There are a large number of optimization problems in physical models where the relationships between model parameters and outputs are unknown or hard to track. These models are named as black-box models in general because they can only be viewed in terms of inputs and outputs, without knowledge of the internal workings. Optimizing the black-box model parameters has become increasingly expensive and time consuming as they have become more complex. Hence, developing effective and efficient black-box model optimization algorithms has become an important task. One powerful algorithm to solve such problem is Bayesian optimization, which can effectively estimates the model parameters that lead to the best performance, and Gaussian Process (GP) has been one of the most widely used surrogate model in Bayesian optimization. However, the time complexity of GP scales cubically with respect to the number of observed model outputs, and GP does not scale well with large parameter dimension either. Consequently, it has been challenging for GP to optimize black-box models that need to query many observations and/or have many parameters. To overcome the drawbacks of GP, in this study, we propose a general Bayesian optimization algorithm that employs a Neural Process (NP) as the surrogate model to perform black-box model optimization, namely, Neural Process for Bayesian Optimization (NPBO). In order to validate the benefits of NPBO, we compare NPBO with four benchmark approaches on a power system parameter optimization problem and a series of seven benchmark Bayesian optimization problems. The results show that the proposed NPBO performs better than the other four benchmark approaches on the power system parameter optimization problem and competitively on the seven benchmark problems.
Temporal action localization (TAL) is a fundamental yet challenging task in video understanding. Existing TAL methods rely on pre-training a video encoder through action classification supervision. This results in a task discrepancy problem for the video encoder -- trained for action classification, but used for TAL. Intuitively, end-to-end model optimization is a good solution. However, this is not operable for TAL subject to the GPU memory constraints, due to the prohibitive computational cost in processing long untrimmed videos. In this paper, we resolve this challenge by introducing a novel low-fidelity end-to-end (LoFi) video encoder pre-training method. Instead of always using the full training configurations for TAL learning, we propose to reduce the mini-batch composition in terms of temporal, spatial or spatio-temporal resolution so that end-to-end optimization for the video encoder becomes operable under the memory conditions of a mid-range hardware budget. Crucially, this enables the gradient to flow backward through the video encoder from a TAL loss supervision, favourably solving the task discrepancy problem and providing more effective feature representations. Extensive experiments show that the proposed LoFi pre-training approach can significantly enhance the performance of existing TAL methods. Encouragingly, even with a lightweight ResNet18 based video encoder in a single RGB stream, our method surpasses two-stream ResNet50 based alternatives with expensive optical flow, often by a good margin.
Model complexity is a fundamental problem in deep learning. In this paper we conduct a systematic overview of the latest studies on model complexity in deep learning. Model complexity of deep learning can be categorized into expressive capacity and effective model complexity. We review the existing studies on those two categories along four important factors, including model framework, model size, optimization process and data complexity. We also discuss the applications of deep learning model complexity including understanding model generalization, model optimization, and model selection and design. We conclude by proposing several interesting future directions.
Epistemic Uncertainty is a measure of the lack of knowledge of a learner which diminishes with more evidence. While existing work focuses on using the variance of the Bayesian posterior due to parameter uncertainty as a measure of epistemic uncertainty, we argue that this does not capture the part of lack of knowledge induced by model misspecification. We discuss how the excess risk, which is the gap between the generalization error of a predictor and the Bayes predictor, is a sound measure of epistemic uncertainty which captures the effect of model misspecification. We thus propose a principled framework for directly estimating the excess risk by learning a secondary predictor for the generalization error and subtracting an estimate of aleatoric uncertainty, i.e., intrinsic unpredictability. We discuss the merits of this novel measure of epistemic uncertainty, and highlight how it differs from variance-based measures of epistemic uncertainty and addresses its major pitfall. Our framework, Direct Epistemic Uncertainty Prediction (DEUP) is particularly interesting in interactive learning environments, where the learner is allowed to acquire novel examples in each round. Through a wide set of experiments, we illustrate how existing methods in sequential model optimization can be improved with epistemic uncertainty estimates from DEUP, and how DEUP can be used to drive exploration in reinforcement learning. We also evaluate the quality of uncertainty estimates from DEUP for probabilistic image classification and predicting synergies of drug combinations.
The recent success of graph neural networks has significantly boosted molecular property prediction, advancing activities such as drug discovery. The existing deep neural network methods usually require large training dataset for each property, impairing their performances in cases (especially for new molecular properties) with a limited amount of experimental data, which are common in real situations. To this end, we propose Meta-MGNN, a novel model for few-shot molecular property prediction. Meta-MGNN applies molecular graph neural network to learn molecular representation and builds a meta-learning framework for model optimization. To exploit unlabeled molecular information and address task heterogeneity of different molecular properties, Meta-MGNN further incorporates molecular structure, attribute based self-supervised modules and self-attentive task weights into the former framework, strengthening the whole learning model. Extensive experiments on two public multi-property datasets demonstrate that Meta-MGNN outperforms a variety of state-of-the-art methods.
Deep Learning (DL) combined with advanced model optimization methods such as RC-NN and Edge2Train has enabled offline execution of large networks on the IoT devices. In this paper, we compare the modern Deep Learning (DL) based video analytics approaches with the standard Computer Vision (CV) based approaches and finally, discuss the best-suited approach for video analytics on IoT devices.
Contrastive self-supervised learning (CSL) has attracted increasing attention for model pre-training via unlabeled data. The resulted CSL models provide instance-discriminative visual features that are uniformly scattered in the feature space. During deployment, the common practice is to directly fine-tune CSL models with cross-entropy, which however may not be the best strategy in practice. Although cross-entropy tends to separate inter-class features, the resulting models still have limited capability for reducing intra-class feature scattering that exists in CSL models. In this paper, we investigate whether applying contrastive learning to fine-tuning would bring further benefits, and analytically find that optimizing the contrastive loss benefits both discriminative representation learning and model optimization during fine-tuning. Inspired by these findings, we propose Contrast-regularized tuning (Core-tuning), a new approach for fine-tuning CSL models. Instead of simply adding the contrastive loss to the objective of fine-tuning, Core-tuning further applies a novel hard pair mining strategy for more effective contrastive fine-tuning, as well as smoothing the decision boundary to better exploit the learned discriminative feature space. Extensive experiments on image classification and semantic segmentation verify the effectiveness of Core-tuning.
Designing proper loss functions for vision tasks has been a long-standing research direction to advance the capability of existing models. For object detection, the well-established classification and regression loss functions have been carefully designed by considering diverse learning challenges. Inspired by the recent progress in network architecture search, it is interesting to explore the possibility of discovering new loss function formulations via directly searching the primitive operation combinations. So that the learned losses not only fit for diverse object detection challenges to alleviate huge human efforts, but also have better alignment with evaluation metric and good mathematical convergence property. Beyond the previous auto-loss works on face recognition and image classification, our work makes the first attempt to discover new loss functions for the challenging object detection from primitive operation levels. We propose an effective convergence-simulation driven evolutionary search algorithm, called CSE-Autoloss, for speeding up the search progress by regularizing the mathematical rationality of loss candidates via convergence property verification and model optimization simulation. CSE-Autoloss involves the search space that cover a wide range of the possible variants of existing losses and discovers best-searched loss function combination within a short time (around 1.5 wall-clock days). We conduct extensive evaluations of loss function search on popular detectors and validate the good generalization capability of searched losses across diverse architectures and datasets. Our experiments show that the best-discovered loss function combinations outperform default combinations by 1.1% and 0.8% in terms of mAP for two-stage and one-stage detectors on COCO respectively. Our searched losses are available at https://github.com/PerdonLiu/CSE-Autoloss.
In recent years, recommender systems play a pivotal role in helping users identify the most suitable items that satisfy personal preferences. As user-item interactions can be naturally modelled as graph-structured data, variants of graph convolutional networks (GCNs) have become a well-established building block in the latest recommenders. Due to the wide utilization of sensitive user profile data, existing recommendation paradigms are likely to expose users to the threat of privacy breach, and GCN-based recommenders are no exception. Apart from the leakage of raw user data, the fragility of current recommenders under inference attacks offers malicious attackers a backdoor to estimate users' private attributes via their behavioral footprints and the recommendation results. However, little attention has been paid to developing recommender systems that can defend such attribute inference attacks, and existing works achieve attack resistance by either sacrificing considerable recommendation accuracy or only covering specific attack models or protected information. In our paper, we propose GERAI, a novel differentially private graph convolutional network to address such limitations. Specifically, in GERAI, we bind the information perturbation mechanism in differential privacy with the recommendation capability of graph convolutional networks. Furthermore, based on local differential privacy and functional mechanism, we innovatively devise a dual-stage encryption paradigm to simultaneously enforce privacy guarantee on users' sensitive features and the model optimization process. Extensive experiments show the superiority of GERAI in terms of its resistance to attribute inference attacks and recommendation effectiveness.
We developed a model and a software package for stochastic simulations of transmission of COVID-19 and other similar infectious diseases, that takes into account contact network structures and geographical distribution of population density, detailed up to a level of location of individuals. Our analysis framework includes a surrogate model optimization process for quick fitting of the model's parameters to the observed epidemic curves for cases, hospitalizations and deaths. This set of instruments (the model, the simulation code, and the optimizer) is a useful tool for policymakers and epidemic response teams who can use it to forecast epidemic development scenarios in local environments (on the scale from towns to large countries) and design optimal response strategies. The simulation code also includes a geospatial visualization subsystem, presenting detailed views of epidemic scenarios directly on population density maps. We used the developed framework to draw predictions for COVID-19 spreading in the canton of Geneva, Switzerland.
Designing deep learning-based solutions is becoming a race for training deeper models with a greater number of layers. While a large-size deeper model could provide competitive accuracy, it creates a lot of logistical challenges and unreasonable resource requirements during development and deployment. This has been one of the key reasons for deep learning models not being excessively used in various production environments, especially in edge devices. There is an immediate requirement for optimizing and compressing these deep learning models, to enable on-device intelligence. In this research, we introduce a black-box framework, Deeplite Neutrino for production-ready optimization of deep learning models. The framework provides an easy mechanism for the end-users to provide constraints such as a tolerable drop in accuracy or target size of the optimized models, to guide the whole optimization process. The framework is easy to include in an existing production pipeline and is available as a Python Package, supporting PyTorch and Tensorflow libraries. The optimization performance of the framework is shown across multiple benchmark datasets and popular deep learning models. Further, the framework is currently used in production and the results and testimonials from several clients are summarized.
In recent years, deep learning-based automated personality trait detection has received a lot of attention, especially now, due to the massive digital footprints of an individual. Moreover, many researchers have demonstrated that there is a strong link between personality traits and emotions. In this paper, we build on the known correlation between personality traits and emotional behaviors, and propose a novel multitask learning framework, SoGMTL that simultaneously predicts both of them. We also empirically evaluate and discuss different information-sharing mechanisms between the two tasks. To ensure the high quality of the learning process, we adopt a MAML-like framework for model optimization. Our more computationally efficient CNN-based multitask model achieves the state-of-the-art performance across multiple famous personality and emotion datasets, even outperforming Language Model based models.
Machine learning, and eventually true artificial intelligence techniques, are extremely important advancements in astrophysics and astronomy. We explore the application of deep learning using neural networks in order to automate the detection of astronomical bodies for future exploration missions, such as missions to search for signatures or suitability of life. The ability to acquire images, analyze them, and send back those that are important, as determined by the deep learning algorithm, is critical in bandwidth-limited applications. Our previous foundational work solidified the concept of using simulator images and deep learning in order to detect planets. Optimization of this process is of vital importance, as even a small loss in accuracy might be the difference between capturing and completely missing a possibly-habitable nearby planet. Through computer vision, deep learning, and simulators, we introduce methods that optimize the detection of exoplanets. We show that maximum achieved accuracy can hit above 98% for multiple model architectures, even with a relatively small training set.
Quantitative assessment of cardiac left ventricle (LV) morphology is essential to assess cardiac function and improve the diagnosis of different cardiovascular diseases. In current clinical practice, LV quantification depends on the measurement of myocardial shape indices, which is usually achieved by manual contouring of the endo- and epicardial. However, this process subjected to inter and intra-observer variability, and it is a time-consuming and tedious task. In this paper, we propose a spatio-temporal multi-task learning approach to obtain a complete set of measurements quantifying cardiac LV morphology, regional-wall thickness (RWT), and additionally detecting the cardiac phase cycle (systole and diastole) for a given 3D Cine-magnetic resonance (MR) image sequence. We first segment cardiac LVs using an encoder-decoder network and then introduce a multitask framework to regress 11 LV indices and classify the cardiac phase, as parallel tasks during model optimization. The proposed deep learning model is based on the 3D spatio-temporal convolutions, which extract spatial and temporal features from MR images. We demonstrate the efficacy of the proposed method using cine-MR sequences of 145 subjects and comparing the performance with other state-of-the-art quantification methods. The proposed method obtained high prediction accuracy, with an average mean absolute error (MAE) of 129 $mm^2$, 1.23 $mm$, 1.76 $mm$, Pearson correlation coefficient (PCC) of 96.4%, 87.2%, and 97.5% for LV and myocardium (Myo) cavity regions, 6 RWTs, 3 LV dimensions, and an error rate of 9.0\% for phase classification. The experimental results highlight the robustness of the proposed method, despite varying degrees of cardiac morphology, image appearance, and low contrast in the cardiac MR sequences.
While federated learning traditionally aims to train a single global model across decentralized local datasets, one model may not always be ideal for all participating clients. Here we propose an alternative, where each client only federates with other relevant clients to obtain a stronger model per client-specific objectives. To achieve this personalization, rather than computing a single model average with constant weights for the entire federation as in traditional FL, we efficiently calculate optimal weighted model combinations for each client, based on figuring out how much a client can benefit from another's model. We do not assume knowledge of any underlying data distributions or client similarities, and allow each client to optimize for arbitrary target distributions of interest, enabling greater flexibility for personalization. We evaluate and characterize our method on a variety of federated settings, datasets, and degrees of local data heterogeneity. Our method outperforms existing alternatives, while also enabling new features for personalized FL such as transfer outside of local data distributions.
Despite major advances in quantitative approaches to natural resource management, there has been resistance to using these tools in the actual practice of managing ecological populations. Given a managed system and a set of assumptions, translated into a model, optimization methods can be used to solve for the most cost effective management actions. However, when the underlying assumptions are not met, such methods can potentially lead to poor decisions. Managers who develop decisions based on past experience and judgment, without the aid of mathematical models, can potentially learn about the system and develop flexible management strategies. However, these strategies are often based on subjective criteria and equally invalid and often unstated assumptions. Given the drawbacks of both methods, it is unclear whether simple quantitative models improve environmental decision making over expert opinion. In this paper, we explore how well students, using their experience and judgment, manage simulated fishery populations in an online computer game and compare their management outcomes to the performance of model-based decisions. We consider harvest decisions generated using four different quantitative models: 1. the model used to produce the simulated population dynamics observed in the game, with all underlying parameter values known [a control], 2. the same model, but with unknown parameter values that must be estimated during the game from observed data, 3. models that are structurally different from those used to simulate the population dynamics and 4. a model that ignores age structure. Humans on average performed much worse than the models in cases 1 - 3. When the models ignored age structure, they generated poorly performing management decisions, but still outperformed students using experience and judgment 66 percent of the time.
Forecasting is challenging since uncertainty resulted from exogenous factors exists. This work investigates the rank position forecasting problem in car racing, which predicts the rank positions at the future laps for cars. Among the many factors that bring changes to the rank positions, pit stops are critical but irregular and rare. We found existing methods, including statistical models, machine learning regression models, and state-of-the-art deep forecasting model based on encoder-decoder architecture, all have limitations in the forecasting. By elaborative analysis of pit stops events, we propose a deep model, RankNet, with the cause effects decomposition that modeling the rank position sequence and pit stop events separately. It also incorporates probabilistic forecasting to model the uncertainty inside each sub-model. Through extensive experiments, RankNet demonstrates a strong performance improvement over the baselines, e.g., MAE improves more than 10% consistently, and is also more stable when adapting to unseen new data. Details of model optimization, performance profiling are presented. It is promising to provide useful forecasting tools for the car racing analysis and shine a light on solutions to similar challenging issues in general forecasting problems.
Large datasets have become commonplace in NLP research. However, the increased emphasis on data quantity has made it challenging to assess the quality of data. We introduce Data Maps---a model-based tool to characterize and diagnose datasets. We leverage a largely ignored source of information: the behavior of the model on individual instances during training (training dynamics) for building data maps. This yields two intuitive measures for each example---the model's confidence in the true class, and the variability of this confidence across epochs---obtained in a single run of training. Experiments across four datasets show that these model-dependent measures reveal three distinct regions in the data map, each with pronounced characteristics. First, our data maps show the presence of "ambiguous" regions with respect to the model, which contribute the most towards out-of-distribution generalization. Second, the most populous regions in the data are "easy to learn" for the model, and play an important role in model optimization. Finally, data maps uncover a region with instances that the model finds "hard to learn"; these often correspond to labeling errors. Our results indicate that a shift in focus from quantity to quality of data could lead to robust models and improved out-of-distribution generalization.
We present a novel active learning algorithm, termed as iterative surrogate model optimization (ISMO), for robust and efficient numerical approximation of PDE constrained optimization problems. This algorithm is based on deep neural networks and its key feature is the iterative selection of training data through a feedback loop between deep neural networks and any underlying standard optimization algorithm. Under suitable hypotheses, we show that the resulting optimizers converge exponentially fast (and with exponentially decaying variance), with respect to increasing number of training samples. Numerical examples for optimal control, parameter identification and shape optimization problems for PDEs are provided to validate the proposed theory and to illustrate that ISMO significantly outperforms a standard deep neural network based surrogate optimization algorithm.
Attention and self-attention mechanisms, are now central to state-of-the-art deep learning on sequential tasks. However, most recent progress hinges on heuristic approaches with limited understanding of attention's role in model optimization and computation, and rely on considerable memory and computational resources that scale poorly. In this work, we present a formal analysis of how self-attention affects gradient propagation in recurrent networks, and prove that it mitigates the problem of vanishing gradients when trying to capture long-term dependencies by establishing concrete bounds for gradient norms. Building on these results, we propose a relevancy screening mechanism, inspired by the cognitive process of memory consolidation, that allows for a scalable use of sparse self-attention with recurrence. While providing guarantees to avoid vanishing gradients, we use simple numerical experiments to demonstrate the tradeoffs in performance and computational resources by efficiently balancing attention and recurrence. Based on our results, we propose a concrete direction of research to improve scalability of attentive networks.