Related papers: EXTENT: Enabling Approximation-Oriented Energy Eff…
The quadratic complexity of self-attention mechanism presents a significant impediment to applying Transformer models to long sequences. This work explores computational principles derived from astrocytes-glial cells critical for biological…
Perpendicular shape anisotropy (PSA) and double magnetic tunnel junctions (DMTJ) offer practical solutions to downscale spin-transfer-torque Magnetic Random-Access Memory (STT-MRAM) beyond 20 nm technology nodes, whilst retaining their…
The scaling of magnetic memory into nanometer size calls for a theoretical model to accurately predict the switching current. Previous models show large discrepancy with experiments in studying the spin-orbit torque switching of…
Conventional computing paradigm struggles to fulfill the rapidly growing demands from emerging applications, especially those for machine intelligence, because much of the power and energy is consumed by constant data transfers between…
This work introduces novel training and post-training compression schemes to reduce external memory access during transformer model inference. Additionally, a new control flow mechanism, called dynamic batching, and a novel buffer…
Power consumption is the main limitation in the development of new high performance random access memory for portable electronic devices. Magnetic RAM (MRAM) with CoFeB/MgO based magnetic tunnel junctions (MTJs) is a promising candidate for…
We present a phenomenological theory of filamentary resistive random access memory (RRAM) describing the commonly observed features of their current-voltage characteristics. Our approach follows the approach of thermodynamic theory…
Fine-grained sparsity promises higher parametric capacity without proportional per-token compute, but often suffers from training instability, load balancing, and communication overhead. We introduce STEM (Scaling Transformers with…
Long Short-Term Memory (LSTM) is one of the most widely used recurrent structures in sequence modeling. It aims to use gates to control information flow (e.g., whether to skip some information or not) in the recurrent computations, although…
As deep neural network (DNN) models are growing exponentially in size, their deployment on resource-constrained edge platforms is becoming increasingly challenging. In-memory-computing (IMC) with non-volatile memories (NVMs) has emerged as…
Ultra-fast \& low-power superconductor single-flux-quantum (SFQ)-based CNN systolic accelerators are built to enhance the CNN inference throughput. However, shift-register (SHIFT)-based scratchpad memory (SPM) arrays prevent a SFQ CNN…
Electric-field control of spin states offers a promising route to ultra-low-power, ultra-fast magnetization switching in spintronic devices such as magnetic tunnel junctions (MTJs). Recent progress in modulating spin-orbit interactions at…
With the staggering increase of edge compute applications like Internet-of-Things (IoT) and artificial intelligence (AI), the demand for fast, energy-efficient on-chip memory is growing. While the fast and mature static random-access memory…
Arrays of Vortex Transitional (VT) memory cells with functional density up to $1 Mbit/cm^2$ have been designed, fabricated, and successfully demonstrated. This progress is due to recent advances in design optimization and in superconductor…
Stochastic magnetic tunnel junctions (s-MTJ) is a promising component of probabilistic bit (p-bit), which plays a pivotal role in probabilistic computers. For a standard cell structure of the p-bit, s-MTJ is desired to be insensitive to…
As conventional technology scaling approaches physical and power limitations, modern computing systems increasingly face performance bottlenecks arising from memory latency, energy consumption, scalability constraints, and data movement…
Superparamagnetic tunnel junctions (sMTJs) are emerging as promising components for stochastic units in neuromorphic computing, owing to their tunable random switching behavior. Conventional MTJ control methods, such as spin-transfer torque…
As the demand for efficient, low-power computing in embedded and edge devices grows, traditional computing methods are becoming less effective for handling complex tasks. Stochastic computing (SC) offers a promising alternative by…
Magneto-Electric FET (MEFET) is a recently developed post-CMOS FET, which offers intriguing characteristics for high speed and low-power design in both logic and memory applications. In this paper, for the first time, we propose a…
Rotating the magnetization of a magnetostrictive nanomagnet with electrically generated mechanical strain dissipates miniscule amount of energy compared to any other rotation method and would have been the ideal method to write bits in…