Abstract
Instance discrimination contrastive learning (CL) has achieved significant success in learning transferable representations. A hardness-aware property related to the temperature $ \tau $ of the CL loss is identified to play an essential role in automatically concentrating on hard negative samples. However, previous work also proves that there exists a uniformity-tolerance dilemma (UTD) in CL loss, which will lead to unexpected performance degradation. Specifically, a smaller temperature helps to learn separable embeddings but has less tolerance to semantically related samples, which may result in suboptimal embedding space, and vice versa. In this paper, we propose a Model-Aware Contrastive Learning (MACL) strategy to escape UTD. For the undertrained phases, there is less possibility that the high similarity region of the anchor contains latent positive samples. Thus, adopting a small temperature in these stages can impose larger penalty strength on hard negative samples to improve the discrimination of the CL model. In contrast, a larger temperature in the well-trained phases helps to explore semantic structures due to more tolerance to potential positive samples. During implementation, the temperature in MACL is designed to be adaptive to the alignment property that reflects the confidence of a CL model. Furthermore, we reexamine why contrastive learning requires a large number of negative samples in a unified gradient reduction perspective. Based on MACL and these analyses, a new CL loss is proposed in this work to improve the learned representations and training with small batch size.
Mind the Gap: Polishing Pseudo labels for Accurate Semi-supervised Object Detection
Authors: Lei Zhang, Yuxuan Sun, Wei Wei
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
Exploiting pseudo labels (e.g., categories and bounding boxes) of unannotated objects produced by a teacher detector have underpinned much of recent progress in semi-supervised object detection (SSOD). However, due to the limited generalization capacity of the teacher detector caused by the scarce annotations, the produced pseudo labels often deviate from ground truth, especially those with relatively low classification confidences, thus limiting the generalization performance of SSOD. To mitigate this problem, we propose a dual pseudo-label polishing framework for SSOD. Instead of directly exploiting the pseudo labels produced by the teacher detector, we take the first attempt at reducing their deviation from ground truth using dual polishing learning, where two differently structured polishing networks are elaborately developed and trained using synthesized paired pseudo labels and the corresponding ground truth for categories and bounding boxes on the given annotated objects, respectively. By doing this, both polishing networks can infer more accurate pseudo labels for unannotated objects through sufficiently exploiting their context knowledge based on the initially produced pseudo labels, and thus improve the generalization performance of SSOD. Moreover, such a scheme can be seamlessly plugged into the existing SSOD framework for joint end-to-end learning. In addition, we propose to disentangle the polished pseudo categories and bounding boxes of unannotated objects for separate category classification and bounding box regression in SSOD, which enables introducing more unannotated objects during model training and thus further improve the performance. Experiments on both PASCAL VOC and MS COCO benchmarks demonstrate the superiority of the proposed method over existing state-of-the-art baselines.
Active Exploration for Inverse Reinforcement Learning
Authors: David Lindner, Andreas Krause, Giorgia Ramponi
Abstract
Inverse Reinforcement Learning (IRL) is a powerful paradigm for inferring a reward function from expert demonstrations. Many IRL algorithms require a known transition model and sometimes even a known expert policy, or they at least require access to a generative model. However, these assumptions are too strong for many real-world applications, where the environment can be accessed only through sequential interaction. We propose a novel IRL algorithm: Active exploration for Inverse Reinforcement Learning (AceIRL), which actively explores an unknown environment and expert policy to quickly learn the expert's reward function and identify a good policy. AceIRL uses previous observations to construct confidence intervals that capture plausible reward functions and find exploration policies that focus on the most informative regions of the environment. AceIRL is the first approach to active IRL with sample-complexity bounds that does not require a generative model of the environment. AceIRL matches the sample complexity of active IRL with a generative model in the worst case. Additionally, we establish a problem-dependent bound that relates the sample complexity of AceIRL to the suboptimality gap of a given IRL problem. We empirically evaluate AceIRL in simulations and find that it significantly outperforms more naive exploration strategies.
Keyword: scaling
An Experimental Evaluation of Machine Learning Training on a Real Processing-in-Memory System
Authors: Juan Gómez-Luna, Yuxin Guo, Sylvan Brocard, Julien Legriel, Remy Cimadomo, Geraldo F. Oliveira, Gagandeep Singh, Onur Mutlu
Abstract
Training machine learning (ML) algorithms is a computationally intensive process, which is frequently memory-bound due to repeatedly accessing large training datasets. As a result, processor-centric systems (e.g., CPU, GPU) suffer from costly data movement between memory units and processing units, which consumes large amounts of energy and execution cycles. Memory-centric computing systems, i.e., with processing-in-memory (PIM) capabilities, can alleviate this data movement bottleneck. Our goal is to understand the potential of modern general-purpose PIM architectures to accelerate ML training. To do so, we (1) implement several representative classic ML algorithms (namely, linear regression, logistic regression, decision tree, K-Means clustering) on a real-world general-purpose PIM architecture, (2) rigorously evaluate and characterize them in terms of accuracy, performance and scaling, and (3) compare to their counterpart implementations on CPU and GPU. Our evaluation on a real memory-centric computing system with more than 2500 PIM cores shows that general-purpose PIM architectures can greatly accelerate memory-bound ML workloads, when the necessary operations and datatypes are natively supported by PIM hardware. For example, our PIM implementation of decision tree is $27\times$ faster than a state-of-the-art CPU version on an 8-core Intel Xeon, and $1.34\times$ faster than a state-of-the-art GPU version on an NVIDIA A100. Our K-Means clustering on PIM is $2.8\times$ and $3.2\times$ than state-of-the-art CPU and GPU versions, respectively. To our knowledge, our work is the first one to evaluate ML training on a real-world PIM architecture. We conclude with key observations, takeaways, and recommendations that can inspire users of ML workloads, programmers of PIM architectures, and hardware designers & architects of future memory-centric computing systems.
Abstract
Applications of machine learning (ML) are growing by the day for many unique and challenging scientific applications. However, a crucial challenge facing these applications is their need for ultra low-latency and on-detector ML capabilities. Given the slowdown in Moore's law and Dennard scaling, coupled with the rapid advances in scientific instrumentation that is resulting in growing data rates, there is a need for ultra-fast ML at the extreme edge. Fast ML at the edge is essential for reducing and filtering scientific data in real-time to accelerate science experimentation and enable more profound insights. To accelerate real-time scientific edge ML hardware and software solutions, we need well-constrained benchmark tasks with enough specifications to be generically applicable and accessible. These benchmarks can guide the design of future edge ML hardware for scientific applications capable of meeting the nanosecond and microsecond level latency requirements. To this end, we present an initial set of scientific ML benchmarks, covering a variety of ML and embedded system techniques.
Monitoring Vegetation From Space at Extremely Fine Resolutions via Coarsely-Supervised Smooth U-Net
Authors: Joshua Fan, Di Chen, Jiaming Wen, Ying Sun, Carla P. Gomes
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Abstract
Monitoring vegetation productivity at extremely fine resolutions is valuable for real-world agricultural applications, such as detecting crop stress and providing early warning of food insecurity. Solar-Induced Chlorophyll Fluorescence (SIF) provides a promising way to directly measure plant productivity from space. However, satellite SIF observations are only available at a coarse spatial resolution, making it impossible to monitor how individual crop types or farms are doing. This poses a challenging coarsely-supervised regression (or downscaling) task; at training time, we only have SIF labels at a coarse resolution (3km), but we want to predict SIF at much finer spatial resolutions (e.g. 30m, a 100x increase). We also have additional fine-resolution input features, but the relationship between these features and SIF is unknown. To address this, we propose Coarsely-Supervised Smooth U-Net (CS-SUNet), a novel method for this coarse supervision setting. CS-SUNet combines the expressive power of deep convolutional networks with novel regularization methods based on prior knowledge (such as a smoothness loss) that are crucial for preventing overfitting. Experiments show that CS-SUNet resolves fine-grained variations in SIF more accurately than existing methods.
Progress and limitations of deep networks to recognize objects in unusual poses
Authors: Amro Abbas, Stéphane Deny
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
Deep networks should be robust to rare events if they are to be successfully deployed in high-stakes real-world applications (e.g., self-driving cars). Here we study the capability of deep networks to recognize objects in unusual poses. We create a synthetic dataset of images of objects in unusual orientations, and evaluate the robustness of a collection of 38 recent and competitive deep networks for image classification. We show that classifying these images is still a challenge for all networks tested, with an average accuracy drop of 29.5% compared to when the objects are presented upright. This brittleness is largely unaffected by various network design choices, such as training losses (e.g., supervised vs. self-supervised), architectures (e.g., convolutional networks vs. transformers), dataset modalities (e.g., images vs. image-text pairs), and data-augmentation schemes. However, networks trained on very large datasets substantially outperform others, with the best network tested$\unicode{x2014}$Noisy Student EfficentNet-L2 trained on JFT-300M$\unicode{x2014}$showing a relatively small accuracy drop of only 14.5% on unusual poses. Nevertheless, a visual inspection of the failures of Noisy Student reveals a remaining gap in robustness with the human visual system. Furthermore, combining multiple object transformations$\unicode{x2014}$3D-rotations and scaling$\unicode{x2014}$further degrades the performance of all networks. Altogether, our results provide another measurement of the robustness of deep networks that is important to consider when using them in the real world. Code and datasets are available at https://github.com/amro-kamal/ObjectPose.
Performance degradation of ImageNet trained models by simple image transformations
Authors: Harsh Maheshwari
Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Abstract
ImageNet trained PyTorch models are generally preferred as the off-the-shelf models for direct use or for initialisation in most computer vision tasks. In this paper, we simply test a representative set of these convolution and transformer based models under many simple image transformations like horizontal shifting, vertical shifting, scaling, rotation, presence of Gaussian noise, cutout, horizontal flip and vertical flip and report the performance drop caused by such transformations. We find that even simple transformations like rotating the image by 10{\deg} or zooming in by 20% can reduce the top-1 accuracy of models like ResNet152 by 1%+. The code is available at https://github.com/harshm121/imagenet-transformation-degradation.
E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context
Abstract
Recently, the image-wise implicit neural representation of videos, NeRV, has gained popularity for its promising results and swift speed compared to regular pixel-wise implicit representations. However, the redundant parameters within the network structure can cause a large model size when scaling up for desirable performance. The key reason of this phenomenon is the coupled formulation of NeRV, which outputs the spatial and temporal information of video frames directly from the frame index input. In this paper, we propose E-NeRV, which dramatically expedites NeRV by decomposing the image-wise implicit neural representation into separate spatial and temporal context. Under the guidance of this new formulation, our model greatly reduces the redundant model parameters, while retaining the representation ability. We experimentally find that our method can improve the performance to a large extent with fewer parameters, resulting in a more than $8\times$ faster speed on convergence. Code is available at https://github.com/kyleleey/E-NeRV.
Can large language models reason about medical questions?
Authors: Valentin Liévin, Christoffer Egeberg Hother, Ole Winther
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Abstract
Although large language models (LLMs) often produce impressive outputs, they also fail to reason and be factual. We set out to investigate how these limitations affect the LLM's ability to answer and reason about difficult real-world based questions. We applied the human-aligned GPT-3 (InstructGPT) to answer multiple-choice medical exam questions (USMLE and MedMCQA) and medical research questions (PubMedQA). We investigated Chain-of-thought (think step by step) prompts, grounding (augmenting the prompt with search results) and few-shot (prepending the question with question-answer exemplars). For a subset of the USMLE questions, a medical domain expert reviewed and annotated the model's reasoning. Overall, GPT-3 achieved a substantial improvement in state-of-the-art machine learning performance. We observed that GPT-3 is often knowledgeable and can reason about medical questions. GPT-3, when confronted with a question it cannot answer, will still attempt to answer, often resulting in a biased predictive distribution. LLMs are not on par with human performance but our results suggest the emergence of reasoning patterns that are compatible with medical problem-solving. We speculate that scaling model and data, enhancing prompt alignment and allowing for better contextualization of the completions will be sufficient for LLMs to reach human-level performance on this type of task.
Shrunk subspaces via operator Sinkhorn iteration
Authors: Cole Franks, Tasuku Soma, Michel X. Goemans
Abstract
A recent breakthrough in Edmonds' problem showed that the noncommutative rank can be computed in deterministic polynomial time, and various algorithms for it were devised. However, only quite complicated algorithms are known for finding a so-called shrunk subspace, which acts as a dual certificate for the value of the noncommutative rank. In particular, the operator Sinkhorn algorithm, perhaps the simplest algorithm to compute the noncommutative rank with operator scaling, does not find a shrunk subspace. Finding a shrunk subspace plays a key role in applications, such as separation in the Brascamp-Lieb polytope, one-parameter subgroups in the null-cone membership problem, and primal-dual algorithms for matroid intersection and fractional matroid matching. In this paper, we provide a simple Sinkhorn-style algorithm to find the smallest shrunk subspace over the complex field in deterministic polynomial time. To this end, we introduce a generalization of the operator scaling problem, where the spectra of the marginals must be majorized by specified vectors. Then we design an efficient Sinkhorn-style algorithm for the generalized operator scaling problem. Applying this to the shrunk subspace problem, we show that a sufficiently long run of the algorithm also finds an approximate shrunk subspace close to the minimum exact shrunk subspace. Finally, we show that the approximate shrunk subspace can be rounded if it is sufficiently close. Along the way, we also provide a simple randomized algorithm to find the smallest shrunk subspace. As applications, we design a faster algorithm for fractional linear matroid matching and efficient weak membership and optimization algorithms for the rank-2 Brascamp-Lieb polytope.
Efficient View Clustering and Selection for City-Scale 3D Reconstruction
Authors: Marco Orsingher, Paolo Zani, Paolo Medici, Massimo Bertozzi
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
Image datasets have been steadily growing in size, harming the feasibility and efficiency of large-scale 3D reconstruction methods. In this paper, a novel approach for scaling Multi-View Stereo (MVS) algorithms up to arbitrarily large collections of images is proposed. Specifically, the problem of reconstructing the 3D model of an entire city is targeted, starting from a set of videos acquired by a moving vehicle equipped with several high-resolution cameras. Initially, the presented method exploits an approximately uniform distribution of poses and geometry and builds a set of overlapping clusters. Then, an Integer Linear Programming (ILP) problem is formulated for each cluster to select an optimal subset of views that guarantees both visibility and matchability. Finally, local point clouds for each cluster are separately computed and merged. Since clustering is independent from pairwise visibility information, the proposed algorithm runs faster than existing literature and allows for a massive parallelization. Extensive testing on urban data are discussed to show the effectiveness and the scalability of this approach.
Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit
Authors: Boaz Barak, Benjamin L. Edelman, Surbhi Goel, Sham Kakade, Eran Malach, Cyril Zhang
Subjects: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Optimization and Control (math.OC); Machine Learning (stat.ML)
Abstract
There is mounting empirical evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning $k$-sparse parities of $n$ bits, a canonical family of problems which pose theoretical computational barriers. In this setting, we find that neural networks exhibit surprising phase transitions when scaling up dataset size and running time. In particular, we demonstrate empirically that with standard training, a variety of architectures learn sparse parities with $n^{O(k)}$ examples, with loss (and error) curves abruptly dropping after $n^{O(k)}$ iterations. These positive results nearly match known SQ lower bounds, even without an explicit sparsity-promoting prior. We elucidate the mechanisms of these phenomena with a theoretical analysis: we find that the phase transition in performance is not due to SGD "stumbling in the dark" until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time); instead, we show that SGD gradually amplifies a Fourier gap in the population gradient.
Keyword: calibration
Angular Gap: Reducing the Uncertainty of Image Difficulty through Model Calibration
Authors: Bohua Peng, Mobarakol Islam, Mei Tu
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
Curriculum learning needs example difficulty to proceed from easy to hard. However, the credibility of image difficulty is rarely investigated, which can seriously affect the effectiveness of curricula. In this work, we propose Angular Gap, a measure of difficulty based on the difference in angular distance between feature embeddings and class-weight embeddings built by hyperspherical learning. To ascertain difficulty estimation, we introduce class-wise model calibration, as a post-training technique, to the learnt hyperbolic space. This bridges the gap between probabilistic model calibration and angular distance estimation of hyperspherical learning. We show the superiority of our calibrated Angular Gap over recent difficulty metrics on CIFAR10-H and ImageNetV2. We further propose Angular Gap based curriculum learning for unsupervised domain adaptation that can translate from learning easy samples to mining hard samples. We combine this curriculum with a state-of-the-art self-training method, Cycle Self Training (CST). The proposed Curricular CST learns robust representations and outperforms recent baselines on Office31 and VisDA 2017.
Keyword: out of distribution detection
There is no result
Keyword: out-of-distribution detection
There is no result
Keyword: expected calibration error
There is no result
Keyword: overconfident
There is no result
Keyword: overconfidence
There is no result
Keyword: confidence
Model-Aware Contrastive Learning: Towards Escaping Uniformity-Tolerance Dilemma in Training
Mind the Gap: Polishing Pseudo labels for Accurate Semi-supervised Object Detection
Active Exploration for Inverse Reinforcement Learning
Keyword: scaling
An Experimental Evaluation of Machine Learning Training on a Real Processing-in-Memory System
FastML Science Benchmarks: Accelerating Real-Time Scientific Edge Machine Learning
Monitoring Vegetation From Space at Extremely Fine Resolutions via Coarsely-Supervised Smooth U-Net
Progress and limitations of deep networks to recognize objects in unusual poses
Performance degradation of ImageNet trained models by simple image transformations
E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context
Can large language models reason about medical questions?
Shrunk subspaces via operator Sinkhorn iteration
Efficient View Clustering and Selection for City-Scale 3D Reconstruction
Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit
Keyword: calibration
Angular Gap: Reducing the Uncertainty of Image Difficulty through Model Calibration