Silas: High Performance, Explainable and V erifiable Machine Learning Hadrien Bride, Zh e H ou Griffith University, Nathan, Brisbane, Australia Jie Dong Dependable Intelligence Pty Ltd, Brisbane, Australia Jin Song Dong National University of Singapore, Singapore Ali Mirjalili Griffith University, Nathan, Brisbane, AustraliaAbstract This paper introduces a new classification tool named Silas, which is built to provide a more transparent and dependable data analytics service. A focus of Silas is on providing a formal foundation of decision trees in order to support logical analysis and verification of learned prediction models. This paper describes the distinct features of Silas: The Model Audit module formally verifies the prediction model against user specifications, the Enforcement Learning module trains prediction models that are guaranteed correct, the Model Insight and Prediction Insight modules reason about the prediction model and explain the decision-making of predictions. We also discuss implementation details ranging from programming paradigm to memory management that help achieve high-performance computation.1. Introduction Machine learning has enjoyed great success in many research areas and industries, including entertainment [1], self-driving cars [2], banking [3], medical diagnosis [4], shopping [5], and among many others. However, the wide adoption of machine learn-Preprint submitted to Elsevier October 4, 2019 arXiv:1910.01382v1 The ramifications of the black-box approach are multifold. First, it may lead to unexpected results that are only observable after the deployment of the algorithm. For instance, Amazon's Alexa offered porn to a child [6], a self-driving car had a deadly accident [7], etc. Some of these accidents result in lawsuits or even lost lives, the cost of which is immeasurable. Second, it prevents the adoption in some applications and industries where an explanation is mandatory or certain specifications must be satisfied. For example, in some countries, it is required by law to give a reason why a loan application is rejected. In recent years, eXplainable AI (XAI) has been gaining attention, and there is a surge of interest in studying how prediction models work and how to provide formal guarantees for the models. A common theme in this space is to use statistical methods to analyse prediction models.

Deep neural network image classifiers are reported to be susceptible to adversarial evasion attacks, which use carefully crafted images created to mislead a classifier. Recently, various kinds of adversarial attack methods have been proposed, most of which focus on adding small perturbations to input images. Despite the success of existing approaches, the way to generate realistic adversarial images with small perturbations remains a challenging problem. In this paper, we aim to address this problem by proposing a novel adversarial method, which generates adversarial examples by imposing not only perturbations but also spatial distortions on input images, including scaling, rotation, shear, and translation. As humans are less susceptible to small spatial distortions, the proposed approach can produce visually more realistic attacks with smaller perturbations, able to deceive classifiers without affecting human predictions. We learn our method by amortized techniques with neural networks and generate adversarial examples efficiently by a forward pass of the networks. Extensive experiments on attacking different types of non-robustified classifiers and robust classifiers with defence show that our method has state-of-the-art performance in comparison with advanced attack parallels.

Recent algorithms with state-of-the-art few-shot classification results start their procedure by computing data features output by a large pretrained model. In this paper we systematically investigate which models provide the best representations for a few-shot image classification task when pretrained on the Imagenet dataset. We test their representations when used as the starting point for different few-shot classification algorithms. We observe that models trained on a supervised classification task have higher performance than models trained in an unsupervised manner even when transferred to out-of-distribution datasets. Models trained with adversarial robustness transfer better, while having slightly lower accuracy than supervised models.

The combination of the re-parameterization trick with the use of variational auto-encoders has caused a sensation in Bayesian deep learning, allowing the training of realistic generative models of images and has considerably increased our ability to use scalable latent variable models. The re-parameterization trick is necessary for models in which no analytical variational bound is available and allows noisy gradients to be computed for arbitrary models. However, for certain standard output layers of a neural network, analytical bounds are available and the variational auto-encoder may be used both without the re-parameterization trick or the need for any Monte Carlo approximation. In this work, we show that using Jaakola and Jordan bound, we can produce a binary classification layer that allows a Bayesian output layer to be trained, using the standard stochastic gradient descent algorithm. We further demonstrate that a latent variable model utilizing the Bouchard bound for multi-class classification allows for fast training of a fully probabilistic latent factor model, even when the number of classes is very large.

A BSTRACT Few-shot learning is the process of learning novel classes using only a few examples and it remains a challenging task in machine learning. Many sophisticated few-shot learning algorithms have been proposed based on the notion that networks can easily overfit to novel examples if they are simply fine-tuned using only a few examples. In this study, we show that in the commonly used low-resolution mini -ImageNet dataset, the fine-tuning method achieves higher accuracy than common few-shot learning algorithms in the 1-shot task and nearly the same accuracy as that of the state-of-the-art algorithm in the 5-shot task. We then evaluate our method with more practical tasks, namely the high-resolution single-domain and cross-domain tasks. With both tasks, we show that our method achieves higher accuracy than common few-shot learning algorithms. We further analyze the experimental results and show that: 1) the retraining process can be stabilized by employing a low learning rate, 2) using adaptive gradient optimizers during fine-tuning can increase test accuracy, and 3) test accuracy can be improved by updating the entire network when a large domain-shift exists between base and novel classes. 1 I NTRODUCTION Previous studies have shown that high image classification performance can be achieved by using deep networks and big datasets (Krizhevsky et al., 2012; Simonyan & Zisserman, 2015; He et al., 2016; Szegedy et al., 2015). However, the performances of these algorithms rely heavily on extensive manually annotated images, and considerable cost is often incurred in preparing these datasets. To avoid this problem, few-shot learning, which is a task of learning novel classes using only a few examples, has been actively researched.

Network modeling is a key enabler to achieve efficient network operation in future self-driving Software-Defined Networks. However, we still lack functional network models able to produce accurate predictions of Key Performance Indicators (KPI) such as delay, jitter or loss at limited cost. In this paper we propose RouteNet, a novel network model based on Graph Neural Network (GNN) that is able to understand the complex relationship between topology, routing and input traffic to produce accurate estimates of the per-source/destination per-packet delay distribution and loss. RouteNet leverages the ability of GNNs to learn and model graph-structured information and as a result, our model is able to generalize over arbitrary topologies, routing schemes and traffic intensity. In our evaluation, we show that RouteNet is able to predict accurately the delay distribution (mean delay and jitter) and loss even in topologies, routing and traffic unseen in the training (worst case $R^{2}$ = 0.878). Also, we present several use-cases where we leverage the KPI predictions of our GNN model to achieve efficient routing optimization and network planning.

Resilient Coverage: Exploring the Local-to-Global Tradeoff Ragesh K. Ramachandran 1, Lifeng Zhou 2 and Gaurav S. Sukhatme 1 Abstract -- We propose a centralized control framework to select suitable robots from a heterogeneous pool and place them at appropriate locations to monitor a region for events of interest. In the event of a robot failure, the framework repositions robots in a user-defined local neighborhood of the failed robot to compensate for the coverage loss. The central controller augments the team with additional robots from the robot pool when simply repositioning robots fails to attain a user-specified level of desired coverage. The size of the local neighborhood around the failed robot and the desired coverage over the region are two settings that can be manipulated to achieve a user-specified balance. We investigate the tradeoff between the coverage compensation achieved through local repositioning and the computation required to plan the new robot locations. We also study the relationship between the size of the local neighborhood and the number of additional robots added to the team for a given user-specified level of desired coverage. The computational complexity of our resilient strategy (tunable resilient coordination), is quadratic in both neighborhood size and number of robots in the team. At first glance, it seems that any desired level of coverage can be efficiently achieved by augmenting the robot team with more robots while keeping the neighborhood size fixed. However, we show that to reach a high level of coverage in a neighborhood with a large robot population, it is more efficient to enlarge the neighborhood size, instead of adding additional robots and repositioning them.

Causal reasoning has been an indispensable capability for humans and other intelligent animals to interact with the physical world. In this work, we propose to endow an artificial agent with the capability of causal reasoning for completing goal-directed tasks. We develop learning-based approaches to inducing causal knowledge in the form of directed acyclic graphs, which can be used to contextualize a learned goal-conditional policy to perform tasks in novel environments with latent causal structures. We leverage attention mechanisms in our causal induction model and goal-conditional policy, enabling us to incrementally generate the causal graph from the agent's visual observations and to selectively use the induced graph for determining actions. Our experiments show that our method effectively generalizes towards completing new tasks in novel environments with previously unseen causal structures.

A BSTRACT In the multi-objective reinforcement learning (MORL) paradigm, the relative importance of each environment objective is often unknown prior to training, so agents must learn to specialize their behavior to optimize different combinations of environment objectives that are specified post-training. These are typically linear combinations, so the agent is effectively parameterized by a weight vector that describes how to balance competing environment objectives. However, many real world behaviors require nonlinear combinations of objectives. Additionally, the conversion between desired behavior and weightings is often unclear. In this work, we explore the use of a language based on propositional logic with quantitative semantics-in place of weight vectors-for specifying nonlinear behaviors in an interpretable way. We use a recurrent encoder to encode logical combinations of objectives, and train a MORL agent to generalize over these encodings. We test our agent in several grid worlds with various objectives and show that our agent can generalize to many never-before-seen specifications with performance comparable to single policy baseline agents. We also demonstrate our agent's ability to generate meaningful policies when presented with novel specifications and quickly specialize to novel specifications. 1 I NTRODUCTION Reinforcement Learning (RL) is a method for learning behavior policies by maximizing expected reward through interactions with an environment. RL has grown in popularity as RL agents have excelled at increasingly complex tasks, including board games (Silver et al., 2016), video games (Mnih et al., 2015), robotic control (Haarnoja et al., 2018), and other high dimensional, complex tasks.

Widely-used deep reinforcement learning algorithms have been shown to fail in the batch setting--learning from a fixed data set without interaction with the environment. Following this result, there have been several papers showing reasonable performances under a variety of environments and batch settings. In this paper, we benchmark the performance of recent off-policy and batch reinforcement learning algorithms under unified settings on the Atari domain, with data generated by a single partially-trained behavioral policy. We find that under these conditions, many of these algorithms underperform DQN trained online with the same amount of data, as well as the partially-trained behavioral policy. To introduce a strong baseline, we adapt the Batch-Constrained Q-learning algorithm to a discrete-action setting, and show it outperforms all existing algorithms at this task.