fp32
single precision, float
FP32 (Single-Precision Floating Point) is the 32-bit numerical format that serves as the baseline precision for neural network training — using 1 sign bit, 8 exponent bits, and 23…
1,257 technical terms and definitions
single precision, float
FP32 (Single-Precision Floating Point) is the 32-bit numerical format that serves as the baseline precision for neural network training — using 1 sign bit, 8 exponent bits, and 23…
mxfp, microscaling, mxfp4, nvfp4, mxfp8, mxfp6, mx format, mx formats, block floating point, fp6, 4-bit precision, 4 bit inference, microscaling format, e2m1
FP4 and the microscaling (MX) formats are how modern accelerators push numeric precision below 8 bits without the model falling apart. A four-bit float has only about sixteen…
8bit, latest
FP8 is an 8-bit floating-point number format used to store and multiply the weights, activations, and sometimes gradients of large neural networks at half the width of FP16. It…
half precision training, bfloat16 fp16 comparison, loss scaling amp, automatic mixed precision fp8
Mixed Precision Training is a deep learning training technique that uses lower-precision floating-point formats (FP16, BF16, or FP8) for computation while maintaining FP32 master…
field programmable gate array, fpga architecture, programmable logic
FPGA means field-programmable gate array: an integrated circuit whose logic functions and connections can be configured after manufacturing. Instead of committing one fixed…
chip design hobby, logisim, ngspice
FPGA alternatives for chip design hobbyists provide accessible paths to learn and practice digital circuit design without semiconductor fabrication — from programmable hardware…
chip design hobby, logisim, ngspice
FPGA alternatives for chip design hobbyists provide accessible paths to learn and practice digital circuit design without semiconductor fabrication — from programmable hardware…
fpga lut logic element, fpga routing fabric, fpga hard ip dsp, xilinx versal acap
FPGA Architecture: Configurable Logic Blocks with Programmable Interconnect — reconfigurable silicon fabric enabling rapid prototyping and domain-specific acceleration without…
hardware
FPGA for AI refers to Field-Programmable Gate Arrays configured as custom neural network accelerators — offering a unique position between general-purpose GPUs and fixed-function…
vivado hls vitis, catapult hls, c to rtl synthesis, pragma hls optimization
FPGA High-Level Synthesis (HLS) is a automated design methodology that converts high-level C/C++ specifications directly into synthesizable RTL, dramatically reducing design time…
fpga-based ml acceleration, fpga-based ml, ml acceleration
FPGA-Based ML Acceleration Introduction & Motivation FPGA-Based ML Acceleration addresses a central problem in ml-hardware process control and reconfiguration-latency prediction…
fpga hls, fpga pipeline, fpga streaming, fpga dataflow, fpga accelerator
FPGA Parallel Computing and HLS is the use of Field-Programmable Gate Arrays as custom hardware accelerators for high-throughput, low-latency parallel computation — leveraging…
fpga emulation, hardware emulation
FPGA prototyping is the practice of implementing a pre-silicon digital design on field-programmable gate arrays to validate functionality, firmware/software integration…
fpga rapid prototyping, fpga rapid, rapid prototyping
FPGA Rapid Prototyping Introduction & Motivation FPGA Rapid Prototyping addresses a central problem in hardware-design process control and prototyping-time-weeks prediction: how…
metrology
Fractal Dimension of Surfaces is a mathematical metric quantifying the self-similar complexity of surface roughness — a fractal dimension between 2 (perfectly smooth plane) and 3…
Fractional Differencing for Memory-Preserving Stationarity Introduction & Motivation Fractional Differencing for Memory-Preserving Stationarity addresses a central problem…
doe
A fractional factorial design is a DOE approach that tests only a carefully selected subset of the full factorial combinations, dramatically reducing the number of experimental…
fracture toughness prediction, fracture toughness, toughness prediction
Fracture Toughness Prediction with Machine Learning Introduction & Motivation Fracture Toughness Prediction with Machine Learning addresses a central problem in fracture toughness…
lithography, shot count optimization, naive fracturing, data volume expansion, stitching artifacts, write time, mask writer primitives, trapezoid decomposition, edge placement accuracy
Fractured data is the mask writer's actual input format — the result of decomposing every complex, OPC-corrected layout polygon into simple rectangles and trapezoids a writer can…
fracturing job deck generation time prediction, fracturing job deck, job deck generation, deck generation time, generation time prediction, fracturing job, job deck
Fracturing Job Deck Generation Time Prediction with Machine Learning Introduction & Motivation Fracturing Job Deck Generation Time Prediction with Machine Learning addresses…
audio & speech
FragmentVC is a voice-conversion method that assembles target-style speech from reference acoustic fragments. - It performs zero-shot style transfer by matching source content…
ferroelectric ram (feram) polarization fatigue and imprint, ferroelectric ram (feram), ram (feram) polarization, (feram) polarization fatigue, polarization fatigue imprint, ram (feram)
Ferroelectric RAM (FeRAM) Polarization Fatigue and Imprint Introduction & Motivation Ferroelectric RAM (FeRAM) Polarization Fatigue and Imprint addresses a central problem…
computer vision
Frame interpolation generates in-between frames to increase video frame rate or create smooth slow motion. How it works: Estimate motion (optical flow) between existing frames…
video generation
Frame interpolation is the process of generating intermediate frames between existing frames to increase frame rate or smooth motion - it is used for motion enhancement, slow…
multimodal ai
Frame Interpolation is generating intermediate frames between existing video frames to increase frame rate or smooth motion - It improves visual continuity in playback and motion…
video understanding
Frame order prediction is the video pretext task that shuffles clips or frames and trains the model to recover correct temporal order - this objective teaches temporal…
framework ecosystem dependency, framework ecosystem, ecosystem dependency
Framework Ecosystem Dependency Introduction & Motivation Framework Ecosystem Dependency addresses a central problem in programming process control…
framework upgrade strategy, framework upgrade, upgrade strategy
Framework Upgrade Strategy Introduction & Motivation Framework Upgrade Strategy addresses a central problem in programming process control and upgrade-complexity-ecosystem-changes…
sep, standard essential patent, licensing, patents, royalty, legal, standards
FRAND licensing is licensing of standard essential patents under fair reasonable and non-discriminatory terms - FRAND frameworks balance patent-holder returns with broad…
Frank-Wolfe Conditional Gradient for Convex Sets Introduction & Motivation Frank-Wolfe Conditional Gradient for Convex Sets addresses a central problem in optimization algorithm…
frank-wolfe conditional gradient prediction, frank-wolfe conditional gradient, conditional gradient prediction, frank-wolfe conditional, conditional gradient, gradient prediction
Frank-Wolfe Conditional Gradient Prediction with Machine Learning Introduction & Motivation Frank-Wolfe Conditional Gradient Prediction with Machine Learning addresses a central…
ai safety
Free Adversarial Training is a method that simultaneous updates both the model parameters and the adversarial perturbation in each gradient computation — reusing the same backward…
plasma dispersion effect, free carrier effect silicon, silicon modulator physics, carrier injection depletion modulator, silicon photonics electro optic
The free-carrier plasma dispersion effect is the physical mechanism that makes silicon photonic modulation possible at all, despite silicon being an otherwise poor material choice…
environmental & sustainability
Free Cooling is cooling strategy that uses favorable ambient conditions to reduce mechanical refrigeration load - It lowers energy consumption by exploiting naturally cool air…
conduction electron, free electron model, Drude model, Sommerfeld model, thermal generation, electron gas, drift velocity
A free electron is an electron that is not bound to any particular atom and so can move through a material and carry electric current when a field is applied. In a metal such as…
healthcare ai
Free Energy Calculations (specifically Free Energy Perturbation, FEP) represent the absolute gold standard in computational drug discovery for quantifying binding affinity…
fto, legal
Freedom to operate is the legal assessment that a product can be made used and sold without infringing active third-party rights - Claim mapping compares planned product features…
semi-supervised learning
FreeMatch is a semi-supervised learning algorithm that uses a self-adaptive global threshold and class-specific thresholds — automatically adjusting confidence thresholds based on…
FreeMatch Self-Adaptive Thresholding and Fairness Introduction & Motivation FreeMatch Self-Adaptive Thresholding and Fairness addresses a central problem in semisupervised…
lyophilization, sublimation drying
Freeze Drying (Lyophilization) in semiconductor processing uses sublimation to dry delicate structures, avoiding surface tension damage from liquid evaporation. What Is Freeze…
device physics
Freeze-out is the extreme low-temperature condition where thermal energy is insufficient to ionize dopant atoms, causing carrier concentration to collapse exponentially…
defects
Frenkel Pair is the fundamental unit of radiation and ion-implant damage — a coupled vacancy-interstitial defect formed when a lattice atom is displaced from its site by…
absolute frequency, count, tally, number of occurrences, frequency count, observed frequency, class frequency
Frequency, in statistics, is the number of times a particular value, category, or class of values occurs in a dataset — the raw count of observations that fall into a given bin.…
frequency dependent behavior prediction, frequency dependent behavior, dependent behavior prediction, frequency dependent, dependent behavior, behavior prediction
Frequency Dependent Behavior Prediction with Machine Learning Introduction & Motivation Frequency Dependent Behavior Prediction with Machine Learning addresses a central problem…
distribution of data, empirical distribution, grouped frequency distribution, relative frequency distribution, class intervals, binning
A frequency distribution is a complete summary of how the observations in a dataset are spread across the possible values or classes, pairing each distinct value (or interval…
clock divider, divide by N, PLL divider, prescaler
Frequency divider. produces an output whose edge rate or phase progression is an integer or fractional relationship to an input clock, most commonly one edge cycle for every N…
Frequency Domain Anomaly Detection Introduction & Motivation Frequency Domain Anomaly Detection addresses a central problem in anomaly detection and {target} scoring: how…
frequency domain noise prediction, frequency domain noise, domain noise prediction, frequency domain, domain noise, noise prediction
Frequency Domain Noise Prediction with Machine Learning Introduction & Motivation Frequency Domain Noise Prediction with Machine Learning addresses a central problem in frequency…
frequency offset correction accuracy prediction, frequency offset correction, offset correction accuracy, correction accuracy prediction, frequency offset
Frequency Offset Correction Accuracy Prediction with Machine Learning Introduction & Motivation Frequency Offset Correction Accuracy Prediction with Machine Learning addresses…
text generation
Frequency penalty is the decoding adjustment that lowers token likelihood proportionally to how often the token has already appeared in the generated output - it controls overuse…