Neurally Augmented ALISTA
Freya Behrens, Jonathan Sauder, Peter Jung

TL;DR
This paper introduces Neurally Augmented ALISTA, which uses an LSTM to adaptively compute parameters for each target, improving sparse reconstruction performance over existing methods especially at higher compression ratios.
Contribution
It proposes a novel adaptive method that combines ALISTA with an LSTM to enhance individual target reconstruction accuracy in compressed sensing.
Findings
Outperforms existing algorithms in sparse reconstruction tasks.
Shows increased margin of improvement at higher compression ratios.
Theoretically justified by revisiting ALISTA's recovery guarantees.
Abstract
It is well-established that many iterative sparse reconstruction algorithms can be unrolled to yield a learnable neural network for improved empirical performance. A prime example is learned ISTA (LISTA) where weights, step sizes and thresholds are learned from training data. Recently, Analytic LISTA (ALISTA) has been introduced, combining the strong empirical performance of a fully learned approach like LISTA, while retaining theoretical guarantees of classical compressed sensing algorithms and significantly reducing the number of parameters to learn. However, these parameters are trained to work in expectation, often leading to suboptimal reconstruction of individual targets. In this work we therefore introduce Neurally Augmented ALISTA, in which an LSTM network is used to compute step sizes and thresholds individually for each target vector during reconstruction. This adaptive…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
Taxonomy
TopicsSparse and Compressive Sensing Techniques · Medical Imaging Techniques and Applications · Photoacoustic and Ultrasonic Imaging
MethodsTanh Activation · Sigmoid Activation · Long Short-Term Memory
