Sparse sequences of neural spikes are posited to underlie aspects of working memory, motor production, and learning. Discovering these sequences in an unsupervised manner is a longstanding problem in statistical neuroscience. Promising recent work utilized a convolutive nonnegative matrix factorization model to tackle this challenge. However, this model requires spike times to be discretized, utilizes a sub-optimal least-squares criterion, and does not provide uncertainty estimates for model predictions or estimated parameters. We address each of these shortcomings by developing a point process model that characterizes fine-scale sequences at the level of individual spikes and represents sequence occurrences as a small number of marked events in continuous time. This ultra-sparse representation of sequence events opens new possibilities for spike train modeling. For example, we introduce learnable time warping parameters to model sequences of varying duration, which have been experimentally observed in neural circuits. We demonstrate these advantages on experimental recordings from songbird higher vocal center and rodent hippocampus.

A channel corresponds to a viewpoint or transformation of an underlying meaning. A pair of parallel sentences in English and French express the same underlying meaning, but through two separate channels corresponding to their languages. In this work, we present the Multichannel Generative Language Model (MGLM). MGLM is a generative joint distribution model over channels. MGLM marginalizes over all possible factorizations within and across all channels. MGLM endows flexible inference, including unconditional generation, conditional generation (where 1 channel is observed and other channels are generated), and partially observed generation (where incomplete observations are spread across all the channels). We experiment with the Multi30K dataset containing English, French, Czech, and German. We demonstrate experiments with unconditional, conditional, and partially conditional generation. We provide qualitative samples sampled unconditionally from the generative joint distribution. We also quantitatively analyze the quality-diversity trade-offs and find MGLM outperforms traditional bilingual discriminative models.

We propose a few-shot learning method for spatial regression. Although Gaussian processes (GPs) have been successfully used for spatial regression, they require many observations in the target task to achieve a high predictive performance. Our model is trained using spatial datasets on various attributes in various regions, and predicts values on unseen attributes in unseen regions given a few observed data. With our model, a task representation is inferred from given small data using a neural network. Then, spatial values are predicted by neural networks with a GP framework, in which task-specific properties are controlled by the task representations. The GP framework allows us to analytically obtain predictions that are adapted to small data. By using the adapted predictions in the objective function, we can train our model efficiently and effectively so that the test predictive performance improves when adapted to newly given small data. In our experiments, we demonstrate that the proposed method achieves better predictive performance than existing meta-learning methods using spatial datasets.

Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.

We present a self-supervised learning approach to learning monocular 3D face reconstruction with a pose guidance network (PGN). First, we unveil the bottleneck of pose estimation in prior parametric 3D face learning methods, and propose to utilize 3D face landmarks for estimating pose parameters. With our specially designed PGN, our model can learn from both faces with fully labeled 3D landmarks and unlimited unlabeled in-the-wild face images. Our network is further augmented with a self-supervised learning scheme, which exploits face geometry information embedded in multiple frames of the same person, to alleviate the ill-posed nature of regressing 3D face geometry from a single image. These three insights yield a single approach that combines the complementary strengths of parametric model learning and data-driven learning techniques. We conduct a rigorous evaluation on the challenging AFLW2000-3D, Florence and FaceWarehouse datasets, and show that our method outperforms the state-of-the-art for all metrics.

In countries experiencing unprecedented waves of urbanization, there is a need for rapid and high-quality urban street design. Our study presents a novel deep learning powered approach, DeepStreet (DS), for automatic street network generation that can be applied to the urban street design with local characteristics. DS is driven by a Convolutional Neural Network (CNN) that enables the interpolation of streets based on the areas of immediate vicinity. Specifically, the CNN is firstly trained to detect, recognize and capture the local features as well as the patterns of the existing street network sourced from the OpenStreetMap. With the trained CNN, DS is able to predict street networks' future expansion patterns within the predefined region conditioned on its surrounding street networks. To test the performance of DS, we apply it to an area in and around the Eixample area in the City of Barcelona, a well-known example in the fields of urban and transport planning with iconic grid-like street networks in the centre and irregular road alignments farther afield. The results show that DS can (1) detect and self-cluster different types of complex street patterns in Barcelona; (2) predict both gridiron and irregular street and road networks. DS proves to have a great potential as a novel tool for designers to efficiently design the urban street network that well maintains the consistency across the existing and newly generated urban street network. Furthermore, the generated networks can serve as a benchmark to guide the local plan-making especially in rapidly-developing cities. Keywords: Urban street network, machine learning, deep learning, Convolutional Neural Network (CNN), Generative Adversarial Network (GAN), image completion, image inpainting

SCI COMMUN### Politics The U.S. presidential race was upended in the first days of October as President Donald Trump tested positive for the pandemic virus and spent 3 days in the hospital. He was aggressively treated with two experimental medicines—monoclonal antibodies and the repurposed antiviral remdesivir—and a steroid used in severe COVID-19 cases. Trump returned to the White House on 5 October saying people should not fear the disease. But public health specialists voiced astonishment when he re-entered the building maskless, trailed by questions about his medical condition and a lack of information about how staff members would be protected from infection. All that followed a rancorous first debate on 29 September between Trump and Democratic challenger Joe Biden. Trump mocked Biden for having worn a mask at other times, despite evidence that the precaution reduces transmission of the virus. The president also left scientists puzzled when he described as a “disaster” Biden's role in the response to the H1N1 swine flu outbreak in 2009. Then-President Barack Obama, whom Biden served under as vice president, declared it a public health emergency 6 weeks before the World Health Organization declared a pandemic. That flu killed an estimated 12,000 Americans—far fewer than the 210,000 U.S. deaths recorded so far from COVID-19. > “Very few don't have some sort of connection to Big Tech.” > > Doctoral student Mohamed Abdalla , in Wired , about a study he led of faculty members specializing in artificial intelligence at four leading research universities. He found 58% (48 of 83) had received a grant or fellowship from one of 14 large technology companies, which may distort research priorities. ### Conservation High-tech fake turtle eggs can spy on poachers and wildlife trafficking routes. The real eggs are a delicacy in Central America, and illicit trading of them adds to other hazards to the survival of turtle species that are threatened. Researchers slipped 101 decoy eggs with GPS trackers embedded (left) into nests on four Costa Rican beaches. The scientists tracked five eggs to learn where the poachers took them; the farthest ended up 137 kilometers inland, the multinational team reported on 5 October in Current Biology . The researchers did not share this information with authorities, noting ethical concerns; many poachers live in poverty, and in Costa Rica, buying the eggs is not illegal. But, the authors say, the study shows that law enforcement agencies could use the method. ### Public health Coronavirus guidelines issued last week by the U.S. Centers for Disease Control and Prevention (CDC) again stirred controversy and concerns that undue political pressure had influenced some of its decisions. CDC announced that on 31 October it will lift an order barring cruise ships from sailing despite a recommendation by its director, Robert Redfield, to extend the ban until February 2021. The industry shut down in March after severe COVID-19 outbreaks occurred on multiple ships. Last week, CDC also drew fire for its updated guidelines on when colleges should test students and faculty and staff members for the pandemic virus. The agency recommended different frequencies of testing, including just a single, initial one, depending on circumstances such as whether students lived in residences with others who tested positive. Critics said the new guidelines should have recommended more regular testing of asymptomatic individuals. CDC addressed another uproar this week by acknowledging evidence that the virus can travel by air and infect people standing more than 2 meters apart in indoor spaces. The agency was faulted last month after it posted, and then withdrew, a draft suggesting otherwise. ### Infectious diseases An international program to reduce the risk of new zoonotic diseases, allowed to expire by the U.S. government in 2019 but extended until last month, will get a successor. On 30 September, the United States Agency for International Development (USAID) awarded a $100 million grant to help countries in Asia and Africa curb viruses jumping from animals to humans. The 5-year Strategies to Prevent Spillover program will have a different focus from its predecessor, PREDICT, whose termination was criticized by the scientific community: Rather than studying the drivers of spillover, it will seek interventions to reduce viral jumps, a USAID spokesperson says. A key goal is to “help partners at the country level build their expertise and ability to take action,” says veterinarian Deborah Kochevar of Tufts University, which leads a 13-institute consortium that won the grant. ### International affairs Yuri Orlov, the Russian physicist who championed human rights in the Soviet Union before being exiled in 1985, died on 27 September at age 96. Orlov helped organize the Soviet Union's branch of Amnesty International in 1973 and 3 years later co-founded the Moscow Helsinki Group, which monitored Soviet adherence to the civil rights provisions of the 1975 Helsinki Accords between the Soviet Union and the West. In 1977, Orlov was arrested and sentenced to 12 years of hard labor and exile in Siberia. After coming to the United States in a prisoner exchange, Orlov, an expert in particle accelerators, worked at Cornell University. He didn't think much of Russian President Vladimir Putin, writing in 2004 that “Russia is flying backward in time.” ### Governance Japan's new prime minister, Yoshihide Suga, has disrupted the process by which scientists are appointed to serve on the governing body of the country's leading academic society, the Science Council of Japan (SCJ). Researchers are criticizing the move as a threat to academic freedom. SCJ makes policy recommendations, promotes scientific literacy and international cooperation, and represents the interests of more than 800,000 scholars in virtually all academic disciplines. The prime minister customarily ratifies appointees recommended by SCJ for its governing body, the General Assembly. But according to an announcement last week, Suga withheld his blessing from six academics, in a list of 105 put forward, who work in the social sciences, law, and the humanities. All six had criticized legislation adopted by Japan's previous government, in which Suga was chief cabinet secretary. His failure to appoint them violated a law governing SCJ, said Satoshi Ihara, secretary general of the Japan Scientists' Association. ### Policy Mexican scientists this week blasted a move by the national legislature to eliminate 109 trust funds run by public research centers and government institutes, one-third of them devoted to science and technology. The government wants to use the money, some $3 billion in total, for the coronavirus pandemic. The funds support everything from student scholarships and emergency maintenance of equipment to major research projects at dozens of government centers. The money also helps pay for biosecurity and biotechnology research, fighting climate change, and disaster relief. On 6 October, Mexico's Chamber of Deputies approved a bill to terminate the funds, but with “reservations” that require further debate; it is expected to pass in the Senate. The plan is “a brutal blow” and the worst hit to Mexican science in 50 years, says Antonio Lazcano, an evolutionary biologist at the National Autonomous University of Mexico, University City. ### Virology There's been a new case of infection with Alaskapox virus, a recently discovered pathogen that's related to smallpox. Alaska state health authorities reported on 30 September that they had found the virus in a woman from the Fairbanks area with a mild, gray skin lesion on one arm, similar to one seen in 2015 in the first known patient, also a woman from Fairbanks. Human infections with pox-viruses are on the rise, presumably because vaccination against smallpox—which offers some protection against related viruses—was halted after that deadly disease was eradicated 40 years ago. But the Alaska cases are no cause for alarm: There is no evidence the virus can be transmitted between humans—scientists think it came from wild mammals—and the lesions went away by themselves. ### Medicine prize goes to discoverers of virus that destroys the liver The Nobel Prize in Physiology or Medicine was awarded this week for the discovery of the hepatitis C virus, one of the most common causes of liver cancer. The prize went to Harvey Alter of the U.S. National Institutes of Health; Michael Houghton of the University of Alberta, Edmonton; and Charles Rice of Rockefeller University. The hepatitis C virus, transmitted via blood, can cause chronic inflammation of the liver that quietly destroys the organ over decades, ultimately leading to cirrhosis and cancer. The laureates did work over 3 decades to identify the virus and show it was responsible for unexplained cases of hepatitis in people who received blood transfusions. They also developed a test to screen blood donations for the virus, which has nearly eliminated the risk of hepatitis from blood transfusions. Their research ultimately led to a successful treatment for the disease, which has cured millions of people. But about 71 million people worldwide still have chronic hepatitis C, and transmission continues via contaminated medical equipment, sharing drug injection needles, and from infected mothers to newborns during birth. The disease causes few acute symptoms, and testing in many developing countries is limited. ### Black hole hunters receive physics prize The Nobel Prize in Physics has been awarded for pioneering discoveries regarding black holes—self-sustaining gravitational fields so intense that nothing, not even light, can escape. Roger Penrose, a mathematician at the University of Oxford, received half of the $1.1 million prize for his theoretical work, conducted in part with the late Stephen Hawking, that proved a black hole would be stable and thus could be a real astrophysical object and not a mere mathematical curiosity. Astronomers Reinhard Genzel of the Max Planck Institute for Extraterrestrial Physics and Andrea Ghez of the University of California, Los Angeles, share the other half of the prize for deducing the presence of the supermassive black hole that lies in the heart of our Galaxy. Since the 1990s, Genzel and Ghez have led rival research groups that observed stars there, 26,000 light-years from Earth. They found ones orbiting a heavy, unseen object, called Sagittarius A*, at incredible speeds—some of the most convincing evidence for a behemoth black hole, with the mass of millions of Suns. ### Fauci: ‘Skunk at the picnic’ On 23 September, in the relative calm before President Donald Trump's coronavirus infection was revealed, Anthony Fauci relaxed at home after tangling earlier that day with U.S. Senator Rand Paul (R–KY) during a hearing on COVID-19. Fauci still had 200 emails in his inbox to read that night, but the head of the National Institute of Allergy and Infectious Diseases, who also serves on the White House's Coronavirus Task Force, sat down with Science to discuss the pandemic and research on vaccines. (Read the full interview at [scim.ag/FauciOctober][1].) Some excerpts: On his showdown with Paul: “I said to myself, you know, ‘I'm sorry, I'm not gonna disrespect him, but I'm not gonna let him get away with saying things that are cherry-picked data.’” (Paul had suggested that the United States follow Sweden's COVID-19 policies because it had a lower death rate from the disease.) On speaking bluntly at the White House: “I'm walking a fine line of being someone who is not hesitant to tell the president and the vice president what they may not want to hear. There are some people in the White House, who, even when I first started telling it like it was in the task force meetings, they were like, ‘Oh my goodness.’ That's when I got that nickname ‘the skunk at the picnic.’ … I say, ‘I'm sorry, I'm not trying to undermine the president. But there is something that's called reality.’” On the state of the pandemic: “Yes, there are parts of the country that are doing well. But this country is a big forest, and when you have fires in some parts of the forest, the entire forest is at risk.” [1]: http://scim.ag/FauciOctober

Wilkinson recently had a close call when a shark trailed him, only inches away. Wilkinson recently had a close call when a shark trailed him, only inches away. The surfer had no idea a shark was trailing him. Near Sharpes Beach in Australia, professional surfer Matt Wilkinson was paddling on his board on Wednesday. Unbeknownst to him, a shark quickly surfaced and began stalking the surfing world champion, at one point only inches away.

Nvidia CEO Jensen Huang is confident that the Nvidia-Arm deal, the biggest in semiconductor history at $40 billion, will get past global regulators. Speaking during a "fireside chat" as part of Arm Dev Summit, the company's developer conference, Huang said that the two companies are complementary and as such, the acquisition would drive innovation forward and be good for customers. We are confident that [the Nvidia-Arm deal] is going to go through," Huang said. "As soon as we explain the rationale of the transaction and our plans to regulators around the world, they will realize that these are two complementary companies." Arm CEO Simon Segars backed Huang up, while noting that the regulatory process will take some time. "This is a deal that's about expanding," Segars said. It's about putting that technology in the hands of people who are going to build really cool stuff with it. So from that point of view, this is about enabling more people to do more things. That's the positive thing, that as regulators do scrutinize this, they're going to be able to see." Extended reach Huang's reasoning behind the Nvidia-Arm deal is that the combination of the world leader in AI compute with the world's most popular CPU architecture will bring AI capabilities to the Arm ecosystem as well as give Nvidia's accelerated computing extended reach.

There has been considerable progress made towards conversational models that generate coherent and fluent responses; however, this often involves training large language models on large dialogue datasets, such as Reddit. These large conversational models provide little control over the generated responses, and this control is further limited in the absence of annotated conversational datasets for attribute specific generation that can be used for fine-tuning the model. In this paper, we first propose and evaluate plug-and-play methods for controllable response generation, which does not require dialogue specific datasets and does not rely on fine-tuning a large model. While effective, the decoding procedure induces considerable computational overhead, rendering the conversational model unsuitable for interactive usage. To overcome this, we introduce an approach that does not require further computation at decoding time, while also does not require any fine-tuning of a large language model. We demonstrate, through extensive automatic and human evaluation, a high degree of control over the generated conversational responses with regard to multiple desired attributes, while being fluent.