NN-LM (Khandelwal et al., 2020) modify the LM output with a non-parametric Work done during internship at Meta. 38th Conference on Neural Information Processing Systems (NeurIPS 2024).

Recent advancements in Large Language Models (LLMs) have significantly enhanced the ability to develop systems that comprehend customer requests and determine the necessary actions to fulfill them. In today's competitive market, delivering superior custome r service is crucial for attracting and retaining clients. Satisfied customers are more likely to become loyal, repeat buyers, and advocate for your brand, leading to increased revenue and market share (Strikingly, 2024) . In industries characterized by intense competition, implementing LLM - based services that effectively address customer needs and enhance satisfaction is becoming a key determinant of a company's growth and success. By leveraging LLMs, businesses can deliver more personalized, efficient, and scalable support, and thereby improve customer experience and foster loyalty (Iopex, 2024) .

We study empirical Bayes estimation in high-dimensional linear regression. To facilitate computationally efficient estimation of the underlying prior, we adopt a variational empirical Bayes approach, introduced originally in Carbonetto and Stephens (2012) and Kim et al. (2022). We establish asymptotic consistency of the nonparametric maximum likelihood estimator (NPMLE) and its (computable) naive mean field variational surrogate under mild assumptions on the design and the prior. Assuming, in addition, that the naive mean field approximation has a dominant optimizer, we develop a computationally efficient approximation to the oracle posterior distribution, and establish its accuracy under the 1-Wasserstein metric. This enables computationally feasible Bayesian inference; e.g., construction of posterior credible intervals with an average coverage guarantee, Bayes optimal estimation for the regression coefficients, estimation of the proportion of non-nulls, etc. Our analysis covers both deterministic and random designs, and accommodates correlations among the features. To the best of our knowledge, this provides the first rigorous nonparametric empirical Bayes method in a high-dimensional regression setting without sparsity.

The deep linear network (DLN) is a model for implicit regularization in gradient based optimization of overparametrized learning architectures. Training the DLN corresponds to a Riemannian gradient flow, where the Riemannian metric is defined by the architecture of the network and the loss function is defined by the learning task. We extend this geometric framework, obtaining explicit expressions for the volume form, including the case when the network has infinite depth. We investigate the link between the Riemannian geometry and the training asymptotics for matrix completion with rigorous analysis and numerics. We propose that under small initialization, implicit regularization is a result of bias towards high state space volume.

All qualified applicants will receive due consideration for employment without any discrimination. All applicants will be evaluated solely on the basis of their ability, competence and their proven capability to perform the functions outlined in the corresponding role. We promote and support a diverse workforce across all levels in the company.

While most executives at financial institutions agree that artificial intelligence (AI) is important to their organization's success, few have fully implemented AI projects. In a recent Cognizant survey of 230 financial services executives, three-quarters said AI is extremely or very important to the success of their organizations. However, only 61% of those were aware of an AI project at their company. Even more telling, only 29% were aware of a project that had been fully implemented. Clearly, AI is quickly becoming a competitive requirement, creating the risk that those who are not implementing or updating AI capabilities will fall behind.

"Deep learning is neither deep, nor is it learning," says Babak Hodjat, the vice president of projects for "Evolutionary AI" at IT services giant Cognizant Technologies. Hodjat's critique is part of a fascinating exploration of AI taking shape at IT services firm Cognizant Technology Solutions, a twenty-five-year-old company based in Teaneck, New Jersey that last year made nearly $16 billion in revenue serving some of the biggest companies in the world. For years, this IT giant has talked about "digital transformation," something that is large and significant but also something hard to get one's mind around because it very often seems vague and undefined. And then in December, Cognizant gave a whole new grounding and precision to that digital work by acquiring certain assets from an eleven-year-old AI startup Sentient Technologies. The company, co-founded by Hodjat, has been pursuing a thrilling line of work in what's called "evolutionary computation," where many algorithms, including conventional artificial neural networks, can be tested in parallel for "fitness," to select an optimal network to perform a task.

Alexa will put up with just about anything. She has a remarkable tolerance for annoying behavior, and she certainly doesn't care if you forget your please and thank yous. But while artificial intelligence technology can blow past such indignities, parents are still irked by their kids' poor manners when interacting with Alexa, the assistant that lives inside the Amazon Echo. "I've found my kids pushing the virtual assistant further than they would push a human," says Avi Greengart, a tech analyst and father of five who lives in Teaneck, New Jersey. "[Alexa] never says'That was rude' or'I'm tired of you asking me the same question over and over again.'"

My Ph.D. dissertation (Calistri 1990) extends traditional methods of plan recognition to handle situations in which agents have flawed plans. This extension involves solving two problems: determining what sorts of mistakes people make when they reason about plans and figuring out how to recognize these mistakes when they occur. I have developed a complete classification of plan-based misconceptions, which categorizes all ways that a plan can fail, and I have developed a probabilistic interpretation of these misconceptions that can be used in principle to guide a best-first search algorithm. I have also developed a program called Pathfinder that embodies a practical implementation of this theory. Pathfinder is a probability-based plan-recognition.