Search for: All records

Socially interactive robots present numerous unique programming challenges for interaction developers. While modern authoring tools succeed at making the authoring experience approachable and convenient for developers from a wide variety of backgrounds, they are less successful at targeting assistance to developers based on the specific task or interaction being authored. We propose interaction templates, a data-driven solution for (1) matching in-progress robot programs to candidate task or interaction models and then (2) providing assistance to developers by using the matched models to generate modifications to in-progress programs. In this paper, we present the various dimensions that define first how interaction templates might be used, then how interaction templates may be represented, and finally how they might be collected.

To verify safety and robustness of neural networks, researchers have successfully applied abstract interpretation , primarily using the interval abstract domain. In this paper, we study the theoretical power and limits of the interval domain for neural-network verification. First, we introduce the interval universal approximation (IUA) theorem. IUA shows that neural networks not only can approximate any continuous function f (universal approximation) as we have known for decades, but we can find a neural network, using any well-behaved activation function, whose interval bounds are an arbitrarily close approximation of the set semantics of f (the result of applying f to a set of inputs). We call this notion of approximation interval approximation . Our theorem generalizes the recent result of Baader et al. from ReLUs to a rich class of activation functions that we call squashable functions . Additionally, the IUA theorem implies that we can always construct provably robust neural networks under ℓ ∞ -norm using almost any practical activation function. Second, we study the computational complexity of constructing neural networks that are amenable to precise interval analysis. This is a crucial question, as our constructive proof of IUA is exponential in the size of the approximation domain. Wemore »boil this question down to the problem of approximating the range of a neural network with squashable activation functions. We show that the range approximation problem (RA) is a Δ 2 -intermediate problem, which is strictly harder than NP -complete problems, assuming coNP ⊄ NP . As a result, IUA is an inherently hard problem : No matter what abstract domain or computational tools we consider to achieve interval approximation, there is no efficient construction of such a universal approximator. This implies that it is hard to construct a provably robust network, even if we have a robust network to start with.« less

Differential privacy is a formal, mathematical def- inition of data privacy that has gained traction in academia, industry, and government. The task of correctly constructing differentially private algorithms is non-trivial, and mistakes have been made in foundational algorithms. Currently, there is no automated support for converting an existing, non-private program into a differentially private version. In this paper, we propose a technique for automatically learning an accurate and differentially private version of a given non-private program. We show how to solve this difficult program synthesis problem via a combination of techniques: carefully picking representative example inputs, reducing the problem to continuous optimization, and mapping the results back to symbolic expressions. We demonstrate that our approach is able to learn foundational al- gorithms from the differential privacy literature and significantly outperforms natural program synthesis baselines.

When a model makes a consequential decision, e.g., denying someone a loan, it needs to additionally generate actionable, realistic feedback on what the person can do to favorably change the decision. We cast this problem through the lens of program synthesis, in which our goal is to synthesize an optimal (realistically cheapest or simplest) sequence of actions that if a person executes successfully can change their classification. We present a novel and general approach that combines search-based program synthesis and test-time adversarial attacks to construct action sequences over a domain-specific set of actions. We demonstrate the effectiveness of our approach on a number of deep neural networks.

Social robots have varied effectiveness when interacting with humans in different interaction contexts. A robot programmed to escort individuals to a different location, for instance, may behave more appropriately in a crowded airport than a quiet library, or vice versa. To address these issues, we exploit ideas from program synthesis and propose an approach to transforming the structure of hand-crafted interaction programs that uses user-scored execution traces as input, in which end users score their paths through the interaction based on their experience. Additionally, our approach guarantees that transformations to a program will not violate task and social expectations that must be maintained across contexts. We evaluated our approach by adapting a robot program to both real-world and simulated contexts and found evidence that making informed edits to the robot's program improves user experience.