Parsons problems have become a mainstay of computer science education. They are heavily used among students, especially in K-12 and provide a small puzzle-like experience for students to practice their skills. Today, while prior work has explored complex issues with accessibility and block languages in general, the 2024 changes to accessibility regulations by the U.S. Department of Justice includes new rules around mobile platforms. These rules are ill-defined and in need of evaluation. In this work, we make several contributions. First, we conducted an evaluation of existing blocks with respect to their regulatory compliance and discuss a new blocks technology that we developed that meets these new mobile guidelines. Second, we conducted three empirical studies using Parsons problems to evaluate the usability of the technology with teachers of the visually impaired (n = 32), high-school students with diverse disabilities (n = 28), and high-school students with blindness or low vision (n = 13).

Touchscreen-based rendering of graphics using vibrations, sonification, and text-to-speech is a promising approach for nonvisual access to graphical information, but extracting trends from complex data representations nonvisually is challenging. This work presents the design of a multimodal feedback scheme with integrated spatial audio for the exploration of histograms and scatter plots on touchscreens. We detail the hardware employed and the algorithms used to control vibrations and sonification adjustments through the change of pitch and directional stereo output. We conducted formative testing with 5 blind or visually impaired participants, and results illustrate that spatial audio has the potential to increase the identification of trends in the data, at the expense of a skewed mental representation of the graph. This design work and pilot study are critical to the iterative, human-centered approach of rendering multimodal graphics on touchscreens and contribute a new scheme for efficiently capturing data trends in complex data representations.

Deep Neural Networks (DNNs) process big datasets achieving high accuracy on incredibly complex tasks. However, this progress has led to a scalability impasse, as DNNs require massive amounts of processing power and local memory to be trained, making them impossible or impractical to be used on a single device. This situation has led to the design of distributed training architectures, where the DNN and the training data can be split among multiple processors. How to choose the appropriate distributed training architecture, however, remains an open question. To help bring insights into this debate, in this work we design a Distributed Training Simulator (DTS) that estimates the training time of a DNN in a distributed architecture through a mathematical model of the distributed architecture and resource-allocation heuristics. We illustrate the power of the proposed DTS through the implementation of five different distributed architectures, Pipeline Learning, Federated Learning, Split Learning, Parallel Split Learning, and Federated Split Learning, and we validate the accuracy of the training estimates using three different datasets of varying complexity and two different DNNs. Finally, we present a trade-off analysis to demonstrate the coherence of DTS estimates for diverse high-performance computing scenarios by comparing these estimates with the behaviors of a real computer cluster.