We present an extension of our Molecular Transformer architecture combined with a hyper-graph exploration strategy for automatic retrosynthesis route planning without human intervention. The single-step retrosynthetic model sets a new state of the art for predicting reactants as well as reagents, solvents and catalysts for each retrosynthetic step. We introduce new metrics (coverage, class diversity, round-trip accuracy and Jensen-Shannon divergence) to evaluate the single-step retrosynthetic models, using the forward prediction and a reaction classification model always based on the transformer architecture. The hypergraph is constructed on the fly, and the nodes are filtered and further expanded based on a Bayesian-like probability. We critically assessed the end-to-end framework with several retrosynthesis examples from literature and academic exams. Overall, the frameworks has a very good performance with few weaknesses due to the bias induced during the training process. The use of the newly introduced metrics opens up the possibility to optimize entire retrosynthetic frameworks through focusing on the performance of the single-step model only.

A BSTRACT To analyze deep ReLU network, we adopt a student-teacher setting in which an over-parameterized student network learns from the output of a fixed teacher network of the same depth, with Stochastic Gradient Descent (SGD). First, we prove that when the gradient is zero (or bounded above by a small constant) at every data point in training, a situation called interpolation setting, there exists many-to-one alignment between student and teacher nodes in the lowest layer under mild conditions. This suggests that generalization in unseen dataset is achievable, even the same condition often leads to zero training error. Second, analysis of noisy recovery and training dynamics in 2-layer network shows that strong teacher nodes (with large fan-out weights) are learned first and subtle teacher nodes are left unlearned until late stage of training. As a result, it could take a long time to converge into these small-gradient critical points. Our analysis shows that over-parameterization plays two roles: (1) it is a necessary condition for alignment to happen at the critical points, and (2) in training dynamics, it helps student nodes cover more teacher nodes with fewer iterations. Although networks with even one-hidden layer can fit any function (Hornik et al., 1989), it remains an open question how such networks can generalize to new data. Different from what traditional machine learning theory predicts, empirical evidence (Zhang et al., 2017) shows more parameters in neural network lead to better generalization. How over-parameterization yields strong generalization is an important question for understanding how deep learning works. In this paper, we analyze multi-layer ReLU networks by adopting teacher-student setting. The fixed teacher network provides the output for the student to learn via SGD. The student is over-parameterized (or over-realized): it has more nodes than the teacher. Therefore, there exists student weights whose gradient at every data point is zero. Here, we want to study the inverse problem: With small gradient at every training sample, can the student weights recover the teachers'? If so, then the generalization performance can be guaranteed if the training converges to such critical points. In this paper, we show that this so-called interpolation setting (Ma et al., 2017; Liu & Belkin, 2018; Bassily et al., 2018) leads to alignment: under certain conditions, each teacher node is provably aligned with at least one student node in the lowest layer. The condition is simply that the teacher node is observed by at least one student node, i.e., teacher's ReLU boundary lies in the activation region of that student. Therefore, more over-parameterization increases the probability of teachers being observed and thus being aligned. Furthermore, in 2-layer case, those student nodes that are not aligned with any teacher have zero contribution to the output and can be pruned.

In our oceans the scale of disasters is measured in millions, billions, and trillions, while solutions amount to single digits: individuals or institutions working to impact a chosen issue with approaches often both brilliant and quixotic. Putting such individuals in close contact with both whales and billionaires is the strange alchemy being attempted by the Sustainable Ocean Alliance's Accelerator at Sea. I and a few other reporters were invited to observe said program, a five-day excursion in Alaska that put recent college graduates, aspiring entrepreneurs, legends of the sea, and soft-spoken financial titans on the same footing: spotting whales from Zodiacs in the morning, learning from one another in the afternoon, and drinking whiskey good and bad under the Northern lights in the pre-dawn dark. In that time I got to know the dozen or so companies in the accelerator, the second batch from the SOA but the first to experience this oddly effective enterprise. And I also gathered from conversations among the group the many challenges facing conservation-focused startups. The picture painted by just about everyone was one of impending doom from a multiplicity of interlinked trends, and as many different approaches to averting or mitigating that doom as people discussing it.

Farming and advanced AI may seem antithetical, but they're not. The venerable farm equipment company has not only long embraced advanced technologies, the company for years has evangelized adoption of high performance clusters and simulation software for product design. And Deere freely states it's an extremely complex undertaking. In a recent article in IEEE, Deere's Julian Sanchez, who heads the Moline, IL, company's intelligent vehicles strategy, said that while the company is working on autonomous driving, "it's not just about driving tractors around." The more difficult problem, he said, is crop classification.

Zyfra says it has developed an automated system using artificial intelligence (AI) to monitor the condition of excavator bucket teeth based on its machine vision BucketControl system. The system is designed to detect the presence or absence of excavator bucket crowns quickly and features functions to alert the excavator operator if a crown is lost or ceases to work. The application, developed jointly by the AI and Mining divisions of Zyfra, uses an on-board controller to acquire images from the camera, process and analyse them using internal software and sends a signal to the operator if a crown is lost or ceases to work. The wear of the tooth is also assessed, and when a critical value is reached, a notification is sent to the dispatcher, according to the company. This data is transmitted to the server in real time, Zyfra added.

While enterprise adoption of AI has grown 270% over the past four years, engineers have been slow to adopt the new technology. Some fear job insecurity and others simply don't understand the technology. But it's through automation that we're able to save time doing mundane, repetitive tasks. And that time can be reinvested in more important things, like design, development, and creativity.

Cross-laminated timber, otherwise known as CLT, is a prefabricated, engineered wood building material with unique and often superior building, aesthetic, environmental, and cost attributes. CLT wood panels are made by pressing perpendicular layers of lumber together with a layer of formaldehyde-free adhesive. The fusion of orthogonal wood layers gives CLT biaxial strength, durability, and stability. CLT can serve as a system-based approach for floors, walls, and roofs to form a high-performance and sustainable timber building of virtually any type. Code Council) adopted tall wood building codes for up to 18 stories.

Just occasionally, Alán Aspuru-Guzik has a movie-star moment, when fans half his age will stop him in the street. "They say, 'Hey, we know who you are'," he laughs. "Then they tell me that they also have a quantum start-up, and would love to talk to me about it." "I don't usually have time to talk, but I'm always happy to give them some tips." That affable approach is not uncommon in the quantum-computing community, says Aspuru-Guzik, who is a computer scientist at the University of Toronto, Canada, and co-founder of quantum-computing company Zapata Computing in Cambridge, Massachusetts.

According to an analysis released by OpenAI, the demand for computing power has increased by more than 300,000 times in the six years after 2012. It grows by about factor of 10 each year, far exceeding the pace set by Moore's Law. As a latecomer to artificial intelligence (AI), Huawei boldly proposed to provide the industry with computing power that is accessible, affordable, and easy to use, to meet the exponentially increasing demand for AI computing. Now, one year after the AI strategy was proposed, has Huawei found a way to address the computing power challenges? In the late 17th century, the British mining industry, particularly the coal mine, was developed to a considerable scale.

By the end of this decade, a milestone is reached in artificial intelligence, with computers now routinely passing the Turing Test.** This test is conducted by a human judge who is made to engage in a natural language conversation with one human and one machine, each of which tries to appear human. Participants are placed in isolated locations. For several decades, information technology had seen exponential growth – leading to vast improvements in computer processing power, memory, bandwidth, voice recognition, image recognition, deep learning and other software algorithms. By the end of the 2020s, it has reached the stage where an independent judge is literally unable to tell which is the real human and which is not.* Answers to certain "obscure" questions posed by the judge may appear childlike from the AI – but they are humanlike nonetheless.*