Predictive uncertainty estimation is an essential next step for the reliable deployment of deep object detectors in safety-critical tasks. In this work, we focus on estimating predictive distributions for bounding box regression output with variance networks. We show that in the context of object detection, training variance networks with negative log likelihood (NLL) can lead to high entropy predictive distributions regardless of the correctness of the output mean. We propose to use the energy score as a non-local proper scoring rule and find that when used for training, the energy score leads to better calibrated and lower entropy predictive distributions than NLL. We also address the widespread use of non-proper scoring metrics for evaluating predictive distributions from deep object detectors by proposing an alternate evaluation approach founded on proper scoring rules. Using the proposed evaluation tools, we show that although variance networks can be used to produce high quality predictive distributions, adhoc approaches used by seminal object detectors for choosing regression targets during training do not provide wide enough data support for reliable variance learning. We hope that our work helps shift evaluation in probabilistic object detection to better align with predictive uncertainty evaluation in other machine learning domains. Deep object detectors are being increasingly deployed as perception components in safety critical robotics and automation applications. For reliable and safe operation, subsequent tasks using detectors as sensors require meaningful predictive uncertainty estimates correlated with their outputs. As an example, overconfident incorrect predictions can lead to non-optimal decision making in planning tasks, while underconfident correct predictions can lead to under-utilizing information in sensor fusion. This paper investigates probabilistic object detectors, extensions of standard object detectors that estimate predictive distributions for output categories and bounding boxes simultaneously. This paper aims to identify the shortcomings of recent trends followed by state-of-the-art probabilistic object detectors, and proposes to provide theoretically founded solutions for identified issues.

The information describing the conditions of a system or a person is constantly evolving and may become obsolete and contradict other information. A database, therefore, must be consistently updated upon the acquisition of new valid observations that contradict obsolete ones contained in the database. In this paper, we propose a novel approach for dealing with the information obsolescence problem. Our approach aims to detect, in real-time, contradictions between observations and then identify the obsolete ones, given a representation model. Since we work within an uncertain environment characterized by the lack of information, we choose to use a Bayesian network as our representation model and propose a new approximate concept, $\epsilon$-Contradiction. The new concept is parameterised by a confidence level of having a contradiction in a set of observations. We propose a polynomial-time algorithm for detecting obsolete information. We show that the resulting obsolete information is better represented by an AND-OR tree than a simple set of observations. Finally, we demonstrate the effectiveness of our approach on a real elderly fall-prevention database and showcase how this tree can be used to give reliable recommendations to doctors. Our experiments give systematically and substantially very good results.

Benefiting from convenient cycling and flexible parking locations, the Dockless Public Bicycle-sharing (DL-PBS) network becomes increasingly popular in many countries. However, redundant and low-utility stations waste public urban space and maintenance costs of DL-PBS vendors. In this paper, we propose a Bicycle Station Dynamic Planning (BSDP) system to dynamically provide the optimal bicycle station layout for the DL-PBS network. The BSDP system contains four modules: bicycle drop-off location clustering, bicycle-station graph modeling, bicycle-station location prediction, and bicycle-station layout recommendation. In the bicycle drop-off location clustering module, candidate bicycle stations are clustered from each spatio-temporal subset of the large-scale cycling trajectory records. In the bicycle-station graph modeling module, a weighted digraph model is built based on the clustering results and inferior stations with low station revenue and utility are filtered. Then, graph models across time periods are combined to create a graph sequence model. In the bicycle-station location prediction module, the GGNN model is used to train the graph sequence data and dynamically predict bicycle stations in the next period. In the bicycle-station layout recommendation module, the predicted bicycle stations are fine-tuned according to the government urban management plan, which ensures that the recommended station layout is conducive to city management, vendor revenue, and user convenience. Experiments on actual DL-PBS networks verify the effectiveness, accuracy and feasibility of the proposed BSDP system.

In this paper, we consider the problem of leveraging textual descriptions to improve generalization of control policies to new scenarios. Unlike prior work in this space, we do not assume access to any form of prior knowledge connecting text and state observations, and learn both symbol grounding and control policy simultaneously. This is challenging due to a lack of concrete supervision, and incorrect groundings can result in worse performance than policies that do not use the text at all. We develop a new model, EMMA (Entity Mapper with Multi-modal Attention) which uses a multi-modal entity-conditioned attention module that allows for selective focus over relevant sentences in the manual for each entity in the environment. EMMA is end-to-end differentiable and can learn a latent grounding of entities and dynamics from text to observations using environment rewards as the only source of supervision. To empirically test our model, we design a new framework of 1320 games and collect text manuals with free-form natural language via crowd-sourcing. We demonstrate that EMMA achieves successful zero-shot generalization to unseen games with new dynamics, obtaining significantly higher rewards compared to multiple baselines. The grounding acquired by EMMA is also robust to noisy descriptions and linguistic variation.

With the advent of deep learning, the number of works proposing new methods or improving existent ones has grown exponentially in the last years. In this scenario, "very deep" models were emerging, once they were expected to extract more intrinsic and abstract features while supporting a better performance. However, such models suffer from the gradient vanishing problem, i.e., backpropagation values become too close to zero in their shallower layers, ultimately causing learning to stagnate. Such an issue was overcome in the context of convolution neural networks by creating "shortcut connections" between layers, in a so-called deep residual learning framework. Nonetheless, a very popular deep learning technique called Deep Belief Network still suffers from gradient vanishing when dealing with discriminative tasks. Therefore, this paper proposes the Residual Deep Belief Network, which considers the information reinforcement layer-by-layer to improve the feature extraction and knowledge retaining, that support better discriminative performance. Experiments conducted over three public datasets demonstrate its robustness concerning the task of binary image classification.

The multinomial probit Bayesian additive regression trees (MPBART) framework was proposed by Kindo et al. (KD), approximating the latent utilities in the multinomial probit (MNP) model with BART (Chipman et al. 2010). Compared to multinomial logistic models, MNP does not assume independent alternatives and the correlation structure among alternatives can be specified through multivariate Gaussian distributed latent utilities. We introduce two new algorithms for fitting the MPBART and show that the theoretical mixing rates of our proposals are equal or superior to the existing algorithm in KD. Through simulations, we explore the robustness of the methods to the choice of reference level, imbalance in outcome frequencies, and the specifications of prior hyperparameters for the utility error term. The work is motivated by the application of generating posterior predictive distributions for mortality and engagement in care among HIV-positive patients based on electronic health records (EHRs) from the Academic Model Providing Access to Healthcare (AMPATH) in Kenya. In both the application and simulations, we observe better performance using our proposals as compared to KD in terms of MCMC convergence rate and posterior predictive accuracy.

In recent years, deep neural networks have had great success in machine learning and pattern recognition. Architecture size for a neural network contributes significantly to the success of any neural network. In this study, we optimize the selection process by investigating different search algorithms to find a neural network architecture size that yields the highest accuracy. We apply binary search on a very well-defined binary classification network search space and compare the results to those of linear search. We also propose how to relax some of the assumptions regarding the dataset so that our solution can be generalized to any binary classification problem. We report a 100-fold running time improvement over the naive linear search when we apply the binary search method to our datasets in order to find the best architecture candidate. By finding the optimal architecture size for any binary classification problem quickly, we hope that our research contributes to discovering intelligent algorithms for optimizing architecture size selection in machine learning.

The right to be forgotten has been legislated in many countries but the enforcement in machine learning would cause unbearable costs: companies may need to delete whole models trained from massive resources because of single individual requests. Existing works propose to remove the influence of the requested datums on the learned models via its influence function which is no longer naturally well-defined in Bayesian inference. To address this problem, this paper proposes a {\it Bayesian inference forgetting} (BIF) framework to extend the applicable domain to Bayesian inference. In the BIF framework, we develop forgetting algorithms for variational inference and Markov chain Monte Carlo. We show that our algorithms can provably remove the influence of single datums on the learned models. Theoretical analysis demonstrates that our algorithms have guaranteed generalizability. Experiments of Gaussian mixture models on the synthetic dataset and Bayesian neural networks on the Fashion-MNIST dataset verify the feasibility of our methods. The source code package is available at \url{https://github.com/fshp971/BIF}.

Federated learning brings potential benefits of faster learning, better solutions, and a greater propensity to transfer when heterogeneous data from different parties increases diversity. However, because federated learning tasks tend to be large and complex, and training times non-negligible, it is important for the aggregation algorithm to be robust to non-IID data and corrupted parties. This robustness relies on the ability to identify, and appropriately weight, incompatible parties. Recent work assumes that a \textit{reference dataset} is available through which to perform the identification. We consider settings where no such reference dataset is available; rather, the quality and suitability of the parties needs to be \textit{inferred}. We do so by bringing ideas from crowdsourced predictions and collaborative filtering, where one must infer an unknown ground truth given proposals from participants with unknown quality. We propose novel federated learning aggregation algorithms based on Bayesian inference that adapt to the quality of the parties. Empirically, we show that the algorithms outperform standard and robust aggregation in federated learning on both synthetic and real data.

To perform Bayesian inference for stochastic simulator models for which the likelihood is not accessible, Likelihood-Free Inference (LFI) relies on simulations from the model. Standard LFI methods can be split according to how these simulations are used: to build an explicit Surrogate Likelihood, or to accept/reject parameter values according to a measure of distance from the observations (Approximate Bayesian Computation (ABC)). In both cases, simulations are adaptively tailored to the value of the observation. Here, we generate parameter-simulation pairs from the model independently on the observation, and use them to learn a conditional exponential family likelihood approximation; to parametrize it, we use Neural Networks whose weights are tuned with Score Matching. With our likelihood approximation, we can employ MCMC for doubly intractable distributions to draw samples from the posterior for any number of observations without additional model simulations, with performance competitive to comparable approaches. Further, the sufficient statistics of the exponential family can be used as summaries in ABC, outperforming the state-of-the-art method in five different models with known likelihood. Finally, we apply our method to a challenging model from meteorology.