Inductive Process Modeling
Mar 10, 2012
Will Bridewell
Research Scientist in Artificial Intelligence
My research interests include the relationship between attention, cognition, and intentional action.
Publications
Previous research on inductive process modeling, which constructs models from knowledge and time-series data, has relied on handcrafted constraints. In this paper, we report an approach to discovering such constraints from a set of models that have been ranked according to their error on observations.
Ljupčo Todorovski,
Will Bridewell,
Pat Langley
We review the paradigm of inductive process modeling and examine its application to human physiology.
Pat Langley,
Will Bridewell
Chunki Park,
Will Bridewell,
Pat Langley
In the past, inductive process modeling systems have been limited to data sets that recorded change over time, but many interesting problems involve both spatial and temporal dynamics. To meet this challenge, we introduce SCISM, an integrated intelligent system which solves the task of inducing process models that account for spatial and temporal variation.
Chunki Park,
Will Bridewell,
Pat Langley
Research on transfer learning attempts to apply previously acquired knowledge to new learning tasks. Working within this area, we report on a procedure that learns and transfers constraints in the context of inductive process modeling.
Will Bridewell,
Ljupčo Todorovski
We report on a system for inductive process modeling that uses structural constraints to reduce its search through the space of candidate models and to produce ones that human scientists find plausible.
Will Bridewell,
Pat Langley
We report some challenges encountered in developing Prometheus, a software environment that supports the construction and revision of explanatory scientific models. Our responses to these challenges include the use of quantitative processes, to encode models and background knowledge, and the combination of AND/OR search through a space of model structures with gradient descent to estimate parameters.
Will Bridewell,
Stuart R. Borrett,
Pat Langley
In this paper, we motivate the need for a broad curriculum in science informatics and describe the content we are planning to incorporate in our courses.
Pat Langley,
Will Bridewell
We discuss a mechanism for transfer learning in the context of inductive process modeling.
Will Bridewell,
Ljupčo Todorovski
We pose a novel research problem for machine learning that involves constructing a process model from continuous data.
Will Bridewell,
Pat Langley,
Ljupčo Todorovski,
Saso Džeroski
We introduce an approach for extracting constraints on process model construction. Results suggest that the learned constraints make sense ecologically and may provide insight into the nature of the modeled domain.
Will Bridewell,
Stuart R. Borrett,
Ljupčo Todorovski
We developed SC-IPM, an inductive process modeler that uses declarative constraints to reduce the size of the model structure space. In this paper, we describe the constraint formalism and how it guides SC-IPM’s search.
Matt Bravo,
Will Bridewell,
Ljupčo Todorovski
We introduce an inductive logic programming approach to learning declarative bias. Results indicate that the bias reduces the size of the search space without removing the most accurate structures.
Will Bridewell,
Ljupčo Todorovski
We introduce a method for representing process-based models that facilitates the discovery of structures that explain observed behavior. Using this approach, a modeler first encodes relevant ecological knowledge into a library of generic entities and processes, then instantiates these theoretical components, and finally assembles candidate models from these elements.
Stuart R. Borrett,
Will Bridewell,
Pat Langley,
Kevin R. Arrigo
Borrett, S.R., Bridewell, W., Langley, P., & Arrigo, K.R. (2006). Process sensitivity analysis for ecological modeling. The Fifth International Conference on Ecological Informatics. Santa Barbara, CA.
Stuart R. Borrett,
Will Bridewell,
Pat Langley,
Kevin R. Arrigo
We present a language for stating process models and background knowledge in terms familiar to scientists, along with an interactive environment for knowledge discovery that lets the user construct, edit, and visualize scientific models, use them to make predictions, and revise them to better fit available data.
Will Bridewell,
Javier N. Sánchez,
Pat Langley,
Dorrit Billman
We review the task of inductive process modeling, which uses domain knowledge to compose explanatory models of continuous dynamic systems, and investigate methods for learning with missing data. Using experiments with synthetic and natural data, we compare an expectation maximization approach with one that simply ignores missing data.
Will Bridewell,
Pat Langley,
Steve Racunas,
Stuart R. Borrett
Dorrit Billman,
Will Bridewell,
Stuart R. Borrett
Langley, P., & Bridewell, W. (2006). Scientific reasoning and artificial intelligence. (Technical Report) Computational Learning Laboratory, CSLI, Stanford University, CA.
Pat Langley,
Will Bridewell
Borrett, S.R., Bridewell, W., Langley, P., & Arrigo, K.R. (2006). A hierarchical process model of the Ross Sea ecosystem. Eos, Transactions, American Geophysical Union, 87(36), Ocean Sciences Meeting Supplement, Abstract OS43K-06.
Stuart R. Borrett,
Will Bridewell,
Pat Langley,
Kevin R. Arrigo
Borrett, S.R., Bridewell, W., & Langley, P. (2006). Computational discovery of process models of aquatic ecosystems. Ninety-First Annual Meeting of the Ecological Society of America. Memphis, TN.
Stuart R. Borrett,
Will Bridewell,
Pat Langley
We note that previous methods for inductive process modeling tend to overfit the training data, which suggests ensemble learning as a likely response. We introduce a new approach that induces a set of process models from different samples of the training data and uses them to guide a final search through the space of model structures.
Will Bridewell,
Narges Bani Asadi,
Pat Langley,
Ljupčo Todorovski
Research on inductive process modeling combines background knowledge with time-series data to construct explanatory models, but previous work has placed few constraints on search through the model space. We present an extended formalism that organizes process knowledge in a hierarchical manner and a system that carries out constrained search using this knowledge.
Ljupčo Todorovski,
Will Bridewell,
Oren Shiran,
Pat Langley