A5061265455- Parallel Computing and Optimization Techniques
- Embedded Systems Design Techniques
- Advanced Neural Network Applications
- Memory and Neural Mechanisms
- Ferroelectric and Negative Capacitance Devices
- Advanced Data Storage Technologies
- Neurogenesis and neuroplasticity mechanisms
- Interconnection Networks and Systems
- Child and Animal Learning Development
- Opioid Use Disorder Treatment
- Anesthesia and Neurotoxicity Research
- Circadian rhythm and melatonin
- Stress Responses and Cortisol
- Substance Abuse Treatment and Outcomes
- Poisoning and overdose treatments
- Neuroscience and Neuropharmacology Research
- Neuroscience and Neural Engineering
- CCD and CMOS Imaging Sensors
- VLSI and Analog Circuit Testing
- Advanced Memory and Neural Computing
- Neural dynamics and brain function
Vassar College
2018-2025
Yale University
2022-2023
Neurosciences Institute
2022
IBM Research - Austin
2020-2021
Poughkeepsie Public Library District
2015-2020
IBM (United States)
2013
Low-precision computation is the key enabling factor to achieve high compute densities (T0PS/W and T0PS/mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) in AI hardware accelerators across cloud edge platforms. However, robust deep learning (DL) model accuracy equivalent high-precision must be maintained. Improvements bandwidth, architecture, power management are also required harness benefit of reduced precision by feeding supporting...
A processor core is presented for AI training and inference products. Leading-edge compute efficiency achieved robust fp16 via efficient heterogeneous 2-D systolic array-SIMD engines leveraging compact DLFloat16 FPUs. Architectural flexibility maintained very high utilization across neural network topologies. modular dual-corelet architecture with a shared scratchpad software-controlled network/memory interface enables scalability to many-core SoCs large-scale systems. The 14nm achieves peak...
Reduced precision computation is a key enabling factor for energy-efficient acceleration of deep learning (DL) applications. This article presents 7-nm four-core mixed-precision artificial intelligence (AI) chip that supports four compute precisions—FP16, Hybrid-FP8 (HFP8), INT4, and INT2—to support diverse application demands training inference. The leverages cutting-edge algorithmic advances to demonstrate leading-edge power efficiency 8-bit floating-point (FP8) INT4 inference without...
The IBM z13™ system is the latest generation of z Systems™ mainframes. z13 microprocessor improves upon zEnterprise® EC12 (zEC12) processor with two vector execution units, higher instruction parallelism, and a simultaneous multithreaded (SMT) architecture that supports concurrent threads. These advances yield performance gains in legacy online transaction processing business analytics workloads. This features an eight-core chip, robust cache hierarchy, large multiprocessor design optimized...
This paper describes the large capacity hierarchical branch predictor in 5.5 GHz IBM zEnterprise EC12 microprocessor. Performance analyses a simulation model and on zEC12 hardware demonstrate benefit of this hierarchy compared to smaller one level predictor. Novel structures algorithms for two prediction are presented. Prediction information about multiple branches is bulk transferred from second into first upon detecting perceived miss level. The does not directly make predictions. Access...
The corticospinal tract (CST) forms a central part of the voluntary motor apparatus in all mammals. Thus, injury, disease, and subsequent degeneration within this pathway result chronic irreversible functional deficits. Current strategies to repair damaged CST are suboptimal because underexplored molecular heterogeneity adult tract. Here, we combine spinal retrograde tracing with single-cell RNA sequencing (scRNAseq) male female mice index neuron (CSN) subtypes that differentially innervate...
Abstract Spatiotemporal patterns of activity in the neocortex are linked to cognitive processes underlying behavior. However, identifying discrete events within highly dynamic cortical network fluctuations remains a critical challenge. Here, we demonstrate novel analytical method track state-dependent β- (15-30Hz) and γ- (30-80Hz) range mouse primary visual cortex (V1). We find that γ selectively associated with enhanced encoding by V1 neurons event rate increases prior visually-cued...
The design of the modern, enterprise-class IBM z15 branch predictor is described. Implemented as a multilevel look-ahead structure, capable predicting direction and target addresses, augmented with multiple auxiliary direction, target, power predictors. Predictions are made asynchronously, later integrated into processor pipeline. optimized for unique workloads executed on these systems, including compute intensive both large instruction data footprint workloads. This paper highlights major...
We report a replication of Patel, Gibson, Ratner, Besson &amp; Holbomb (1998). The results our are largely consistent with the conclusions original study. found evidence P600 component event-related potential (ERP) in response to syntactic violations language and harmonic inconsistencies music. There were some minor differences spatial distribution on scalp between original. experiment was pre-registered at https://osf.io/g3b5j/register/565fb3678c5e4a66b5582f67. conducted this as part an...