & Schwarz also provides fully compliant EMI test receivers, signal generators, antennas, software and other essential system components and service for EMC testing.

The post Rohde & Schwarz extends the broadband amplifier range to 18 GHz appeared first on ELE Times.

Empowering a New Era of Physical AI and Robotics Development in the Asia-Pacific Region

EDOM Technology announced the official distribution of NVIDIA’s latest NVIDIA Jetson Thor module and developer kit, built for physical AI and general robotics. This move is set to accelerate technological upgrades and local deployment of applications such as intelligent robotics, AMR (Autonomous Mobile Robot), AIoT, and smart manufacturing across the region.

Jetson Thor is the most powerful edge AI module in the NVIDIA Jetson series. Built on NVIDIA Blackwell GPU architecture, it delivers over 2,070 TFLOPS of AI inference capability, specifically designed for humanoid robots, AMRs, and industrial smart devices. Its highly integrated computing architecture supports multi-sensor fusion, Transformer model inference, and real-time motion control, enabling a deep integration of generative AI and the physical world. Jetson Thor seamlessly integrates with NVIDIA Isaac ROS, NVIDIA Omniverse, and NVIDIA Isaac GR00T, forming a complete AI toolchain from data generation and simulation training to edge deployment. This significantly accelerates the adoption and commercialization of Physical AI applications, making it a key enabler of next-generation edge AI and robotics intelligence.

As NVIDIA’s long-standing partner and authorized distributor of Jetson series modules in the Asia-Pacific, EDOM brings around 30 years of experience in distribution and technical integration, covering AI modules, embedded systems, sensor integration, industrial automation, and component applications.

EDOM provides comprehensive product offerings of the Jetson Thor platform, including the Jetson AGX Thor Developer kit and Jetson T5000 module. Equipped with NVIDIA Holoscan Sensor Bridge for real-time data processing, along with high-speed interfaces such as GMSL, MIPI, 25GbE, 5G, and Wi-Fi modules, as well as high-performance storage interfaces, these solutions effectively meet the stringent low-latency and high-bandwidth demands of edge computing. Additionally, EDOM supports custom hardware design and system integration reference solutions, fully assisting customers in accelerating product development and deployment processes.

Jeffrey Yu, CEO at EDOM Technology stated:
“Jetson Thor represents a major breakthrough in NVIDIA’s physical AI and robotics applications. We are honored to be the authorized distributor for Jetson Thor in the Asia-Pacific. By combining technical supports, educational resources, and platform ecosystems, we aim to help customers accelerate innovation and advance the deployment of generative and physical AI technologies.”

With the launch of Jetson Thor, the module is expected to see wide adoption in fast-growing physical AI and robotics sectors across Asia-Pacific, including smart manufacturing, AMRs, smart transportation, and service robots. For example:

• In high-precision AOI (Automated Optical Inspection), Jetson Thor can process large-scale image data in real time and perform inference, improving yield rates and automation in factories.
• In AMR factory logistics, through multi-sensor fusion and real-time motion control, it enables autonomous navigation and smart scheduling in complex environments.
• In humanoid and companion robots, Jetson Thor’s integration with GR00T multimodal models and visual recognition enables highly interactive scenarios, ideal for applications in aging societies and public services.
• With support for multiple GMSL cameras and high-speed Ethernet, Jetson Thor is also well-suited for smart city traffic nodes, performing real-time image analysis and behavior recognition.

These applications demonstrate Jetson Thor’s powerful computing capabilities and provide developers and enterprises in Asia-Pacific a complete path from AI training to edge deployment.

EDOM will continue to act as a critical bridge between technology and the market, working with system developers, integrators, and academic institutions. By driving the local deployment of the NVIDIA Jetson platform across key sectors—such as smart transportation, AIoT, and smart manufacturing—EDOM is accelerating the development and implementation of generative AI and Physical AI throughout the Asia-Pacific region.

The post EDOM Strengthens NVIDIA Jetson Thor Distribution Across APAC appeared first on ELE Times.