Projects

ICSI hosts basic, pre-competitive research of fundamental importance to computer science and engineering. Projects are chosen based on the interests of the Institute’s principal investigators and the strengths of its researchers and affiliated UC Berkeley faculty.

Recent projects are listed below; the full list of each group's projects is accessible via the links listed in the sidebar.

Resilient Dynamic Autoencoders for Modeling and Predicting Earthquake Threats

Large earthquakes generate strong ground motions and tsunamis that may lead to a significant number of casualties and cause severe impacts on social resilience in seismically active regions including the West Coast of the United States. Early warning systems have been developed to mitigate immediate threats by detecting first-arriving ground motions near an earthquake epicenter and forecasting the intensity and timing of strong destructive ground motions. To further improve the efficacy and accuracy of these systems, deep learning methods have strong potential, but it is crucial to significantly extend the forecast horizons of existing models.

RDL
Creating An Extensible Internet Through Interposition

In the past decade, the construction of Internet-scale distributed applications has been a battle between the immovable object of the Internet architecture and the irresistible force of the need for better application performance and control.

Extensible Internet
Liquid Wireless Networking for Real-Time Data-Intensive Rural Applications

Liquid Wireless NetworkingRural broadband is a foundation for rural economy and quality of life, and many rural applications require real-time data-intensive communications. To address the rural broadband challenge, wireless networks are essential building blocks. However, rural wireless is subject to environmental factors such as weather, terrain, foliage, and crop types and densities, and there exist complex network dynamics and uncertainties (e.g., spatiotemporal uncertainties of wireless links).

Research Initiatives, TCS
DASS: Developer Implementation of Privacy in Software Systems

DASSRecent years have seen a surge in privacy regulations across the globe. The main objective of these regulations is to protect user data and users’ rights by providing guidelines for organizations to follow. The assumption is that such guidelines will provide developers with a clear and concise framework for writing privacy-conscious code. However, even after the introduction of these regulatory frameworks, society continues to experience blatant violations of user privacy.

Usable Security and Privacy
Real-Time Data Reduction Codesign at the Extreme Edge for Science

Real-Time Data Reduction CodesignThis project focuses on intelligent ML-based data reduction and processing as close as possible to the data source. Per sensor compression and efficient aggregation of information while preserving scientific fidelity can have a huge impact on data rates further downstream and the way that experiments are designed and operated. The research team is concentrating on powerful, specialized compute hardware at the extreme edge—such as FPGAs, ASICs, and systems-on-chip—which are typical initial processing layers of many experiments.

Big Data
An Extensible Internet for Science Applications and Beyond

extensible internet reseaechThere are many science experiments that generate huge amounts of data, ranging from terabytes to petabytes and beyond. Some of these science applications require their data to be processed in complex workflows that involve different functions (data generation, analysis, and storage) and span multiple collaborating sites. While progress has been made in addressing these needs – particularly the Data Transfer Nodes (DTNs) developed by ESnet – more work is needed to provide the detailed management and specific data-handling functionality these complex workflows require.

Extensible Internet
Previous Research: Representation Learning

We study learning mid-level representation from natural non-curated data to achieve efficient and generalizing performance on downstream visual tasks such as recognition, segmentation, and detection.  We exploit instance discrimination, instance grouping, model bias and variance analysis, pixel-to-segment contrastive learning, and visual memory to handle open-set recognition, long-tail distribution, open compound domain adaptation, unsupervised or weakly supervised recognition and segmentation.

Vision
Previous Research: Complex-valued Deep Learning

Complex-valued data is ubiquitous in physics and signal processing applications, and complex-valued representations in deep learning have appealing theoretical properties. While these aspects have long been recognized, complex-valued deep learning lags far behind its real-valued counterpart.  Existing methods ignore the rich geometry of complex-valued data, instead opting to use the same techniques and architectures as real-valued data, with undesirable consequences such as decreased robustness, larger model sizes, and poor generalization.

Vision
Previous Research: Sound and Vision Integration

Sound carries complementary information to vision and can help scene understanding and navigation. We train a model to tell individual sounds apart without using labels, which can be used to accelerate subsequent training on supervised sound event classification, and to explain how song birds such as zebra finch can develop communication without any external supervision.  We also demonstrate with a low-cost real system that learns echolocation and generates depth images only from sound.

Vision
Previous Research: Implicit Deep Learning

Most modern neural networks are defined explicitly as a sequence of layers with various connections.  Any desired property such as translational equivariance needs to be hard-coded into the architecture, which is inflexible and restrictive.  In contrast, implicit models are defined as a set of constraints to satisfy or criteria to optimize at the test time.  This framework can help express a large class of operations such as test-time optimization, planning, dynamics, constraints, and feedback.  Our research explores implicit models to integrate invariance and equivariance constraints in co

Vision
Previous Research: Sketch Recognition and Photo Synthesis

Sketches are rapidly executed freehand drawings that make an intuitive and powerful visual expression.  While they lack visual details and have spatial/geometrical distortions, humans can effortlessly envision objects from sketches.  We study translations between sketches, photos, and 3D models at both the object level and the scene level.

Vision
Previous Research: 3D Point Cloud Parsing

Deep neural networks are widely used for understanding 3D point clouds. At each point convolution layer, features are computed from local neighborhoods of 3D points and combined for subsequent processing in order to extract semantic information. We study a novel approach to learn different non-rigid transformations of the input point cloud so that optimal local neighborhoods can be adopted at each layer. 

Vision
Previous Research: Model Learning and Compression

Traditional deep neural network learning seeks an optimal model in a large model space.  However, the optimal model ends up with a lot of redundancy which is then removed during model compression.  We seek an optimal model in a reduced model space without jeopardizing optimality.  We study several techniques such as tied block convolution (TBC), light-cost regularizer (OCNN), and recurrent parameter generator (RPG) where smaller and leaner models are optimized and can be deployed directly with more robustness and generalizability.

Vision
Previous Research: Vision-Based Reinforcement Learning

Vision-based reinforcement learning (RL) is successful, but how to generalize it to unknown test environments remains challenging.  It not only needs to process high-dimensional visual inputs, but it is also required to deal with significant variations in new test scenarios, e.g. color/texture changes or moving distractors.

Vision
Previous Research: Robotic Manipulation and Locomotion

Existing methods for robotic manipulation and locomotion overlook real world constraints such as data availability, data efficiency, and data quality.  We explore novel approaches that incorporate curriculum learning, latent space information extraction, and invariant states to improve the generalizability of learned control policies against environmental and robot configurational changes. 

Vision
Previous Research: Machine Learning for Medical Applications

We explore unsupervised representation learning to not only reduce labeling bias inherent in supervised image-based medical diagnosis, but also allow data-driven discovery of novel pathological, physiological, and camera-related domains.  We also explore machine learning to generate realistic healthy and tumor medical scans to study human visual perception in radiology. 

Vision
Previous Research: Machine Learning in the Hyperbolic Space

We study how hyperbolic space can be used to facilitate the formation of hierarchical representations from natural data without any supervision.  We demonstrate that hyperbolic neural networks outperform standard  Euclidean counterparts when their optimization process is improved with a restricted feature space, resulting in higher classification performance, more adversarial robustness, and better out-of-distribution detection capability.

Vision
Previous Research: Adversarial Robustness

We study how to increase both classification accuracy and robustness with image-wise and pixel-wise representation learning, where perceptual organization is incorporated.

Vision
Agent Based Modeling at the Boundary of Law and Software

ICSI researchers are studying how agent-based models (ABMs) of social contexts can improve the design and regulation of accountable software systems. Agent-based modeling is a social scientific research method that involves bottom-up modeling of complex systems and computationally determining their emergent properties by running simulations. The ICSI researchers are using ABMs to model elements of the social and regulatory environment in which a software system operates.

Networking and Security
Scalable linear algebra and neural network theory

While deep learning methods have in no doubt transformed certain applications of machine learning (ML) such as Computer Vision (CV) and Natural Language Processing (NLP), its promised impact on many other areas has yet to be seen. The reason for this is the flip side of why it has been successful where it has.

Big Data

Pages