More Like this

1. Test-Time Training Can Close the Natural Distribution Shift Performance Gap in Deep Learning Based Compressed Sensing

   Mohammad Zalbagi Darestani; Jiayu Liu; Reinhard Heckel (September 2022, International Conference on Machine Learning)

   Deep learning based image reconstruction methods outperform traditional methods. However, neural networks suffer from a performance drop when applied to images from a different distribution than the training images. For example, a model trained for reconstructing knees in accelerated magnetic resonance imaging (MRI) does not reconstruct brains well, even though the same network trained on brains reconstructs brains perfectly well. Thus there is a distribution shift performance gap for a given neural network, defined as the difference in performance when training on a distribution P and training on another distribution Q, and evaluating both models on Q. In this work, we propose a domain adaptation method for deep learning based compressive sensing that relies on self-supervision during training paired with test-time training at inference. We show that for four natural distribution shifts, this method essentially closes the distribution shift performance gap for state-of-the-art architectures for accelerated MRI. 

   more » « less
2. Learning and Communications Co-Design for Remote Inference Systems: Feature Length Selection and Transmission Scheduling

   https://doi.org/10.1109/JSAIT.2023.3322620  

   Shisher, Md Kamran; Ji, Bo; Hou, I-Hong; Sun, Yin (January 2023, IEEE Journal on Selected Areas in Information Theory)

   In this paper, we consider a remote inference system, where a neural network is used to infer a time-varying target (e.g., robot movement), based on features (e.g., video clips) that are progressively received from a sensing node (e.g., a camera). Each feature is a temporal sequence of sensory data. The inference error is determined by (i) the timeliness and (ii) the sequence length of the feature, where we use Age of Information (AoI) as a metric for timeliness. While a longer feature can typically provide better inference performance, it often requires more channel resources for sending the feature. To minimize the time-averaged inference error, we study a learning and communication co-design problem that jointly optimizes feature length selection and transmission scheduling. When there is a single sensor-predictor pair and a single channel, we develop low-complexity optimal co-designs for both the cases of time-invariant and time-variant feature length. When there are multiple sensor-predictor pairs and multiple channels, the co-design problem becomes a restless multi-arm multi-action bandit problem that is PSPACE-hard. For this setting, we design a low-complexity algorithm to solve the problem. Trace-driven evaluations demonstrate the potential of these co-designs to reduce inference error by up to 10000 times. 

   more » « less
3. Test-Time Training with Self-Supervision for Generalization under Distribution Shifts

   Sun, Yu; Wang, Xiaolong; Liu, Zhuang; Miller, John; Efros, Alexei A.; Hardt, Moritz (April 2020, ICML 2020)

   null (Ed.)

   In this paper, we propose Test-Time Training, a general approach for improving the performance of predictive models when training and test data come from different distributions. We turn a single unlabeled test sample into a self-supervised learning problem, on which we update the model parameters before making a prediction. This also extends naturally to data in an online stream. Our simple approach leads to improvements on diverse image classification benchmarks aimed at evaluating robustness to distribution shifts. 

   more » « less
4. Online Structure Learning for Feed-Forward and Recurrent Sum-Product Networks

   Kalra, Agastya; Rashwan, Abdullah; Hsu, Wei-Shou; Poupart, Pascal; Doshi, Prashant; Trimponias, Georgios (December 2018, Advances in Neural Information Processing Systems (NeurIPS))

   Sum-product networks have recently emerged as an attractive representation due to their dual view as a special type of deep neural network with clear semantics and a special type of probabilistic graphical model for which inference is always tractable. Those properties follow from some conditions (i.e., completeness and decomposability) that must be respected by the structure of the network. As a result, it is not easy to specify a valid sum-product network by hand and therefore structure learning techniques are typically used in practice. This paper describes a new online structure learning technique for feed-forward and recurrent SPNs. The algorithm is demonstrated on real-world datasets with continuous features for which it is not clear what network architecture might be best, including sequence datasets of varying length. 

   more » « less
5. Testable Learning with Distribution Shift

   Klivans, A; Stavropoulos, K; Vasilyan, A (May 2024, https://doi.org/10.48550/arXiv.2311.15142)

   We revisit the fundamental problem of learning with distribution shift, in which a learner is given labeled samples from training distribution D, unlabeled samples from test distribution D′ and is asked to output a classifier with low test error. The standard approach in this setting is to bound the loss of a classifier in terms of some notion of distance between D and D′. These distances, however, seem difficult to compute and do not lead to efficient algorithms. We depart from this paradigm and define a new model called testable learning with distribution shift, where we can obtain provably efficient algorithms for certifying the performance of a classifier on a test distribution. In this model, a learner outputs a classifier with low test error whenever samples from D and D′ pass an associated test; moreover, the test must accept if the marginal of D equals the marginal of D′. We give several positive results for learning well-studied concept classes such as halfspaces, intersections of halfspaces, and decision trees when the marginal of D is Gaussian or uniform on {±1}d. Prior to our work, no efficient algorithms for these basic cases were known without strong assumptions on D′. For halfspaces in the realizable case (where there exists a halfspace consistent with both D and D′), we combine a moment-matching approach with ideas from active learning to simulate an efficient oracle for estimating disagreement regions. To extend to the non-realizable setting, we apply recent work from testable (agnostic) learning. More generally, we prove that any function class with low-degree L2-sandwiching polynomial approximators can be learned in our model. We apply constructions from the pseudorandomness literature to obtain the required approximators. 

   more » « less