Many students rely on examples when learning to program, but they often face barriers when incorporating these examples into their own code and learning the concepts they present. As a step towards designing effective example interfaces that can support student learning, we investigate novices' needs and strategies when using examples to write code. We conducted a study with 12 pairs of high school students working on open-ended game design projects, using a system that allows students to browse examples based on their functionality, and to view and copy the example code. We analyzed interviews, screen recordings, and log data, identifying 5 moments when novices request examples, and 4 strategies that arise when students use examples. We synthesize these findings into principles that can inform the design of future example systems to better support students.

This study presents the concept testing results of a low-fidelity user-centered design (UCD) tool applied to the U.S. military acquisition request process. A purposive, random, small-sampled population of United States Marines used a UCD-based Design Worksheet to aid their own design process as they prepared a simulated acquisition request for equipment modifications. The results indicated that Marines engaged in the robust discovery and design exploration while using the Design Worksheet. While Marines had some challenges with understanding how to navigate the Design Worksheet, they found it valuable as a prompt for a useful, more productive thinking process. Importantly, Marines indicated that they would adopt the Design Worksheet to complete a real acquisition design request. Considerations for how the Design Worksheet supported their equipment modification analysis, and whether similar HCI or UCD tools could be pursued and adopted in comparable design environments are discussed.

Project-based learning can encourage and motivate students to learn through exploring their own interests, but introduces special challenges for novice programmers. Recent research has shown that novice students perceive themselves to be "bad at programming, especially when they do not know how to start writing a program, or need to create a plan before getting started. In this paper, we present PlanIT, a guided planning tool integrated with the Snap! programming environment designed to help novices plan and program their open-ended projects. Within PlanIT, students can add a description for their project, use a to do list to help break down the steps of implementation, plan important elements of their program including actors, variables, and events, and view related example projects. We report findings from a pilot study of high school students using PlanIT, showing that students who used the tool learned to make more specific and actionable plans. Results from student interviews show they appreciate the guidance that PlanIT provides, as well as the affordances it offers to more quickly create program elements.

Batch Normalization (BN) is a well-known technique used in training deep neural networks. The main idea behind batch normalization is to normalize the features of the layers (i.e., transforming them to have a mean equal to zero and a variance equal to one). Such a procedure encourages the optimization landscape of the loss function to be smoother, and improves the learning of the networks for both speed and performance. In this paper, we demonstrate that the performance of the network can be improved, if the distributions of the features of the output in the same layer are similar. As normalizing based on mean and variance does not necessarily make the features to have the same distribution, we propose a new normalization scheme: Batch Normalization with Skewness Reduction (BNSR). Comparing with other normalization approaches, BNSR transforms not just only the mean and variance, but also the skewness of the data. By tackling this property of a distribution, we are able to make the output distributions of the layers to be further similar. The nonlinearity of BNSR may further improve the expressiveness of the underlying network. Comparisons with other normalization schemes are tested on the CIFAR-100 and ImageNet datasets. Experimental results show that the proposed approach can outperform other state-of-the-arts that are not equipped with BNSR.