skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Principal Component Networks: Parameter Reduction Early in Training
In this paper, we show that hidden layer activations in overparameterized neural networks for image classification exist primarily in subspaces smaller than the actual model width. We further show that these subspaces can be identified early in training. Based on these observations, we show how to efficiently find small networks that exhibit similar accuracy to their overparameterized counterparts after only a few training epochs. We term these network architectures Principal Component Networks (PCNs). We evaluate PCNs on CIFAR-10 and ImageNet for VGG and ResNet style architectures and find that PCNs consistently reduce parameter counts with little accuracy loss, thus providing the potential to reduce the compu- tational costs of deep neural network training.  more » « less
Award ID(s):
1815538
PAR ID:
10385069
Author(s) / Creator(s):
;
Date Published:
Journal Name:
International Conference on Machine Learning Workshop on Hardware Aware Efficient Training
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; (, Advances in neural information processing systems)
    null (Ed.)
    Several works have aimed to explain why overparameterized neural networks generalize well when trained by Stochastic Gradient Descent (SGD). The consensus explanation that has emerged credits the randomized nature of SGD for the bias of the training process towards low-complexity models and, thus, for implicit regularization. We take a careful look at this explanation in the context of image classification with common deep neural network architectures. We find that if we do not regularize explicitly, then SGD can be easily made to converge to poorly-generalizing, high-complexity models: all it takes is to first train on a random labeling on the data, before switching to properly training with the correct labels. In contrast, we find that in the presence of explicit regularization, pretraining with random labels has no detrimental effect on SGD. We believe that our results give evidence that explicit regularization plays a far more important role in the success of overparameterized neural networks than what has been understood until now. Specifically, by penalizing complicated models independently of their fit to the data, regularization affects training dynamics also far away from optima, making simple models that fit the data well discoverable by local methods, such as SGD. 
    more » « less
  2. ; ; ; ; ; (, SPIE Defense + Commercial Sensing)
    Markopoulos, Panos P.; Ouyang, Bing (Ed.)
    We consider the problem of unsupervised (blind) evaluation and assessment of the quality of data used for deep neural network (DNN) RF signal classification. When neural networks train on noisy or mislabeled data, they often (over-)fit to the noise measurements and faulty labels, which leads to significant performance degradation. Also, DNNs are vulnerable to adversarial attacks, which can considerably reduce their classification performance, with extremely small perturbations of their input. In this paper, we consider a new method based on L1-norm principal-component analysis (PCA) to improve the quality of labeled wireless data sets that are used for training a convolutional neural network (CNN), and a deep residual network (ResNet) for RF signal classification. Experiments with data generated for eleven classes of digital and analog modulated signals show that L1-norm tensor conformity curation of the data identifies and removes from the training data set inappropriate class instances that appear due to mislabeling and universal black-box adversarial attacks and drastically improves/restores the classification accuracy of the identified deep neural network architectures. 
    more » « less
  3. ; ; ; ; ; ; (, Journal of Cheminformatics)
    Abstract Modern data mining techniques using machine learning (ML) and deep learning (DL) algorithms have been shown to excel in the regression-based task of materials property prediction using various materials representations. In an attempt to improve the predictive performance of the deep neural network model, researchers have tried to add more layers as well as develop new architectural components to create sophisticated and deep neural network models that can aid in the training process and improve the predictive ability of the final model. However, usually, these modifications require a lot of computational resources, thereby further increasing the already large model training time, which is often not feasible, thereby limiting usage for most researchers. In this paper, we study and propose a deep neural network framework for regression-based problems comprising of fully connected layers that can work with any numerical vector-based materials representations as model input. We present a novel deep regression neural network, iBRNet, with branched skip connections and multiple schedulers, which can reduce the number of parameters used to construct the model, improve the accuracy, and decrease the training time of the predictive model. We perform the model training using composition-based numerical vectors representing the elemental fractions of the respective materials and compare their performance against other traditional ML and several known DL architectures. Using multiple datasets with varying data sizes for training and testing, We show that the proposed iBRNet models outperform the state-of-the-art ML and DL models for all data sizes. We also show that the branched structure and usage of multiple schedulers lead to fewer parameters and faster model training time with better convergence than other neural networks. Scientific contribution: The combination of multiple callback functions in deep neural networks minimizes training time and maximizes accuracy in a controlled computational environment with parametric constraints for the task of materials property prediction. 
    more » « less
  4. (, Electronic proceedings in theoretical computer science)
    The integration of low-level perception with high-level reasoning is one of the oldest problems in Artificial Intelligence. Recently, several proposals were made to implement the reasoning process in complex neural network architectures. While these works aim at extending neural networks with the capability of reasoning, a natural question that we consider is: can we extend answer set programs with neural networks to allow complex and high-level reasoning on neural network outputs? As a preliminary result, we propose NeurASP – a simple extension of answer set programs by embracing neural networks where neural network outputs are treated as probability distributions over atomic facts in answer set programs. We show that NeurASP can not only improve the perception accuracy of a pre-trained neural network, but also help to train a neural network better by giving restrictions through logic rules. However, training with NeurASP would take much more time than pure neural network training due to the internal use of a symbolic reasoning engine. For future work, we plan to investigate the potential ways to solve the scalability issue of NeurASP. One potential way is to embed logic programs directly in neural networks. On this route, we plan to first design a SAT solver using neural networks, then extend such a solver to allow logic programs. 
    more » « less
  5. ; ; ; (, Information and Inference: A Journal of the IMA)
    Abstract Empirical evidence suggests that for a variety of overparameterized nonlinear models, most notably in neural network training, the growth of the loss around a minimizer strongly impacts its performance. Flat minima—those around which the loss grows slowly—appear to generalize well. This work takes a step towards understanding this phenomenon by focusing on the simplest class of overparameterized nonlinear models: those arising in low-rank matrix recovery. We analyse overparameterized matrix and bilinear sensing, robust principal component analysis, covariance matrix estimation and single hidden layer neural networks with quadratic activation functions. In all cases, we show that flat minima, measured by the trace of the Hessian, exactly recover the ground truth under standard statistical assumptions. For matrix completion, we establish weak recovery, although empirical evidence suggests exact recovery holds here as well. We complete the paper with synthetic experiments that illustrate our findings. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email