We establish a direct connection between general tensor networks and deep feed-forward artificial neural networks. The core of our results is the construction of neural-network layers that efficiently perform tensor contractions, and that use commonly adopted non-linear activation functions. The resulting deep networks feature a number of edges that closely matches the contraction complexity of the tensor networks to be approximated. In the context of many-body quantum states, this result establishes that neural-network states have strictly the same or higher expressive power than practically usable variational tensor networks. As an example, we show that all matrix product states can be efficiently written as neural-network states with a number of edges polynomial in the bond dimension and depth logarithmic in the system size. The opposite instead does not hold true, and our results imply that there exist quantum states that are not efficiently expressible in terms of matrix product states or PEPS, but that are instead efficiently expressible with neural network states.

Given a deep learning model trained on data from a source site, how to deploy the model to a target hospital automatically? How to accommodate heterogeneous medical coding systems across different hospitals? Standard approaches rely on existing medical code mapping tools, which have significant practical limitations. To tackle this problem, we propose AutoMap to automatically map the medical codes across different EHR systems in a coarse-to-fine manner: (1) Ontology-level Alignment: We leverage the ontology structure to learn a coarse alignment between the source and target medical coding systems; (2) Code-level Refinement: We refine the alignment at a fine-grained code level for the downstream tasks using a teacher-student framework. We evaluate AutoMap using several deep learning models with two real-world EHR datasets: eICU and MIMIC-III. Results show that AutoMap achieves relative improvements up to 3.9% (AUC-ROC) and 8.7% (AUC-PR) for mortality prediction, and up to 4.7% (AUC-ROC) and 3.7% (F1) for length-of-stay estimation. Further, we show that AutoMap can provide accurate mapping across coding systems. Lastly, we demonstrate that AutoMap can adapt to the two challenging scenarios: (1) mapping between completely different coding systems and (2) between completely different hospitals.

Alphabet's drone service Wing this morning announced another milestone, as it hit 200,000 commercial deliveries. The number, which the firm says excludes test flights, comes half-a-year after it hit 100,000. Australia, which has been the primary market for testing and commercial deployment, comprises 30,000 of those deliveries in the first two months of this year. Broken down further, Wing says it was up to more than 1,000 deliveries in a day, or one delivery every 25 seconds or so. The big round number arrives as it announces a commercial partnership with Coles, one of Australia's leading supermarket chains.

Jobs in Artificial Intelligence (AI), Machine Learning (ML), Data Science, Data Engineering, Big Data, NLP, Deep Learning ,Karkidi

SARDI scientist Rhiannon Schilling showcases a demonstration application created by using paddock data and machine learning models. PINPOINTING the cause of paddock-yield variability using large data sets and innovative machine-learning models is the focus of a project led by the University of Adelaide and funded by the Grains Research and Development Corporation (GRDC). South Australian Research and Development Institute (SARDI) agriculture scientist Rhiannon Schilling gave an update of the project at ABARES Outlook online this week. Ms Schilling said the research looked at the challenge of working out what is behind variability in crop growth and yield across paddocks. "Often there has been a focus on improving grain yields of our varieties; but when we drive around and have a look at our paddocks, we can see that we are not always achieving uniform crop growth and yield across our paddocks," Ms Schilling said.

The BBC's Reith Lectures in 2022 featured Professor Stuart Russell from the University of California, Berkeley discussing artificial intelligence: its promises and its threats. Toby Walsh, Professor of AI at the University of NSW responds. He says autonomous vehicles have the potential to save 1,000 lives a year in Australia, people who currently die in road accidents as drivers drive whilst tired, drunk, or distracted. But autonomous war machines could wreak devastation. Toby Walsh says we need to stay in control and must not make ourselves redundant.

This thesis rigorously studies fundamental reinforcement learning (RL) methods in modern practical considerations, including robust RL, distributional RL, and offline RL with neural function approximation. The thesis first prepares the readers with an overall overview of RL and key technical background in statistics and optimization. In each of the settings, the thesis motivates the problems to be studied, reviews the current literature, provides computationally efficient algorithms with provable efficiency guarantees, and concludes with future research directions. The thesis makes fundamental contributions to the three settings above, both algorithmically, theoretically, and empirically, while staying relevant to practical considerations.

This study proposes a data condensation method for multivariate kernel density estimation by genetic algorithm. First, our proposed algorithm generates multiple subsamples of a given size with replacement from the original sample. The subsamples and their constituting data points are regarded as $\it{chromosome}$ and $\it{gene}$, respectively, in the terminology of genetic algorithm. Second, each pair of subsamples breeds two new subsamples, where each data point faces either $\it{crossover}$, $\it{mutation}$, or $\it{reproduction}$ with a certain probability. The dominant subsamples in terms of fitness values are inherited by the next generation. This process is repeated generation by generation and brings the sparse representation of kernel density estimator in its completion. We confirmed from simulation studies that the resulting estimator can perform better than other well-known density estimators.

On 30 July 2021, the Federal Court of Australia decided that AI systems can be inventors. In a word-first determination of Thaler v Commissioner of Patents,{[2021] FCA 879, ('Thaler')}, the Honourable Justice Beach found that AI systems can be the inventors on a patent application under Australian patent law. The decision has been appealed to the Full Bench of the Federal Court, which may decide to overrule it. For now, however, the decision is binding in Australia. Read on to find out what a patent is and an overview of the decision.

The detection of computer-generated text is an area of rapidly increasing significance as nascent generative models allow for efficient creation of compelling human-like text, which may be abused for the purposes of spam, disinformation, phishing, or online influence campaigns. Past work has studied detection of current state-of-the-art models, but despite a developing threat landscape, there has been minimal analysis of the robustness of detection methods to adversarial attacks. To this end, we evaluate neural and non-neural approaches on their ability to detect computer-generated text, their robustness against text adversarial attacks, and the impact that successful adversarial attacks have on human judgement of text quality. We find that while statistical features underperform neural features, statistical features provide additional adversarial robustness that can be leveraged in ensemble detection models. In the process, we find that previously effective complex phrasal features for detection of computer-generated text hold little predictive power against contemporary generative models, and identify promising statistical features to use instead. Finally, we pioneer the usage of $\Delta$MAUVE as a proxy measure for human judgement of adversarial text quality.