PUG: Photorealistic and Semantically Controllable Synthetic Data for Representation Learning

Synthetic image datasets offer unmatched advantages for designing and evaluating deep neural networks: they make it possible to (i) render as many data samples as needed, (ii) precisely control each scene and yield granular ground truth labels (and captions), (iii) precisely control distribution shifts between training and testing to isolate variables of interest for sound experimentation. Despite such promise, the use of synthetic image data is still limited -- and often played down -- mainly due to their lack of realism. Most works therefore rely on datasets of real images, which have often been scraped from public images on the internet, and may have issues with regards to privacy, bias, and copyright, while offering little control over how objects precisely appear. In this work, we present a path to democratize the use of photorealistic synthetic data: we develop a new generation of interactive environments for representation learning research, that offer both controllability and realism. We use the Unreal Engine, a powerful game engine well known in the entertainment industry, to produce PUG (Photorealistic Unreal Graphics) environments and datasets for representation learning. In this paper, we demonstrate the potential of PUG to enable more rigorous evaluations of vision models.

Multi-Task Learning Improves Performance In Deep Argument Mining Models

The successful analysis of argumentative techniques from user-generated text is central to many downstream tasks such as political and market analysis. Recent argument mining tools use state-of-the-art deep learning methods to extract and annotate argumentative techniques from various online text corpora, however each task is treated as separate and different bespoke models are fine-tuned for each dataset. We show that different argument mining tasks share common semantic and logical structure by implementing a multi-task approach to argument mining that achieves better performance than state-of-the-art methods for the same problems. Our model builds a shared representation of the input text that is common to all tasks and exploits similarities between tasks in order to further boost performance via parameter-sharing. Our results are important for argument mining as they show that different tasks share substantial similarities and suggest a holistic approach to the extraction of argumentative techniques from text.

Objectives Matter: Understanding the Impact of Self-Supervised Objectives on Vision Transformer Representations

Joint-embedding based learning (e.g., SimCLR, MoCo, DINO) and reconstruction-based learning (e.g., BEiT, SimMIM, MAE) are the two leading paradigms for self-supervised learning of vision transformers, but they differ substantially in their transfer performance. Here, we aim to explain these differences by analyzing the impact of these objectives on the structure and transferability of the learned representations. Our analysis reveals that reconstruction-based learning features are significantly dissimilar to joint-embedding based learning features and that models trained with similar objectives learn similar features even across architectures. These differences arise early in the network and are primarily driven by attention and normalization layers. We find that joint-embedding features yield better linear probe transfer for classification because the different objectives drive different distributions of information and invariances in the learned representation. These differences explain opposite trends in transfer performance for downstream tasks that require spatial specificity in features. Finally, we address how fine-tuning changes reconstructive representations to enable better transfer, showing that fine-tuning re-organizes the information to be more similar to pre-trained joint embedding models.

A Cookbook of Self-Supervised Learning

Self-supervised learning, dubbed the dark matter of intelligence, is a promising path to advance machine learning. Yet, much like cooking, training SSL methods is a delicate art with a high barrier to entry. While many components are familiar, successfully training a SSL method involves a dizzying set of choices from the pretext tasks to training hyper-parameters. Our goal is to lower the barrier to entry into SSL research by laying the foundations and latest SSL recipes in the style of a cookbook. We hope to empower the curious researcher to navigate the terrain of methods, understand the role of the various knobs, and gain the know-how required to explore how delicious SSL can be.

UNet Based Pipeline for Lung Segmentation from Chest X-Ray Images

Biomedical image segmentation is one of the fastest growing fields which has seen extensive automation through the use of Artificial Intelligence. This has enabled widespread adoption of accurate techniques to expedite the screening and diagnostic processes which would otherwise take several days to finalize. In this paper, we present an end-to-end pipeline to segment lungs from chest X-ray images, training the neural network model on the Japanese Society of Radiological Technology (JSRT) dataset, using UNet to enable faster processing of initial screening for various lung disorders. The pipeline developed can be readily used by medical centers with just the provision of X-Ray images as input. The model will perform the preprocessing, and provide a segmented image as the final output. It is expected that this will drastically reduce the manual effort involved and lead to greater accessibility in resource-constrained locations.

Robust Planning for Human-Robot Joint Tasks with Explicit Reasoning on Human Mental State

We consider the human-aware task planning problem where a human-robot team is given a shared task with a known objective to achieve. Recent approaches tackle it by modeling it as a team of independent, rational agents, where the robot plans for both agents' (shared) tasks. However, the robot knows that humans cannot be administered like artificial agents, so it emulates and predicts the human's decisions, actions, and reactions. Based on earlier approaches, we describe a novel approach to solve such problems, which models and uses execution-time observability conventions. Abstractly, this modeling is based on situation assessment, which helps our approach capture the evolution of individual agents' beliefs and anticipate belief divergences that arise in practice. It decides if and when belief alignment is needed and achieves it with communication. These changes improve the solver's performance: (a) communication is effectively used, and (b) robust for more realistic and challenging problems.

Beyond neural scaling laws: beating power law scaling via data pruning

Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how both in theory and practice we can break beyond power law scaling and reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this new exponential scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling performance on ResNets trained on CIFAR-10, SVHN, and ImageNet. Given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning.

An Ensemble Model for Face Liveness Detection

In this paper, we present a passive method to detect face presentation attack a.k.a face liveness detection using an ensemble deep learning technique. Face liveness detection is one of the key steps involved in user identity verification of customers during the online onboarding/transaction processes. During identity verification, an unauthenticated user tries to bypass the verification system by several means, for example, they can capture a user photo from social media and do an imposter attack using printouts of users faces or using a digital photo from a mobile device and even create a more sophisticated attack like video replay attack. We have tried to understand the different methods of attack and created an in-house large-scale dataset covering all the kinds of attacks to train a robust deep learning model. We propose an ensemble method where multiple features of the face and background regions are learned to predict whether the user is a bonafide or an attacker.

A Comparative study of Hyper-Parameter Optimization Tools

Most of the machine learning models have associated hyper-parameters along with their parameters. While the algorithm gives the solution for parameters, its utility for model performance is highly dependent on the choice of hyperparameters. For a robust performance of a model, it is necessary to find out the right hyper-parameter combination. Hyper-parameter optimization (HPO) is a systematic process that helps in finding the right values for them. The conventional methods for this purpose are grid search and random search and both methods create issues in industrial-scale applications. Hence a set of strategies have been recently proposed based on Bayesian optimization and evolutionary algorithm principles that help in runtime issues in a production environment and robust performance. In this paper, we compare the performance of four python libraries, namely Optuna, Hyper-opt, Optunity, and sequential model-based algorithm configuration (SMAC) that has been proposed for hyper-parameter optimization. The performance of these tools is tested using two benchmarks. The first one is to solve a combined algorithm selection and hyper-parameter optimization (CASH) problem The second one is the NeurIPS black-box optimization challenge in which a multilayer perception (MLP) architecture has to be chosen from a set of related architecture constraints and hyper-parameters. The benchmarking is done with six real-world datasets. From the experiments, we found that Optuna has better performance for CASH problem and HyperOpt for MLP problem.

Joint Power-control and Antenna Selection in User-Centric Cell-Free Systems with Mixed Resolution ADC

In this paper, we propose a scheme for the joint optimization of the user transmit power and the antenna selection at the access points (AP)s of a user-centric cell-free massive multiple-input-multiple-output (UC CF-mMIMO) system. We derive an approximate expression for the achievable uplink rate of the users in a UC CF-mMIMO system in the presence of a mixed analog-to-digital converter (ADC) resolution profile at the APs. Using the derived approximation, we propose to maximize the uplink sum rate of UC CF-mMIMO systems subject to energy constraints at the APs. An alternating-optimization solution is proposed using binary particle swarm optimization (BPSO) and successive convex approximation (SCA). We also study the impact of various system parameters on the performance of the system.