X Web Viewer

US Defense Secretary Pete Hegseth said the US is seeking a framework for defense cooperation with Southeast Asia as Washington seeks to bolster partners as a counterweight to Beijing

Trees crashed onto highways, electricity poles were uprooted and thousands took shelter against Arnab, a powerful cyclone which made landfall on India's east coast, authorities said

From the suburbs of Paris to the squares of Belgrade, thousands have taken to the streets. The demands are various but often overlapping — frustrations with shabby schooling facilities, perceived lack of economic opportunity, cost of living, corruption, housing and a sense of not being listened to. At times those protests triggered violence — with thousands of shops in France alone being damaged or looted in recent weeks — and police firing stun grenades and water cannons. This weekend they also brought real change — student-led protests culminated in the toppling in an election of Serbia’s long-time leader Aleksandar Vucic after nearly two years of demonstrations against perceived corruption and creeping authoritarianism

Did softmax and cross-entropy have numerical bugs nobody actually fixed? A new paper from Google DeepMind says yes. Title: A Fix for Softmax and Cross-Entropy URL: ❓ What is the problem? Subtracting the max logit prevents overflow, but two bugs remain. In Softmax Collapse, the target gradient rounds to zero once the correct logit runs far ahead. In Softmax Cancellation, distillation subtracts nearly equal probabilities. ❓ How is it fixed? By rewriting in log space. The max logit is set to 0, the loss is softplus(-m), and the target gradient is defined as the negated sum of the others, so gradients always sum to zero. The distillation side uses expm1 and log1p to avoid the subtraction. Short JAX code is a drop-in replacement. ❓ Does it help? On a²+b² mod 97 grokking, standard test accuracy drops from 61% to 23% at the precision cliff, while the fix reaches 85%. In soft-target distillation, the fix keeps KL at exactly 0 in every precision. ❓ What did we learn? Slingshot spikes and grokking failures have a numerical cause, and the margin ceiling is about log(1/ε) from Adam's ε and weight decay. The experiments are small-scale, though, and large-scale impact is untested. #DeepLearning# #MachineLearning#

European trade-finance firm Zvilo Group is growing its sub-Saharan African business, with plans to make the continent its biggest market by the end of the decade as it expands into nine countries.