Arrow Research search

Author name cluster

Randall Davis

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

26 papers
2 author rows

Possible papers

26

IJCAI Conference 2016 Conference Paper

Balancing Appearance and Context in Sketch Interpretation

  • Yale Song
  • Randall Davis
  • Kaichen Ma
  • Dana L. Penney

We describe a sketch interpretation system that detects and classifies clock numerals created by subjects taking the Clock Drawing Test, a clinical tool widely used to screen for cognitive impairments (e. g. , dementia). We describe how it balances appearance and context, and document its performance on some 2, 000 drawings (about 24K clock numerals) produced by a wide spectrum of patients. We calibrate the utility of different forms of context, describing experiments with Conditional Random Fields trained and tested using a variety of features. We identify context that contributes to interpreting otherwise ambiguous or incomprehensible strokes. We describe ST-slices, a novel representation that enables "unpeeling" the layers of ink that result when people overwrite, which often produces ink impossible to analyze if only the final drawing is examined. We characterize when ST-slices work, calibrate their impact on performance, and consider their breadth of applicability.

PDF Details

IJCAI Conference 2015 Conference Paper

Continuous Body and Hand Gesture Recognition for Natural Human-Computer Interaction: Extended Abstract

  • Yale Song
  • Randall Davis

We present a new approach to gesture recognition that tracks body and hands simultaneously and recognizes gestures continuously from an unsegmented and unbounded input stream. Our system estimates 3D coordinates of upper body joints and classifies the appearance of hands into a set of canonical shapes. A novel multi-layered filtering technique with a temporal sliding window is developed to enable online sequence labeling and segmentation. Experimental results on the NATOPS dataset show the effectiveness of the approach. We also report on our recent work on multimodal gesture recognition and deep-hierarchical sequence representation learning that achieve the state-ofthe-art performances on several real-world datasets.

PDF Details

IJCAI Conference 2013 Conference Paper

One-Class Conditional Random Fields for Sequential Anomaly Detection

  • Yale Song
  • Zhen Wen
  • Ching-Yung Lin
  • Randall Davis

Sequential anomaly detection is a challenging problem due to the one-class nature of the data (i. e. , data is collected from only one class) and the temporal dependence in sequential data. We present One-Class Conditional Random Fields (OCCRF) for sequential anomaly detection that learn from a one-class dataset and capture the temporal dependence structure, in an unsupervised fashion. We propose a hinge loss in a regularized risk minimization framework that maximizes the margin between each sequence being classified as “normal” and “abnormal. ” This allows our model to accept most (but not all) of the training data as normal, yet keeps the solution space tight. Experimental results on a number of real-world datasets show our model outperforming several baselines. We also report an exploratory study on detecting abnormal organizational behavior in enterprise social networks.

PDF Details DOI

ICRA Conference 2010 Conference Paper

A voice-commandable robotic forklift working alongside humans in minimally-prepared outdoor environments

  • Seth J. Teller
  • Matthew R. Walter
  • Matthew E. Antone
  • Andrew Correa
  • Randall Davis
  • Luke Fletcher
  • Emilio Frazzoli
  • James R. Glass

One long-standing challenge in robotics is the realization of mobile autonomous robots able to operate safely in existing human workplaces in a way that their presence is accepted by the human occupants. We describe the development of a multi-ton robotic forklift intended to operate alongside human personnel, handling palletized materials within existing, busy, semi-structured outdoor storage facilities.

Details

IJCAI Conference 2009 Conference Paper

  • David Tyler Bischel
  • Thomas Stahovich
  • Eric Peterson
  • Randall Davis
  • Aaron Adler

Mechanical design tools would be considerably more useful if we could interact with them in the way that human designers communicate design ideas to one another, i. e. , using crude sketches and informal speech. Those crude sketches frequently contain pen strokes of two different sorts, one type portraying device structure, the other denoting gestures, such as arrows used to indicate motion. We report here on techniques we developed that use information from both sketch and speech to distinguish gesture strokes from non-gestures — a critical first step in understanding a sketch of a device. We collected and analyzed unconstrained device descriptions, which revealed six common types of gestures. Guided by this knowledge, we developed a classifier that uses both sketch and speech features to distinguish gesture strokes from nongestures. Experiments with our techniques indicate that the sketch and speech modalities alone produce equivalent classification accuracy, but combining them produces higher accuracy.

PDF Details

IJCAI Conference 2009 Conference Paper

  • Tom Y. Ouyang
  • Randall Davis

There is increasing interest in building systems that can automatically interpret hand-drawn sketches. However, many challenges remain in terms of recognition accuracy, robustness to different drawing styles, and ability to generalize across multiple domains. To address these challenges, we propose a new approach to sketched symbol recognition that focuses on the visual appearance of the symbols. This allows us to better handle the range of visual and stroke-level variations found in freehand drawings. We also present a new symbol classifier that is computationally efficient and invariant to rotation and local deformations. We show that our method exceeds state-of-the-art performance on all three domains we evaluated, including handwritten digits, PowerPoint shapes, and electrical circuit symbols.

PDF Details

NeurIPS Conference 2009 Conference Paper

Learning from Neighboring Strokes: Combining Appearance and Context for Multi-Domain Sketch Recognition

  • Tom Ouyang
  • Randall Davis

We propose a new sketch recognition framework that combines a rich representation of low level visual appearance with a graphical model for capturing high level relationships between symbols. This joint model of appearance and context allows our framework to be less sensitive to noise and drawing variations, improving accuracy and robustness. The result is a recognizer that is better able to handle the wide range of drawing styles found in messy freehand sketches. We evaluate our work on two real-world domains, molecular diagrams and electrical circuit diagrams, and show that our combined approach significantly improves recognition performance.

PDF Details

AAAI Conference 2004 Conference Paper

Automatically Transforming Symbolic Shape Descriptions for Use in Sketch Recognition

  • Tracy Hammond
  • Randall Davis

Sketch recognition systems are currently being developed for many domains, but can be time consuming to build if they are to handle the intricacies of each domain. This paper presents the first translator that takes symbolic shape descriptions (written in the LADDER sketch language) and automatically transforms them into shape recognizers, editing recognizers, and shape exhibitors for use in conjunction with a domain independent sketch recognition system. This transformation allows us to build a single domain independent recognition system that can be customized for multiple domains. We have tested our framework by writing several domain descriptions and automatically created a domain specific sketch recognition system for each domain.

PDF Details

AAAI Conference 2004 Conference Paper

Perceptually Based Learning of Shape Descriptions for Sketch Recognition

  • Olya Veselova
  • Randall Davis

We are interested in enabling a generic sketch recognition system that would allow more natural interaction with design tools in various domains, such as mechanical engineering, military planning, logic design, etc. We would like to teach the system the symbols for a particular domain by simply drawing an example of each one – as easy as it is to teach a person. Studies in cognitive science suggest that, when shown a symbol, people attend preferentially to certain geometric features. Relying on such biases, we built a system capable of learning descriptions of hand-drawn symbols from a single example. The generalization power is derived from a qualitative vocabulary reflecting human perceptual categories and a focus on perceptually relevant global properties of the symbol. Our user study shows that the system agrees with the subjects’ majority classification about as often as any individual subject did.

PDF Details

IJCAI Conference 2003 Conference Paper

LADDER: A Language to Describe Drawing, Display, and Editing in Sketch Recognition

  • Tracy Hammond
  • Randall Davis

We have created LADDER, the first language to describe how sketched diagrams in a domain are drawn, displayed, and edited. The difficulty in creating such a language is choosing a set of predefined entities that is broad enough to support a wide range of domains, while remaining narrow enough to be comprehensible. The language consists of predefined shapes, constraints, editing behaviors, and display methods, as well as a syntax for specifying a domain description sketch grammar and extending the language, ensuring that shapes and shape groups from many domains can be described. The language allows shapes to be built hierarchically (e. g. , an arrow is built out of three lines), and includes the concept of "abstract shapes", analogous to abstract classes in an object oriented language. Shape groups describe how multiple domain shapes interact and can provide the sketch recognition system with information to be used in top-down recognition. Shape groups can also be used to describe "chain-reaction" editing commands that effect multiple shapes at once. To test that recognition is feasible using this language, we have built a simple domain-independent sketch recognition system that parses the domain descriptions and generates the code necessary to recognize the shapes.

PDF

AIJ Journal 2000 Journal Article

Qualitative rigid-body mechanics

  • Thomas F. Stahovich
  • Randall Davis
  • Howard Shrobe

We present a theory of qualitative rigid-body mechanics and describe a program that uses this theory to compute qualitative rigid-body dynamic simulations. The program works directly from a qualitative representation of geometry—qualitative configuration space—without need of any metric information. The program can handle devices that are composed of an arbitrary number of fixed-axis components and springs, with driving inputs coming from both applied motions and forces. The program employs quasi-static assumptions, i. e. , inertia-free motion, inelastic collisions, and frictionless contacts. It employs a new qualitative representation for forces that reduces ambiguity in force sums and hence reduces unnecessary branching.

Details DOI

AIJ Journal 1998 Journal Article

Generating multiple new designs from a sketch

  • Thomas F. Stahovich
  • Randall Davis
  • Howard Shrobe

We describe a program called SketchIT that transforms a single sketch of a mechanical device into multiple families of new designs. It represents each of these families with a “BEP-Model”, a parametric model augmented with constraints that ensure the device produces the desired behavior. The program is based on qualitative configuration space (qc-space), a novel representation that captures mechanical behavior while abstracting away its implementation. The program employs a paradigm of abstraction and resynthesis: it abstracts the initial sketch into qc-space, then uses a library of primitive mechanical interactions to map from qc-space to new implementations.

Details DOI

AAAI Conference 1993 Conference Paper

Multiple Dimensions of Generalization In Model-Based Troubleshooting

  • Randall Davis

Two observations motivate our work: (a) modelbased diagnosis programs are powerful but do not learn from experience, and (b) one of the longterm trends in learning research has been the increasing use of knowledge to guide and inform the process of induction. We have developed a knowledge-guided learning method, based in EBL, that allows a model-based diagnosis program to selectively accumulate and generalize its experience. Our work is novel in part because it produces several different kinds of generalizations from a single example. Where previous work in learning has for the most part intensively explored one or another specific kind of generalization, our work has focused on accumulating and using multiple different grounds for generalization, i. e. , multiple domain theories. As a result our system not only learns from a single example (as in all EBL), it can learn multiple things from a single example. Simply saying there ought to be multiple grounds for generalization only opens up the possibility of exploring more than one domain theory. We provide some guidance in determining which grounds to explore by demonstrating that in the domain of physical devices, causal models are a rich source of useful domain theories. We also caution that adding more knowledge can sometimes degrade performance. Hence we need to select the grounds for generalization carefully and analyze the resulting rules to ensure that they improve performance. We illustrate one such quantitative analysis in the context of a model-based troubleshooting program, measuring and analyzing the gain resulting from the generalizations produced.

PDF Details

AAAI Conference 1983 Conference Paper

Diagnosis Via Causal Reasoning: Paths of Interaction and the Locality Principle

  • Randall Davis

Interest has grown recently in developing expert systems that reason "from first principles", i.e., capable of the kind of problem solving exhibited by an engineer who can diagnose a malfunctioning device by reference to its schematics, even though he may never have seen that device before. In developing such a system for troubleshooting digital electronics, we have argued for the importance of pathways of causal interaction as a key concept. We have also suggested using a layered set of interaction paths as a way of constraining and guiding the diagnostic process. We report here on the implementation and use of these ideas. We show how they make it possible for our system to generate a few sharply constrained hypotheses in diagnosing a bridge fault. Abstracting from this example, we find a number of interesting general principles at work. We suggest that diagnosis can be viewed as the interaction of simulation and inference and we find that the concept of locality proves to be extremely useful in understanding why bridge faults are difficult to diagnose and why multiple representations are useful.

PDF Details

AAAI Conference 1982 Conference Paper

Diagnosis Based on Description of Structure and Function

  • Randall Davis
  • Walter Hamscher
  • Mark Shirley

While expert systems have traditionally been built using large coliections of rules based on empirlcal associations, interest has grown recently in the use of systems that reason from representations of structure and function. Our work explores the use of such models in troubleshooting digital electronics. We describe our work to date on (i) a language for describing structure, (ii) a language for describing function, and (i/i) a set of prlnctples for troubleshooting that uses the two descriptions to guide its investigation. In discussing troubleshooting we show why the traditional approach --- test generation --- solves a different [JrdJklll dnti vve &SCllSS a Ilumber of its pIdC, hd ShOrt~Olllill~S. We consider next the style of debugging known as violated expectations and demonstrate why it is a fundclmental advance over traditional test generation. Further exploration of this approach. however, demonstrates that it is incapable of dealing with commonly known classes of faults. We explain the shortcoming as arisirlg from the use of a fault model that is both implicit and inseparable from the basic troubleshooting metl~odology. We argue for the importance of fault models that are explicit, separated from the troubleshooting mechanism, and retractable in much the same sense that inferences are retracted in current systems.

PDF Details