Search for: All records

The ground state electron density — obtainable using Kohn-Sham Density Functional Theory (KS-DFT) simulations — contains a wealth of material information, making its prediction via machine learning (ML) models attractive. However, the computational expense of KS-DFT scales cubically with system size which tends to stymie training data generation, making it difficult to develop quantifiably accurate ML models that are applicable across many scales and system configurations. Here, we address this fundamental challenge by employing transfer learning to leverage the multi-scale nature of the training data, while comprehensively sampling system configurations using thermalization. Our ML models are less reliant on heuristics, and being based on Bayesian neural networks, enable uncertainty quantification. We show that our models incur significantly lower data generation costs while allowing confident — and when verifiable, accurate — predictions for a wide variety of bulk systems well beyond training, including systems with defects, different alloy compositions, and at multi-million-atom scales. Moreover, such predictions can be carried out using only modest computational resources.

 

Voice conversion (VC) aims at altering a person's voice to make it sound similar to the voice of another person while preserving linguistic content. Existing methods suffer from a dilemma between content intelligibility and speaker similarity; i.e., methods with higher intelligibility usually have a lower speaker similarity, while methods with higher speaker similarity usually require plenty of target speaker voice data to achieve high intelligibility. In this work, we propose a novel method Phoneme Hallucinator that achieves the best of both worlds. Phoneme Hallucinator is a one-shot VC model; it adopts a novel model to hallucinate diversified and high-fidelity target speaker phonemes based just on a short target speaker voice (e.g. 3 seconds). The hallucinated phonemes are then exploited to perform neighbor-based voice conversion. Our model is a text-free, any-to-any VC model that requires no text annotations and supports conversion to any unseen speaker. Quantitative and qualitative evaluations show that Phoneme Hallucinator outperforms existing VC methods for both intelligibility and speaker similarity.

 

This paper describes the design of an online learning platform that empowers musical creation and performance with Python code. For this platform we have developed an innovative computational note- book paradigm that we call TunePad playbooks. While playbooks borrow ideas from popular computational notebooks like Jupyter, we have designed them from the ground up to support creative mu- sical expression including live performances. After discussing our design principles and features, we share findings from a series of artifact-centered interviews conducted with experienced TunePad users. Our results show how systems like ours might flexibly sup- port a variety of creative workflows, while suggesting opportunities for future work in this area. 

Over the last decade, large multitouch displays have become commonplace in museums and other public spaces. While there is preliminary evidence that exhibits based on tangible technologies can be more attractive and engaging for visitors than displays alone, very little empirical research has directly compared tangible to large multitouch displays in museums. In this paper, we present a study comparing the use of a tangible and a multitouch tabletop interface in an exhibit designed to explore musical rhythms. From an observation pool of 791 museum visitors, a total of 227 people in 82 groups interacted with one of the two versions of our exhibit. We share the exhibit design, experimental setup, and methods and measures. Our findings highlight advantages of tangible interaction in terms of attracting and engaging children and families. However, the two exhibits were equally effective at supporting collaborative interaction within visitor groups. We conclude with a discussion of the implications for museum exhibit design vis-à-vis visitor engagement and learning. 

TunePad is a free, online platform designed with the goal of empowering diverse communities of learners to create and share music through code. We are interested in the idea of music as a pervasive form of literacy with abundant connections to concepts of computer programming. Over the past three years we have developed and refined successive prototypes with over 500 middle school and high school students in a variety of learning spaces including schools, libraries, summer camps, and other out-of-school programs. This paper shares the current TunePad design along with data from three summer camps for middle school students that involved daily work with the platform. Through these camps we saw significant gains in learners’ attitudes around computer programming as measured through pre-post surveys. We also share a theoretical perspective on music and coding as an intersection of literacies that we reflect on through student-created artifacts.