Transformers have achieved significant success in natural language modeling because of their exceptional capabilities to combine contextual information and global knowledge, yet their theoretical basis remains unclear. In this paper, we first propose Sparse Contextual Bigram (SCB), a natural extension to the classical bigram model, where the generation of the next token depends on a sparse set of earlier positions determined by the last token. We investigate the training dynamics and sample complexity of learning SCB using a one-layer linear transformer with a gradient-based algorithm. We show that when trained from scratch, the training process can be split into an initial sample-intensive stage where the correlation is boosted from zero to a nontrivial value, followed by a more sample-efficient stage of further improvement. Additionally, we prove that, provided a nontrivial correlation between the downstream and pretraining tasks, finetuning from a pretrained model allows us to bypass the initial sample-intensive stage. We also empirically demonstrate that our algorithm can outperform SGD in our setting.

In order to reduce the computational complexity of large language models, great efforts have been made to to improve the efficiency of transformer models such as linear attention and flash-attention. However, the model size and corresponding computational complexity are constantly scaled up in pursuit of higher performance. In this work, we present MemoryFormer, a novel transformer architecture which significantly reduces the computational complexity (FLOPs) from a new perspective. We eliminate nearly all the computations of the transformer model except for the necessary computation required by the multi-head attention operation. This is made possible by utilizing an alternative method for feature transformation to replace the linear projection of fully-connected layers. Specifically, we first construct a group of in-memory lookup tables that store a large amount of discrete vectors to replace the weight matrix used in linear projection.

We consider a clustering problem where a learner seeks to partition a finite set by querying a faulty oracle. This models applications where learners crowdsource information from non-expert human workers or conduct noisy experiments to determine group structure. The learner aims to exactly recover a partition by submitting queries of the form ``are $u$ and $v$ in the same group?''

Multimodal Large Language Models (MLLMs) have demonstrated a wide range of capabilities across many domains including Embodied AI. In this work, we study how to best ground a MLLM into different embodiments and their associated action spaces, including both continuous and discrete actions. For continuous actions, a set of learned tokenizations that capture an action at various resolutions allows for sufficient modeling precision, yielding the best performance on downstream tasks. For discrete actions, semantically aligning these actions with the native output token space of the MLLM leads to the strongest performance. We arrive at these lessons via a thorough study of seven action grounding approaches on five different environments, encompassing over 114 embodied tasks.

Large language models (LLMs) have achieved significant success across various domains. However, training these LLMs typically involves substantial memory and computational costs during both forward and backward propagation. While parameter-efficient fine-tuning (PEFT) considerably reduces the training memory associated with parameters, it does not address the significant computational costs and activation memory. In this paper, we propose Dropping Backward Propagation (DropBP), a novel approach designed to reduce computational costs and activation memory while maintaining accuracy. DropBP randomly drops layers during backward propagation, which is essentially equivalent to training shallow submodules generated by undropped layers and residual connections. Additionally, DropBP calculates the sensitivity of each layer to assign an appropriate drop rate, thereby stabilizing the training process. DropBP is not only applicable to full fine-tuning but can also be orthogonally integrated with all types of PEFT by dropping layers during backward propagation. Specifically, DropBP can reduce training time by 44% with comparable accuracy to the baseline, accelerate convergence to the same perplexity by 1.5$\times$, and enable training with a sequence length 6.2$\times$ larger on a single NVIDIA-A100 GPU.

The premise of semi-supervised learning (SSL) is that combining labeled and unlabeled data yields significantly more accurate models.Despite empirical successes, the theoretical understanding of SSL is still far from complete. In this work, we study SSL for high dimensional sparse Gaussian classification. To construct an accurate classifier a key task is feature selection, detecting the few variables that separate the two classes.For this SSL setting, we analyze information theoretic lower bounds for accurate feature selection as well as computational lower bounds, assuming the low-degree likelihood hardness conjecture. Our key contribution is the identification of a regime in the problem parameters (dimension, sparsity, number of labeled and unlabeled samples) where SSL is guaranteed to be advantageous for classification.Specifically, there is a regime where it is possible to construct in polynomial time an accurate SSL classifier.However, any computationally efficient supervised or unsupervised learning schemes, that separately use only the labeled or unlabeled data would fail. Our work highlights the provable benefits of combining labeled and unlabeled data for classification and feature selection in high dimensions. We present simulations that complement our theoretical analysis.

NTT Global Data Centers is working on 34 projects to double its capacity to 4 gigawatts within as little as two years, CEO Doug Adams said, as it races to meet surging global demand driven by the AI boom. NTT Global Data Centers, the world's third-largest data center provider outside of China, is working to double its capacity to 4 gigawatts to meet the rising global demand for the critical digital infrastructure amid an artificial intelligence boom. The unit of Japan's NTT is working on 34 projects that will double its capacity in as soon as two years, according to the data center business's Chief Executive Officer Doug Adams. Capacity will continue to increase from there, and will be "well over 5 gigawatts" in five years, Adams said in an interview. NTT GDC has seen increasing demand from companies moving more of their software and operations to the cloud as well as businesses hunting for extra capacity to run AI programs. The business's revenue is expected to keep growing at more than 20% a year, Adams said, declining to give a specific time period.

Human image animation involves generating videos from a character photo, allowing user control and unlocking the potential for video and movie production. While recent approaches yield impressive results using high-quality training data, the inaccessibility of these datasets hampers fair and transparent benchmarking. Moreover, these approaches prioritize 2D human motion and overlook the significance of camera motions in videos, leading to limited control and unstable video generation. To demystify the training data, we present HumanVid, the first large-scale high-quality dataset tailored for human image animation, which combines crafted real-world and synthetic data. For the real-world data, we compile a vast collection of real-world videos from the internet. We developed and applied careful filtering rules to ensure video quality, resulting in a curated collection of 20K high-resolution (1080P) human-centric videos. Human and camera motion annotation is accomplished using a 2D pose estimator and a SLAM-based method. To expand our synthetic dataset, we collected 10K 3D avatar assets and leveraged existing assets of body shapes, skin textures and clothings. Notably, we introduce a rule-based camera trajectory generation method, enabling the synthetic pipeline to incorporate diverse and precise camera motion annotation, which can rarely be found in real-world data.

Given a graph $G$ and a positive integer $k$, the Graphlet Sampling problem asks to sample a connected induced $k$-vertex subgraph of $G$ uniformly at random.Graphlet sampling enhances machine learning applications by transforming graph structures into feature vectors for tasks such as graph classification and subgraph identification, boosting neural network performance, and supporting clustered federated learning by capturing local structures and relationships.A recent work has shown that the problem admits an algorithm that preprocesses $G$ in time $O(nk^2 \log k + m)$, and draws one sample in expected time $k^{O(k)} \log n$, where $n=|V(G)|$ and $m=|E(G)|$. Such an algorithm relies on the assumption that the input graph fits into main memory and it does not seem to be straightforward to adapt it to very large graphs. We consider Graphlet Sampling in the semi-streaming setting, where we have a memory of $M = \Omega(n \log n)$ words, and $G$ can be only read through sequential passes over the edge list. We develop a semi-streaming algorithm that preprocesses $G$ in $p={O}(\log n)$ passes and samples $\Theta(M k^{-O(k)})$ independent uniform $k$-graphlets in $O(k)$ passes. For constant $k$, both phases run in time $O((n+m)\log n)$. We also show that the tradeoff between memory and number of passes of our algorithms is near-optimal. Our extensive evaluation on very large graphs shows the effectiveness of our algorithms.

Gesture synthesis is a vital realm of human-computer interaction, with wide-ranging applications across various fields like film, robotics, and virtual reality. Recent advancements have utilized the diffusion model to improve gesture synthesis. However, the high computational complexity of these techniques limits the application in reality. In this study, we explore the potential of state space models (SSMs).Direct application of SSMs in gesture synthesis encounters difficulties, which stem primarily from the diverse movement dynamics of various body parts. The generated gestures may also exhibit unnatural jittering issues.To address these, we implement a two-stage modeling strategy with discrete motion priors to enhance the quality of gestures.Built upon the selective scan mechanism, we introduce MambaTalk, which integrates hybrid fusion modules, local and global scans to refine latent space representations.Subjective and objective experiments demonstrate that our method surpasses the performance of state-of-the-art models.