We present Caffe con Troll (CcT), a fully compatible end-to-end version of the popular framework Caffe with rebuilt internals. We built CcT to examine the performance characteristics of training and deploying general-purpose convolutional neural networks across different hardware architectures. We find that, by employi…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
169,181 papers · 148 categories
Trend · papers per month
15 results for “Caffe”
Barista simplifies deep neural network design and training.
problem Complexity and error-proneness of editing deep learning frameworks.
method Graphical user interface for Caffe framework.
result Researchers can focus on solving problems without coding.
Training deep networks is a time-consuming process, with networks for object recognition often requiring multiple days to train. For this reason, leveraging the resources of a cluster to speed up training is an important area of work. However, widely-popular batch-processing computational frameworks like MapReduce and …
μ-cuDNN accelerates deep learning frameworks with micro-batching.
problem Slow performance and memory constraints in GPU kernels.
method Divides layers' mini-batch computation into micro-batches using Dynamic Programming and Integer Linear Programming.
result Achieved speedups of 1.63x for AlexNet and 1.21x for ResNet-18 on P100-SXM2 GPU.
Poseidon: An Efficient Communication Architecture for Distributed Deep Learning on GPU Clusterscs.LG
Deep learning models can take weeks to train on a single GPU-equipped machine, necessitating scaling out DL training to a GPU-cluster. However, current distributed DL implementations can scale poorly due to substantial parameter synchronization over the network, because the high throughput of GPUs allows more data batc…
DLPaper2Code converts deep learning papers into executable code.
problem Lack of open source implementations and difficulty in re-implementing research papers.
method Extracts and converts design flow diagrams and tables from research papers into executable code.
result The approach achieves over 93% accuracy in converting design visualizations.
A new method for verifying deep learning architectures on FPGAs is proposed.
problem Design-time verification of deep learning architectures on FPGAs.
method 2-Level 3-Way (2L-3W) hardware-software co-verification methodology.
result Layer-by-layer similarity scores of 99% accuracy for successful mappings.
Improved DNN training with BCD converging faster than SGD.
problem Training deep neural networks efficiently and accurately.
method Developed a smooth multi-convex formulation and BCD algorithm for training Tikhonov regularized DNNs.
result BCD algorithm converges globally to a stationary point with R-linear rate.
A new method for efficient CNN compression by incrementally assigning regularization factors to weights.
problem Efficiently compressing ConvNets while maintaining performance.
method IncReg: Incremental Regularization for Parameter Pruning.
result IncReg achieves comparable or better results than state-of-the-art methods.
High demand for computation resources severely hinders deployment of large-scale Deep Neural Networks (DNN) in resource constrained devices. In this work, we propose a Structured Sparsity Learning (SSL) method to regularize the structures (i.e., filters, channels, filter shapes, and layer depth) of DNNs. SSL can: (1) l…
Transforms improve CNNs' invariance to image transformations.
problem Current CNN models lack robustness to spatial transformations.
method Randomly transform feature maps during training to learn invariant representations.
result Significant improvements on benchmark tasks, including image recognition and retrieval.
Patch ranking improves CNN performance by focusing on object content, not location.
problem CNNs lack rotation and translation invariance, limiting model capacity.
method Patch ranking before convolution and pooling to encode invariance.
result Patch ranking module improves CNN performance on various tasks.
Continuous Dropoutcs.CV
Continuous dropout mimics brain neuron firing rates to prevent feature detector co-adaptation.
problem Preventing overfitting in deep neural networks.
method Extending binary dropout to continuous dropout, inspired by brain neuron firing rates.
result Continuous dropout improves test performance by preventing feature detector co-adaptation.
This paper studies learning rate policies for deep neural networks, offering metrics and tools for better tuning.
problem Effective tuning of learning rates for deep neural networks is challenging and crucial for achieving high accuracy.
method Comprehensive study of 13 learning rate functions, proposing metrics for evaluation, and developing LRBench for benchmarking and selection.
result Identification of good learning rate policies with effective ranges and step sizes for various LR update schedules.
New method reduces deep learning complexity on IoT devices.
problem High computational complexity limits deep learning on IoT devices.
method Local quantization region for low-bit data representation.
result Models retain accuracy with reduced computational complexity.