Matching Component Analysis for Transfer Learning

Abstract

We introduce a new Procrustes-type method called matching component analysis to isolate components in data for transfer learning. Our theoretical results describe the sample complexity of this method, and we demonstrate through numerical experiments that our approach is indeed well suited for transfer learning. IntroductionMany state-of-the-art classification algorithms require a large training set that is statistically similar to the test set. For example, deep learning-based approaches require a large number of representative samples in order to find near-optimal network weights and biases [4,9]. Similarly, template-based approaches require large dictionaries of training images so that each test image can be represented by an element of the dictionary [21,26,17,6]. For each technique, if test images cannot be represented in a feature space that has been determined from the training set, then classification accuracy is poor.In applications such as synthetic aperture radar (SAR) automatic target recognition (ATR), it is infeasible to collect the volume of data necessary to naively train high-accuracy classification networks. Additionally, due to varying operating conditions, the features measured in SAR imagery are different from those extracted from electro-optical (EO) imagery [13]. As such, off-the-shelf networks that have been pre-trained on the popular EObased ImageNet [3] or CIFAR-10 [8] datasets are insufficient for performing accurate ATR tasks in different imaging domains. In fact, recent work has demonstrated that pre-trained networks fail to effectively generalize to random perturbations on test sets [19,18]. To build more representative training sets, additional data are often generated using modeling and simulation software. However, due to various model errors, simulated data often misrepresent the real-world scattering observed in measured imagery. Thus, even though it is possible to augment training sets with a large amount of simulated data, the inherent differences in sensor modalities and data representations make modifying classification networks a non-trivial task [20].In this paper, we introduce matching component analysis (MCA) to help remedy this situation. Given a small number of images from the training domain and matching *

Bridging a Gap in SAR-ATR: Training on Fully Synthetic and Testing on Measured Data

“…Transfer Learning. Another popular technique for leveraging synthetic data is to perform transfer learning (TL) [12], [13], [14]. In most of these works, one trains an initial source model on synthetic SAR data to learn useful and generalizable features of the targets.…”

Covariance-Generalized Matching Component Analysis for Data Fusion and Transfer Learning

“…For the special case C i C T i = I k for i ∈ {1, 2} investigated in [1], we find that A = I k in (34b), so that our solution (34a) reduces to…”

“…In contrast to the identity matrix covariance constraint C i C T i ≡ I k of [1], the generalized covariance constraint (1d) may be used to encode (rather than destroy) both (E1) uncertainty information about the ability of the measurement process underlying the i th data domain to provide information about elements of the common domain, and…”

“…The matching component analysis (MCA) transfer learning technique was originally developed as a data augmentation strategy for building large, representative machine learning training sets within a data-limited environment [1]. Specifically, MCA maps a training domain and a testing domain into a low-dimensional, common domain using only a small number of matched train-test image pairs.…”

“…The optimization problem (1) is our covariance-generalized version of the optimization problem (2.1) of [1] (in which C i C T i ≡ I k ); we seek to minimize the expected distance between the images under g i of the random variables X i within the common domain of dimension k, while centering the resulting mappings g i via (1c) and preserving a prescribed covariance structure via (1d).…”