What question did this study set out to answer?

To explore the use of ULTRARAM memory devices for on-chip training and inference in neural networks.

March 25, 2026Journal of Applied PhysicsOpen Access

Artificial synapse based on ULTRARAM memory device for neuromorphic applications

View Full Paper

Ask AI

Bookmark

To explore the use of ULTRARAM memory devices for on-chip training and inference in neural networks.
Utilized ULTRARAM-based synaptic arrays for memory storage and weight updating
Developed a physics-based compact model for ULTRARAM to simulate charge trapping
Applied a circuit-level macro-model to benchmark learning performance on area, latency, and energy
Achieved 91% training accuracy during on-chip simulations
Demonstrated 1.8× improvement in area and 1.52× improvement in energy compared to CMOS designs
Exhibited high endurance of ULTRARAM with over 10 million programming/erase cycles and over 1000 years of retention

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Monolithic arrays using ULTRARAM enhance training accuracy and energy efficiency in neural networks, implying significant advancements in memory technology.

Cite This Study

Kumar et al. (2026) studied this question.

synapsesocial.com/papers/69c37bf3b34aaaeb1a67ed1c https://doi.org/10.1063/5.0314826

View Full Paper

Ask AI

Bookmark

Implication

Monolithic arrays using ULTRARAM enhance training accuracy and energy efficiency in neural networks, implying significant advancements in memory technology.

Artificial synapse based on ULTRARAM memory device for neuromorphic applications

Key Points

Discussion

Implication

Cite This Study

Artificial synapse based on ULTRARAM memory device for neuromorphic applications

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider