In this paper, we propose an efficient saliency map generation method, called Group score-weighted Class Activation Mapping (Group-CAM), which adopts the "split-transform-merge" strategy to generate saliency maps. Specifically, for an input image, the class activations are firstly split into groups. In each group, the sub-activations are summed and de-noised as an initial mask. After that, the initial masks are transformed with meaningful perturbations and then applied to preserve sub-pixels of the input (i.e., masked inputs), which are then fed into the network to calculate the confidence scores. Finally, the initial masks are weighted summed to form the final saliency map, where the weights are confidence scores produced by the masked inputs. Group-CAM is efficient yet effective, which only requires dozens of queries to the network while producing target-related saliency maps. As a result, Group-CAM can be served as an effective data augment trick for fine-tuning the networks. We comprehensively evaluate the performance of Group-CAM on common-used benchmarks, including deletion and insertion tests on ImageNet-1k, and pointing game tests on COCO2017. Extensive experimental results demonstrate that Group-CAM achieves better visual performance than the current state-of-the-art explanation approaches. The code is available at https://github.com/wofmanaf/Group-CAM.

The unification of statistical (data-driven) and symbolic (knowledge-driven) methods is widely recognised as one of the key challenges of modern AI. Recent years have seen large number of publications on such hybrid neuro-symbolic AI systems. That rapidly growing literature is highly diverse and mostly empirical, and is lacking a unifying view of the large variety of these hybrid systems. In this paper we analyse a large body of recent literature and we propose a set of modular design patterns for such hybrid, neuro-symbolic systems. We are able to describe the architecture of a very large number of hybrid systems by composing only a small set of elementary patterns as building blocks. The main contributions of this paper are: 1) a taxonomically organised vocabulary to describe both processes and data structures used in hybrid systems; 2) a set of 15+ design patterns for hybrid AI systems, organised in a set of elementary patterns and a set of compositional patterns; 3) an application of these design patterns in two realistic use-cases for hybrid AI systems. Our patterns reveal similarities between systems that were not recognised until now. Finally, our design patterns extend and refine Kautz' earlier attempt at categorising neuro-symbolic architectures.

This week, we will briefly touch upon the topic of artificial intelligence, or AI, discuss what it is, why it is so important to the American empire and national security, and how the ancient Chamorro people of the Marianas can prepare themselves to be ready for future possible job and entrepreneurial opportunities at the intersection of warfare and technology. Part of the answer is found in a report released by the congressionally established National Security Commission on Artificial Intelligence, which outlined the national importance of AI and its application to all facets of American--and by implication, American colonial--society. AI will become more important as Guam continues to move toward technological solutions for future energy, food technology and security opportunities. Now is the time for the governments of Guam and the CNMI to consider creating a Marianas Artificial Intelligence, Security and Emerging Technologies Understanding advisory board to learn and more completely seek to comprehend the nature of AI, how it is currently used and how it presents opportunities and vulnerabilities to every aspect of Pacific island life. A most dangerous aspect of rapidly emerging AI enabled technology and networks is that nation state adversaries such as China and Russia may outpace the United States on this front over the next 10 to 15 years.

Australia could once again have a globally competitive manufacturing sector by using automation driven by artificial intelligence (AI). That's the view of University of Adelaide researchers who are aiming to play a major role in the development of AI which is poised to reshape the global economy, bringing challenges and opportunities. The authors of the latest Economic Issues paper--"The impact of AI on the future of work and workers"--published by the South Australian Centre for Economic Studies (SACES) and the Australian Institute of Machine Learning (AIML), both research centers at the University of Adelaide, maintain that AI "has reached a global tipping point and we need to plan for it." The authors, Professor Anton van den Hengel and Dr. Paul Dalby, Director and Business Development Manager, respectively, of the AIML, and SACES Research Associate, Dr. Andreas Cebulla, describe AI as "the automation of tasks normally requiring human intelligence." "AI has the potential to temper the impact of globalization which has seen industry leaving developed countries seeking lower cost manufacturing options offshore," the authors say.

The banning of video game Disco Elysium from sale in Australia has renewed calls for the Australian government to overhaul the classification system to move away from the "moral panic" associated with video games. On Friday afternoon, the Australian classification board announced Disco Elysium – The Final Cut was refused classification on the grounds the game was found to "depict, express or otherwise deal with matters of sex, drug misuse or addiction, crime, cruelty, violence or revolting or abhorrent phenomena" in a way that offended "against the standards of morality, decency and propriety generally accepted by reasonable adults". It ruled the game should not be classified. The post-war murder mystery role-playing game has won over a dozen industry awards since its release in 2019. The game has been available in Australia for two years through the Steam online games store, but the game's developers, ZA/UM planned to launch the game on consoles this month, meaning before it could be sold in stores in Australia, it had to go to the classification board for review.

This partnership suggests the chain is relying more heavily on AI and machine learning systems to improve its system. "Machine learning models can provide significant value to an organization in several business applications, but without a solid [machine learning] operations pipeline, that value cannot be truly realized," Zack Fragoso, manager, data science and AI at Domino's, said in the release. The chain began testing AI in 2018 with an Alexa-like voice recognition application called DOM. In 2019, Domino's began piloting Dragontail Systems' AI technology at its Australian and New Zealand locations to scan each pizza and ensure they measured up to quality standards. The pilot was expanded after improving quality scores by 15% in its first month.

We study liquid democracy, a collective decision making paradigm that allows voters to transitively delegate their votes, through an algorithmic lens. In our model, there are two alternatives, one correct and one incorrect, and we are interested in the probability that the majority opinion is correct. Our main question is whether there exist delegation mechanisms that are guaranteed to outperform direct voting, in the sense of being always at least as likely, and sometimes more likely, to make a correct decision. Even though we assume that voters can only delegate their votes to better-informed voters, we show that local delegation mechanisms, which only take the local neighborhood of each voter as input (and, arguably, capture the spirit of liquid democracy), cannot provide the foregoing guarantee. By contrast, we design a non-local delegation mechanism that does provably outperform direct voting under mild assumptions about voters.

Speaker recognition refers to audio biometrics that utilizes acoustic characteristics for automatic speaker recognition. These systems have emerged as an essential means of verifying identity in various scenarios, such as smart homes, general business interactions, e-commerce applications, and forensics. However, the mismatch between training and real-world data causes a shift of speaker embedding space and severely degrades the recognition performance. Various complicated neural architectures are presented to address speaker recognition in the wild but neglect the requirements of storage and computation. To address this issue, we propose a neural architecture search-based efficient time-delay neural network (EfficientTDNN) to improve inference efficiency while maintaining recognition accuracy. The proposed EfficientTDNN contains three phases. First, supernet design is to construct a dynamic neural architecture that consists of sequential cells and enables network pruning. Second, progressive training is to optimize randomly sampled subnets that inherit the weights of the supernet. Third, three search methods, including manual grid search, random search, and model predictive evolutionary search, are introduced to find a trade-off between accuracy and efficiency. Results of experiments on the VoxCeleb dataset show EfficientTDNN provides a huge search space including approximately $10^{13}$ subnets and achieves 1.66% EER and 0.156 DCF$_{0.01}$ with 565M MACs. Comprehensive investigation suggests that the trained supernet generalizes cells unseen during training and obtains an acceptable balance between accuracy and efficiency.

Prediction tasks about students have practical significance for both student and college. Making multiple predictions about students is an important part of a smart campus. For instance, predicting whether a student will fail to graduate can alert the student affairs office to take predictive measures to help the student improve his/her academic performance. With the development of information technology in colleges, we can collect digital footprints which encode heterogeneous behaviors continuously. In this paper, we focus on modeling heterogeneous behaviors and making multiple predictions together, since some prediction tasks are related and learning the model for a specific task may have the data sparsity problem. To this end, we propose a variant of LSTM and a soft-attention mechanism. The proposed LSTM is able to learn the student profile-aware representation from heterogeneous behavior sequences. The proposed soft-attention mechanism can dynamically learn different importance degrees of different days for every student. In this way, heterogeneous behaviors can be well modeled. In order to model interactions among multiple prediction tasks, we propose a co-attention mechanism based unit. With the help of the stacked units, we can explicitly control the knowledge transfer among multiple tasks. We design three motivating behavior prediction tasks based on a real-world dataset collected from a college. Qualitative and quantitative experiments on the three prediction tasks have demonstrated the effectiveness of our model.

Text-based games simulate worlds and interact with players using natural language. Recent work has used them as a testbed for autonomous language-understanding agents, with the motivation being that understanding the meanings of words or semantics is a key component of how humans understand, reason, and act in these worlds. However, it remains unclear to what extent artificial agents utilize semantic understanding of the text. To this end, we perform experiments to systematically reduce the amount of semantic information available to a learning agent. Surprisingly, we find that an agent is capable of achieving high scores even in the complete absence of language semantics, indicating that the currently popular experimental setup and models may be poorly designed to understand and leverage game texts. To remedy this deficiency, we propose an inverse dynamics decoder to regularize the representation space and encourage exploration, which shows improved performance on several games including Zork I. We discuss the implications of our findings for designing future agents with stronger semantic understanding.