Climate and weather prediction traditionally relies on complex numerical simulations of atmospheric physics. Deep learning approaches, such as transformers, have recently challenged the simulation paradigm with complex network forecasts. However, they often act as data-driven black-box models that neglect the underlying physics and lack uncertainty quantification. We address these limitations with ClimODE, a spatiotemporal continuous-time process that implements a key principle of advection from statistical mechanics, namely, weather changes due to a spatial movement of quantities over time. ClimODE models precise weather evolution with value-conserving dynamics, learning global weather transport as a neural flow, which also enables estimating the uncertainty in predictions. Our approach outperforms existing data-driven methods in global and regional forecasting with an order of magnitude smaller parameterization, establishing a new state of the art.

One of the objectives of continual learning is to prevent catastrophic forgetting in learning multiple tasks sequentially, and the existing solutions have been driven by the conceptualization of the plasticity-stability dilemma. However, the convergence of continual learning for each sequential task is less studied so far. In this paper, we provide a convergence analysis of memory-based continual learning with stochastic gradient descent and empirical evidence that training current tasks causes the cumulative degradation of previous tasks. We propose an adaptive method for nonconvex continual learning (NCCL), which adjusts step sizes of both previous and current tasks with the gradients. The proposed method can achieve the same convergence rate as the SGD method when the catastrophic forgetting term which we define in the paper is suppressed at each iteration. Further, we demonstrate that the proposed algorithm improves the performance of continual learning over existing methods for several image classification tasks.

Washington, DC – The response from US President Joe Biden's administration to Iran's historic missile and drone attack on Israel has been two-fold: Washington has re-upped its pledge to always stand by its "ironclad" ally Israel, while also appealing to the government of Benjamin Netanyahu not to take further action that could drag the region into wider war. The days ahead will show if those two options are compatible, or if the two governments' priorities are on collision course, analysts told Al Jazeera. In the short term, the Iranian attack is a coup for both Israel and its backers in the US: From their perspective, it offers renewed justification for military support to Israel while weakening the world's focus on alleged abuses committed in Gaza in seven months of war, according to Trita Parsi, the executive vice president of the Washington-based Quincy Institute. But defiance from Netanyahu to US calls for restraint could find the Biden administration further hamstrung by its political and ideological commitments to Israel, which could eventually drag Washington into a wider war, he added. "The Israelis have been told by Biden to take this as a win and stop here," Parsi told Al Jazeera.

Video game conventions are typically boisterous affairs, as thousands of visitors queue under a constellation of screens for the chance to play one of the hundreds of as-yet unreleased titles on display. Now in its 10th year, Somerset House's Now Play This is to mainstream exhibitions what folk festivals are to raves. None of the experimental games presented here are destined to be advertised on the sides of buses, not least because many are one-offs that use bespoke controllers – a hatching of thick ropes and copper bands, or an old suitcase lined with speakers – connected to laptops via an umbilical tangle of wires. Few of these games adhere to the conventional rules or fashions seen in mainstream video game design, either. They might have no "win state", or provide an "open play" tool set with which visitors can create their own rules.

The Palestinian Hamas group has expressed support for Iran after it launched hundreds of missiles and drones at Israel in retaliation for a deadly assault on its consulate in the Syrian capital, Damascus. In a statement on Saturday, Hamas, which governs the Palestinian enclave of Gaza, said it affirmed "the natural right" of countries and peoples in the Middle East to defend themselves "in the face of Zionist aggressions". "The military operation carried out by Iran against the Zionist entity is a natural right and a due response to the crime of targeting the consulate in Damascus," it said. The Iranian salvo, fired late on Saturday night, consisted of more than 300 cruise missiles, ballistic missiles and drones, according to the Israeli military. Some 99 percent of the projectiles were intercepted, an Israeli spokesman said, with help from France, the United Kingdom and the United States.

Fox News chief national security correspondent Jennifer Griffin reports that the U.S. military has shot down'dozens' of ballistic and cruise missiles as well as attack drones. International world leaders collectively condemned Iran's drone and missile attacks on Israel, calling for peace in the Middle East. British Prime Minister Rishi Sunak said he condemned Iran's "reckless" attack against Israel. "I condemn in the strongest terms the Iranian regime's reckless attack against Israel," Sunak said in a statement. "Iran has once again demonstrated that it is intent on sowing chaos in its own backyard."

Fox News chief national security correspondent Jennifer Griffin reports that the U.S. military has shot down'dozens' of ballistic and cruise missiles as well as attack drones. Israelis breathed easier in the early hours of Sunday, after an attack from Iranian drones and missiles largely fizzled in the face of the Jewish state's powerful air defenses with assistance from the U.S. military and other allies. A senior US official told Fox News that the U.S. military has shot down "at least dozens" of Iranian drones en route to Israel early Sunday local time, adding that U.S. military engagement is ongoing. "U.S. forces in the region continue to shoot down Iranian-launched drones targeting Israel. Our forces remain postured to provide additional defensive support and to protect U.S. forces operating in the region," a U.S. military official told Fox News.

Trump's third visit to the battleground state comes just one day after his election integrity joint press conference with House Speaker Mike Johnson. Following Iran's attack on Israel, former President Trump shared a heated message to those in attendance at his rally in Pennsylvania on Saturday, vowing it should not have happened. "Before going any further, I want to say God bless the people of Israel. That's because we show great weakness," Trump said to open his speech. "The weakness that we've shown is unbelievable, and it would not have happened if we were in office. They know that, and everybody knows that."

On Saturday, Iran launched more than 200 drones and cruise missiles at Israel. As the drones made their way across the Middle East en route to their target, Israel has invoked a number of defense systems to impede their progress. None will be more important than the Iron Dome. The Iron Dome, operational for well over a decade, comprises at least 10 missile-defense batteries strategically distributed around the country. When radar detects incoming objects, it sends that information back to a command-and-control center, which will track the threat to assess whether it's a false alarm, and where it might hit if it's not.

Recent work to enhance data partitioning strategies for more realistic model evaluation face challenges in providing a clear optimal choice. This study addresses these challenges, focusing on morphological segmentation and synthesizing limitations related to language diversity, adoption of multiple datasets and splits, and detailed model comparisons. Our study leverages data from 19 languages, including ten indigenous or endangered languages across 10 language families with diverse morphological systems (polysynthetic, fusional, and agglutinative) and different degrees of data availability. We conduct large-scale experimentation with varying sized combinations of training and evaluation sets as well as new test data. Our results show that, when faced with new test data: (1) models trained from random splits are able to achieve higher numerical scores; (2) model rankings derived from random splits tend to generalize more consistently.