arXiv:2609.28358v1 Announce Type: cross Abstract: Microscaling quantization techniques are increasingly used to represent neural network parameters with 8 bits or fewer while preserving near-full pre...
arXiv:2608.20725v1 Announce Type: cross Abstract: Convolution is a principal computational bottleneck in deep neural networks, and its efficiency depends on tight integration between algorithms and G...
arXiv:2605. 24391v2 Announce Type: replace-cross Abstract: As the demand for deep learning grows, cost reduction through quantization has become essential for both training and inference.
FAME is an FPGA-based platform that evaluates approximate multipliers directly in hardware, eliminating slow CPU/GPU LUT emulation and reducing evaluation time for DNN inference. It also introduces a pattern-guided retraining method that uses multiplier-specific patterns to recover accuracy losses. Experiments on ResNet‑18 and MobileNetV2 over ImageNet show up to 3.47× faster multiplier evaluation and a 65.5% accuracy improvement over prior retraining approaches.
arXiv:2606. 00130v2 Announce Type: replace-cross Abstract: Large deep neural networks are costly to store and deploy because inference must move and evaluate many parameters.
FORGE is a forward‑only test‑time adaptation technique designed for integer‑only vision models running on microcontrollers. It restores batch‑normalization statistics after BN folding by re‑normalizing each convolution’s per‑channel output using only forward‑pass estimates, enabling adaptation on deployed, folded integer models. The method achieves accuracy gains comparable to gradient‑based TENT, requires adapting only a few layers, works with single‑sample streaming, and has been validated on an ESP32‑S3 with minimal energy and latency overhead.
arXiv:2606. 00130v1 Announce Type: cross Abstract: We study Automatically Differentiable Nonlinear Tensor Networks (ADNTNs), a family of structured weight generators whose compact core tensors are trained end-to-end by reverse-mode automatic differentiation (AD).
arXiv:2607. 10611v1 Announce Type: new Abstract: Training with quantized weights can reduce costs but often results in degraded accuracy, especially when optimization is carried out in low precision, without storing high-precision copies.
ShiftQAT is a quantization‑aware training workflow designed for the MAX7800x family of fixed‑function CNN accelerators. It enforces the accelerator’s numerical contract—power‑of‑two, per‑tensor symmetric scales, 8‑bit activations and biases, integer‑only arithmetic, and a single statically configured shift per layer—by training weights and biases on a joint power‑of‑two grid and resolving dependencies so that activation scales propagate statically. Compared to other methods, ShiftQAT uniquely satisfies all contract constraints while matching the accuracy of less constrained QAT approaches, achieving ImageNet performance within 2.9 percentage points of FP32 at W4A8 and enabling fast, energy‑efficient inference on MAX78002 silicon.
arXiv:2509. 10334v2 Announce Type: replace-cross Abstract: Vision Transformers (ViTs) have recently achieved strong results in semantic segmentation, yet their deployment on resource-constrained devices remains limited due to their high memory footprint and computational cost.
arXiv:2610.15092v1 Announce Type: cross Abstract: Quantization plays a crucial role in efficiently deploying deep learning models on resource-constrained devices. Post-training quantization does not...
The paper introduces RAMP, a method for robust adaptive mixed‑precision quantization of vision models on edge CPUs. It evaluates 13 sensitivity metrics across four neural networks, finding that Jensen‑Shannon Divergence consistently identifies layers that can be safely quantized. Using K‑Means clustering on these metrics, RAMP achieves near‑lossless accuracy with an average 1.81× speed‑up, while cautioning against excluding low‑speed‑up layers that can fragment the computational graph.