We consider a covariate shift problem where one has access to several different training datasets for the same learning problem and a small validation set which possibly differs from all the individual training distributions. The distribution shift is due, in part, to \emph{unobserved} features in the datasets. The objective, then, is to find the best mixture distribution over the training datasets (with only observed features) such that training a learning algorithm using this mixture has the best validation performance. Our proposed algorithm, \textsf{Mix\&Match}, combines stochastic gradient descent (SGD) with optimistic tree search and model re-use (evolving partially trained models with samples from different mixture distributions) over the space of mixtures, for this task. We prove a novel high probability bound on the final SGD iterate without relying on a global gradient norm bound, and use it to show the advantages of model re-use. Additionally, we provide simple regret guarantees for our algorithm with respect to recovering the optimal mixture, given a total budget of SGD evaluations.

I would mention that the presence of unobserved feature is something that was not dealt with before, but more general and cannot be handle from existing work.

The reviewers generally liked this paper and also provided a number of suggestions for improvement. Please take these recommendations seriously when revising the paper. In particular, I agree with Reviewer 4 that the informal theorem statements in the main body obscure many details. Theorem 2, in particular, seems to be simultaneously too formal (do we need all these exact numeric constants?), while also obscuring important details. Overall, the ideas are interesting but I found the paper somehow a bit messy to read.

We consider a covariate shift problem where one has access to several different training datasets for the same learning problem and a small validation set which possibly differs from all the individual training distributions. The distribution shift is due, in part, to \emph{unobserved} features in the datasets. The objective, then, is to find the best mixture distribution over the training datasets (with only observed features) such that training a learning algorithm using this mixture has the best validation performance. Our proposed algorithm, \textsf{Mix\&Match}, combines stochastic gradient descent (SGD) with optimistic tree search and model re-use (evolving partially trained models with samples from different mixture distributions) over the space of mixtures, for this task. We prove a novel high probability bound on the final SGD iterate without relying on a global gradient norm bound, and use it to show the advantages of model re-use.

As Qualcomm-powered Windows on Arm PCs begin appearing here at Computex, ushering in a generation of AI-infused Copilot laptops, it seemed appropriate to interview a major player in the push. Instead, I mean Arm, the semiconductor design company that licenses CPUs to companies like Qualcomm, Apple, and Samsung. Arm dominates in smartphones and tablets, and now, true PC contention finally seems possible. I sat down with chief executive Rene Haas in Taipei, touching upon everything from NPUs, to how Arm solved its Windows app gap, to why Intel, AMD, and Qualcomm don't matter to the success of Windows on Arm PCs. And he has nothing but praise for Apple's M-series Macs, which he says "woke up the industry on the art of the possible" with Arm laptops. "I think Apple silicon has really proven that you could build a first-class laptop and have no compromises," Haas said. This interview has been slightly edited for length and clarity.

A team of MIT engineers is designing a kit of universal robotic parts that an astronaut could easily mix and match to build different robot "species" to fit various missions on the moon. When astronauts begin to build a permanent base on the moon, as NASA plans to do in the coming years, they'll need help. Robots could potentially do the heavy lifting by laying cables, deploying solar panels, erecting communications towers, and building habitats. But if each robot is designed for a specific action or task, a moon base could become overrun by a zoo of machines, each with its own unique parts and protocols. To avoid a bottleneck of bots, a team of MIT engineers is designing a kit of universal robotic parts that an astronaut could easily mix and match to rapidly configure different robot "species" to fit various missions on the moon.

Rather, it's because people value their scarcity, prospect their future price, appreciate their proof-of-concept of a new artform, or love the artist who created them. However, a new trend in the art world, AI-generated art, may soon disrupt the multibillion-dollar NFT space. First, let's look at how collections of NFTs are created today. In most high-profile NFT projects, artists design multiple classes of varying attributes such as hair color, background color, or skin tone. Then, artists will mix and match these attributes to create a collection of "unique" NFTs, where no two NFTs look the same.

Epic Games, maker of the video game'Fornite,' has released a sneak peek of its browser-based software tool that lets developers create'digital humans.' Powered by the firm's Unreal Engine, the MetaHuman Creator provides dozens of hairstyles, ear types, lip shades and more, allowing users to mix and match to create 3D characters that move and speak as if they were humans - all of which can be completed in less than an hour. The system works by pulling human appearance and motion from a massive library of variants, which are used as a starting point for developers who can then mix and match the given presets. Epic says that'when you are happy with your human,' you can download the digital creation to use in films, video games and a variety of apps. Epic Games, maker of the video game'Fornite,' has released a sneak peek of its browser-based software tool called MetaHuman that lets developers create'digital humans' Creating digital humans is not a breakthrough, as there are a number applications capable of the task, but Epic says MetaHuman allows users to design 3D characters in less than one hour – others can take weeks to months.

We consider a co-variate shift problem where one has access to several marginally different training datasets for the same learning problem and a small validation set which possibly differs from all the individual training distributions. This co-variate shift is caused, in part, due to unobserved features in the datasets. The objective, then, is to find the best mixture distribution over the training datasets (with only observed features) such that training a learning algorithm using this mixture has the best validation performance. Our proposed algorithm, ${\sf Mix\&Match}$, combines stochastic gradient descent (SGD) with optimistic tree search and model re-use (evolving partially trained models with samples from different mixture distributions) over the space of mixtures, for this task. We prove simple regret guarantees for our algorithm with respect to recovering the optimal mixture, given a total budget of SGD evaluations. Finally, we validate our algorithm on two real-world datasets.