Recent advancements in State Space Models (SSMs) have attracted significant interest, particularly in models optimized for parallel training and handling long-range dependencies. Architectures like Mamba have scaled to billions of parameters with selective SSM. To facilitate broader applications using Mamba, exploring its efficiency is crucial. While token reduction techniques offer a straightforward post-training strategy, we find that applying existing methods directly to SSMs leads to substantial performance drops. Through insightful analysis, we identify the reasons for this failure and the limitations of current techniques. In response, we propose a tailored, unified post-training token reduction method for SSMs. Our approach integrates token importance and similarity, thus taking advantage of both pruning and merging, to devise a fine-grained intra-layer token reduction strategy. Extensive experiments show that our method improves the average accuracy by 5.7% to 13.1% on six benchmarks with Mamba-2 compared to existing methods, while significantly reducing computational demands and memory requirements.

Rethinking Token Reduction for State Space Models

Mobile devices are becoming an important carrier for deep learning tasks, as they are being equipped with powerful, high-end mobile CPUs and GPUs. However, it is still a challenging task to execute 3D Convolutional Neural Networks (CNNs) targeting for real-time performance, besides high inference accuracy. The reason is more complex model structure and higher model dimensionality overwhelm the available computation/storage resources on mobile devices. A natural way may be turning to deep learning weight pruning techniques. However, the direct generalization of existing 2D CNN weight pruning methods to 3D CNNs is not ideal for fully exploiting mobile parallelism while achieving high inference accuracy.

This paper proposes RT3D, a model compression and mobile acceleration framework for 3D CNNs, seamlessly integrating neural network weight pruning and compiler code generation techniques. We propose and investigate two structured sparsity schemes i.e., the vanilla structured sparsity and kernel group structured (KGS) sparsity that are mobile acceleration friendly. The vanilla sparsity removes whole kernel groups, while KGS sparsity is a more fine-grained structured sparsity that enjoys higher flexibility while exploiting full on-device parallelism. We propose a reweighted regularization pruning algorithm to achieve the proposed sparsity schemes. The inference time speedup due to sparsity is approaching the pruning rate of the whole model FLOPs (floating point operations). RT3D demonstrates up to 29.1x speedup in end-to-end inference time comparing with current mobile frameworks supporting 3D CNNs, with moderate 1%~1.5% accuracy loss. The end-to-end inference time for 16 video frames could be within 150 ms, when executing representative C3D and R(2+1)D models on a cellphone. For the first time, real-time execution of 3D CNNs is achieved on off-the-shelf mobiles.

RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices

This presentation covers the overview of our NSF medium project and reports on the recent progress including novel deep neural network pruning and quantization methods recently published.

	https://underline.io/speakers/55839-xue-lin

A Unified Software/Hardware Framework of DNN Computation and Storage Reduction Using ADMM

This poster session includes Main Conference posters and Findings from the following areas:

Low-resource Methods for NLP • Interpretability and Analysis of Models for NLP

In-Person Poster Session D (Jasmine)

poster

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

Main Track

technical paper

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2021

You are welcome and encouraged to view the posters and start asking questions in the Q&A boxes next to each of them.

Break/Poster Presentations 2

 Electronic design and design automation (EDA) is a key enabling technology in the advancement of information processing that we are enjoying today. EDA has provided fundamental design principles and tools empowering the exponential growth of the number of transistors on an integrated circuit chip over the past 50 years. Without EDA, it would not have been possible for billion-transistor chips to be a reality! 

After five decades of growth, the field of EDA now faces some unique challenges. On one hand, we are approaching the physical limit of transistor scaling, and the golden days of exponential growth in transistor density are over. The exorbitant cost of integrated circuit fabrication and the excessive complexity of EDA tools make it difficult to bring design automation innovations to practice. On the other hand, the recent rise of machine learning and artificial intelligence applications and emerging architectures and technologies calls for more customized solutions to designing electronic systems. 

## **Objectives of the Workshop:**

This workshop, funded by the National Science Foundation (NSF), will assemble a diverse group of leading researchers from both academia and industry to identify future research directions in design and design automation of electronic systems and architectures in response to the challenges and opportunities. It will feature plenary talks and panels as well as roundtable discussions. Collectively, the workshop attendees will explore answers to overarching questions such as 

•	what are the high-risk and high-return research topics, 

•	what other fields adjacent to EDA research should be aggressively pursued for highly interdisciplinary collaboration, and

•	where the research funding should come from and how it should be distributed to encourage more transformative research. 

The workshop will also include a satellite event with a roundtable panel discussion on strategies to improve foundry access to US-based EDA researchers. A final report, based on the deliberations of the workshop, will be produced to the government recommending future research directions.


Xue Lin

3

6

SHORT BIO

Presentations

Rethinking Token Reduction for State Space Models

RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices

A Unified Software/Hardware Framework of DNN Computation and Storage Reduction Using ADMM

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES