Bernard Widrow, pioneer of adaptive neural networks and longtime Stanford engineering professor, dies at 95

Bernard “Bernie” Widrow, professor emeritus of electrical engineering in Stanford’s School of Engineering and a foundational figure in adaptive signal processing and neural networks, died on September 30, 2025. He was 95.

Widrow’s groundbreaking contributions helped shape technologies that enabled modern computing, communication, and artificial intelligence. “Professor Widrow and his students played an important role in the early days of what we now call AI,” said Marcian “Ted” Hoff (PhD ’62), early Intel employee and architect of the first microprocessor. “The concept of computers that learn from experience was an important aspect of his work. He was a pioneer.”

Over a six-decade career at Stanford, he was a trailblazing researcher, wide-ranging inventor, and beloved mentor whose impact reached across generations. “Everybody in signal processing and neural networks in the last decades has benefited in some form from his trailblazing work,” said Bernd Girod, professor emeritus of electrical engineering. 

“He was an incredibly kind and approachable mentor,” said Gregory Kovacs (PhD ’90, MD ’92), professor emeritus of electrical engineering, a former student, and longtime co-teacher with Widrow. “He was genius-class brilliant, outgoing, charming, as well as humble and unbelievably kind to students.”

Widrow was known for his warmth, sharp wit, and personable, friendly presence. Colleagues remembered Widrow’s clarity of thought, curiosity, and charming eccentricities that made him both a captivating storyteller and a deeply human mentor. His contagious laugh, attentive conversations, and genuine care for others brought warmth and joy to every interaction, his daughter Leslie Derbin said.

A pioneering career

Born on December 24, 1929, in Norwich, Connecticut, Widrow grew up in an academically focused family that valued curiosity and hard work. He became obsessed with electronics, spending years building batteries and repairing broken radios. His early interest led him to the Massachusetts Institute of Technology, where he earned his bachelor’s degree in electrical engineering in 1951, followed by a master’s degree in 1953, and a doctoral degree in 1956, working with Bill Linvill.

He served on the MIT faculty before joining Stanford in 1959, at the invitation of his mentor's twin brother, John Linvill, who was then chair of electrical engineering and a key figure in shaping the department’s modern identity. He arrived during a period of bold expansion led by School of Engineering Dean and University Provost Fred Terman, widely regarded as the “father of Silicon Valley,” whose vision transformed Stanford Engineering into a hub for innovation.

Working with Hoff, his first doctoral student, Widrow developed the Least Mean Squares algorithm, a method for machines to learn by measuring their errors and correcting them step by step. This simple idea allowed devices to adapt to changing conditions and improve their performance, becoming one of the most widely used learning tools in engineering. It ultimately helped power high-speed communication systems, modems, mobile phones, and the early internet.

The two created some of the earliest neural networks, called the adaptive linear neuron (ADALINE) and many ADALINEs (MADALINE), that could improve their performance by noticing errors and making minor adjustments. These systems helped introduce the idea that a machine could recognize patterns and make better decisions over time, an approach that influenced modern AI.

Their work was so fundamental that examples of these early networks have been displayed in major museums. Two ADALINEs are part of the Smithsonian’s National Museum of American History collection, and Widrow’s work was also featured in a Computer History Museum exhibit tracing the history of AI.

Widrow’s research extended far beyond his early work on neural networks. He went on to create methods that helped machines and devices adjust themselves in real time – a process now known as adaptive filtering. This allowed devices to tune out unwanted noise and focus on the signals that matter, which in turn led to clearer heart and brain monitoring, more stable aircraft sensors, and more dependable digital communication systems.

Martin Hellman (MA ’67, PhD ’69), professor emeritus of electrical engineering at Stanford, recalled encountering Widrow’s work as a graduate student in the 1960s, noting his amazement at a film demonstration of active noise canceling for submarine sonar.

Widrow’s achievements received widespread recognition. He received the IEEE Alexander Graham Bell Medal, the IEEE Neural Networks Pioneer Medal, and the Benjamin Franklin Medal for Engineering from the Franklin Institute. He was elected to the National Academy of Engineering and inducted into the Silicon Valley Engineering Hall of Fame, and he served as president and a longtime board member of the International Neural Network Society.

Over his career, he authored more than 125 technical papers, wrote or co-authored numerous books, and held 21 patents. Widrow mentored 89 doctoral students, many of whom became leaders in academia, industry, and government. Known for his patience, clarity, and humor in the workplace, he created a laboratory environment affectionately called “the zoo,” which pulsed with energy, experimentation, and community.

An innovative teacher

Widrow was equally influential in the classroom. He designed and taught EE 280 with Kovacs, an innovative course exploring the intersection of engineering and medicine – even though Widrow had never taken a biology class. Class projects addressed real clinical issues, including fetal monitoring, hearing loss, and cardiovascular devices.

Students in his EE 280 course gave their final presentations at the Widrow family’s home on campus, a tradition that included a barbecue, swim party, volleyball, and lively conversation that reflected the collaborative spirit he created with his students.

Widrow retired at age 80 but continued advising doctoral students and writing. Into his 90s, he stayed mentally sharp, engaged, and eager to explore new ideas. Late in life, he co-authored his final book, Cognitive Memory: Human Memory | Machine Memory, which explores how the brain – and computers – make memories.

He is survived by his daughters, Leslie Derbin, (AB ’82), and Debbie Sklarin (AB ’85, MA ’87), and her husband, Rick (AB ’85); his grandsons, Jeffrey and Adam; and his sister, Elinor Semel. He was predeceased by his wife, Ronna, and son-in-law, Robert.