Michael Park

Michael Park, Erin Leahey, Russell J. Funk

We respond to Holst et al.'s critique that the decline in scientific disruptiveness documented in Park et al. (Nature, 2023) is an artifact of including works with zero backward citations. Using their advocated dataset, metric, and exclusion criteria, we find declines equivalent to major benchmark transformations in science. Their own regression model--designed to address their concerns about zero-citation works--yields large and significant declines for both papers and patents (p<0.001), a result found in their supplementary tables yet left unaddressed, despite directly contradicting their central claim. Their critique is further undermined by severe quality issues in their data, which contain three times more zero-citation works than ours. We trace this excess to their inclusion of at least 2.8 million editorials, obituaries, and comments, 1.5 million books and proceedings, and 254,000 product and artistic reviews--in all, 20% of their sample is non-research content that almost by definition lacks backward citations. Simple keyword searches confirm the problem's severity, identifying among others 456 For Dummies guides, 50 Dr. Seuss and Curious George books, and the Captain Underpants series--all zero-citation entries in their sample. Applying granular document type classification to their data reveals that such non-research content fell from 40% to 8% of their sample between 1945 and 2010--a shift sufficient to generate the decline in zero-citation prevalence they attribute to metadata errors in our study. Standard practice excludes such content to guard against the metadata quality concerns at the center of their critique--concerns their dataset exemplifies rather than addresses. Declining disruptiveness has been documented in nearly 100 studies across multiple databases, metrics, and non-citation-based measures. The weight of evidence does not support an artifact-based explanation.

Tim Lou, Michael Park, Mohammad Ramezanali et al.

In this note we examine the autoregressive generalization of the FNet algorithm, in which self-attention layers from the standard Transformer architecture are substituted with a trivial sparse-uniformsampling procedure based on Fourier transforms. Using the Wikitext-103 benchmark, we demonstratethat FNetAR retains state-of-the-art performance (25.8 ppl) on the task of causal language modelingcompared to a Transformer-XL baseline (24.2 ppl) with only half the number self-attention layers,thus providing further evidence for the superfluity of deep neural networks with heavily compoundedattention mechanisms. The autoregressive Fourier transform could likely be used for parameterreduction on most Transformer-based time-series prediction models.

Zhigen Zhao, Ziyi Zhou, Michael Park et al.

This study proposes a Task and Motion Planning (TAMP) method with symbolic decisions embedded in a bilevel optimization. This TAMP method exploits the discrete structure of sequential manipulation for long-horizon and versatile tasks in dynamically changing environments. At the symbolic planning level, we propose a scalable decision-making method for long-horizon manipulation tasks using the Planning Domain Definition Language (PDDL) with causal graph decomposition. At the motion planning level, we devise a trajectory optimization (TO) approach based on the Alternating Direction Method of Multipliers (ADMM), suitable for solving constrained, large-scale nonlinear optimization in a distributed manner. Distinct from conventional geometric motion planners, our approach generates highly dynamic manipulation motions by incorporating the full robot and object dynamics. Furthermore, in lieu of a hierarchical planning approach, we solve a holistically integrated bilevel optimization problem involving costs from both the low-level TO and the high-level search. Simulation and experimental results demonstrate dynamic manipulation for long-horizon object sorting tasks in clutter and on a moving conveyor belt.

Michael Park

In this note we explore a fully unsupervised deep-learning framework for simulating non-linear structural equation models from observational training data. The main contribution of this note is an architecture for applying moment-matching loss functions to the edges of a causal Bayesian graph, resulting in a generative conditional-moment-matching graph-neural-network. This framework thus enables automated sampling of latent space conditional probability distributions for various graphical interventions, and is capable of generating out-of-sample interventional probabilities that are often faithful to the ground truth distributions well beyond the range contained in the training set. These methods could in principle be used in conjunction with any existing autoencoder that produces a latent space representation containing causal graph structures.