Search for: computing

Deep learning’s performance has been extensively recognized recently. Graph neural networks (GNNs) are designed to deal with graph-structural data that classical deep learning does not easily manage. Since most GNNs were created using distinct theories, direct comparisons are impossible. Prior research has primarily concentrated on categorizing existing models, with little attention paid to their intrinsic connections. The purpose of this study is to establish a unified framework that integrates GNNs based on spectral graph and approximation theory. The framework incorporates a strong integration between spatial- and spectral-based GNNs while tightly associating approaches that exist within each respective domain.

 

Recent spectral graph sparsificationresearch aims to construct ultra-sparse subgraphs for preserving the original graph spectral (structural) properties, such as the first few Laplacian eigenvalues and eigenvectors, which has led to the development of a variety of nearly linear time numerical and graph algorithms. However, there is very limited progress in the spectral sparsification of directed graphs. In this work, we prove the existence of nearly linear-sized spectral sparsifiers for directed graphs under certain conditions. Furthermore, we introduce a practically efficient spectral algorithm (diGRASS) for sparsifying real-world, large-scale directed graphs leveraging spectral matrix perturbation analysis. The proposed method has been evaluated using a variety of directed graphs obtained from real-world applications, showing promising results for solving directed graph Laplacians, spectral partitioning of directed graphs, and approximately computing (personalized) PageRank vectors.

 

Over the academic year 2022–23, we discussed the teaching of software performance engineering with more than a dozen faculty across North America and beyond. Our outreach was centered on research-focused faculty with an existing interest in this course material. These discussions revealed an enthusiasm for making software performance engineering a more prominent part of a curriculum for computer scientists and engineers. Here, we discuss how MIT’s longstanding efforts in this area may serve as a launching point for community development of a software performance engineering curriculum, challenges in and solutions for providing the necessary infrastructure to universities, and future directions. 

Current deep-learning techniques for processing sets are limited to a fixed cardinality, causing a steep increase in computational complexity when the set is large. To address this, we have taken techniques used to model long-term dependencies from natural language processing and combined them with the permutation equivariant architecture, Set Transformer (STr). The result is Set Transformer XL (STrXL), a novel deep learning model capable of extending to sets of arbitrary cardinality given fixed computing resources. STrXL’s extension capability lies in its recurrent architecture. Rather than processing the entire set at once, STrXL processes only a portion of the set at a time and uses a memory mechanism to provide additional input from the past. STrXL is particularly applicable to processing sets of high-throughput sequencing (HTS) samples of DNA sequences as their set sizes can range into hundreds of thousands. When tasked with classifying HTS prairie soil samples and MNIST digits, results show that STrXL exhibits an expected memory size-accuracy trade-off that scales proportionally with the complexity of downstream tasks, but, unlike STr, is capable of generalizing to sets of arbitrary cardinality. 

Current algorithmic fairness tools focus on auditing completed models, neglecting the potential downstream impacts of iterative decisions about cleaning data and training machine learning models. In response, we developed Retrograde, a JupyterLab environment extension for Python that generates real-time, contextual notifications for data scientists about decisions they are making regarding protected classes, proxy variables, missing data, and demographic differences in model performance. Our novel framework uses automated code analysis to trace data provenance in JupyterLab, enabling these notifications. In a between-subjects online experiment, 51 data scientists constructed loan-decision models with Retrograde providing notifications continuously throughout the process, only at the end, or never. Retrograde’s notifications successfully nudged participants to account for missing data, avoid using protected classes as predictors, minimize demographic differences in model performance, and exhibit healthy skepticism about their models. 

Human intention prediction plays a critical role in human–robot collaboration, as it helps robots improve efficiency and safety by accurately anticipating human intentions and proactively assisting with tasks. While current applications often focus on predicting intent once human action is completed, recognizing human intent in advance has received less attention. This study aims to equip robots with the capability to forecast human intent before completing an action, i.e., early intent prediction. To achieve this objective, we first extract features from human motion trajectories by analyzing changes in human joint distances. These features are then utilized in a Hidden Markov Model (HMM) to determine the state transition times from uncertain intent to certain intent. Second, we propose two models including a Transformer and a Bi-LSTM for classifying motion intentions. Then, we design a human–robot collaboration experiment in which the operator reaches multiple targets while the robot moves continuously following a predetermined path. The data collected through the experiment were divided into two groups: full-length data and partial data before state transitions detected by the HMM. Finally, the effectiveness of the suggested framework for predicting intentions is assessed using two different datasets, particularly in a scenario when motion trajectories are similar but underlying intentions vary. The results indicate that using partial data prior to the motion completion yields better accuracy compared to using full-length data. Specifically, the transformer model exhibits a 2% improvement in accuracy, while the Bi-LSTM model demonstrates a 6% increase in accuracy.