Colloquium

The evolution of cooperative behavior, which is costly to the donor but beneficial for the recipient, is among the most intriguing questions in evolutionary biology. As presented in the prisoners’ dilemma, cooperation is in many cases advantageous for the group, but disadvantageous for each of the individuals. Thus, according to natural selection we expect cooperative behavior to be eliminated. Yet cooperation is prevalent in nature.

I will discuss Autour, an audio augmented reality app that helps blind people serendipitously discover locations of interest around them while they walk down the street. I will then discuss my ongoing research, MIMIC, a mobile system that gives each member of a couple a visceral, background sense of their partner's current activity through frequent, simple vibration (haptic) patterns on the leg. In both cases, the focus is on moving away from needing to look at a smartphone screen to be aware of useful information. I will discuss some specific issues that designing and testing MIMIC has raised for haptic information delivery, and experiments to address these shortcomings.

Last, if there is interest, I will also touch on an alternative form of haptics, Electrical Muscle Stimulation (EMS), and how it could potentially be used to warp temperature perception in a VR/AR application.

Creative exploration – the search for new and valuable patterns in a space of solutions - is central in the development of art, science and technology. The talk will present two paradigms for studying creative exploration using a computational approach. These paradigms distill complex human behaviors into simple experimental setting, allowing for detailed analysis and modelling, aimed at revealing basic principles underlying creative exploration.

The creative foraging game is a high-resolution experimental paradigm that quantifies creative exploration in a well-defined space of geometric shapes. Individuals explore the space of 10-connected squares (~36k possible shapes), searching for patterns they find beautiful and interesting. Participants alternate between exploration along meandering paths and exploitation of categories of similar shapes. Within a category, but not in exploration, people move along optimal paths. Participants discover new categories through ambiguous shapes that belong to two categories, an experimental proxy for creative leaps. Furthermore, in creative foraging people leave a category of similar shapes far before depleting it, showing that people in creative foraging do not follow predictions of Optimal Foraging Theory.

The second experimental paradigm, the mirror game, is a simplified setup for studying joint improvisation. In joint improvisation, such as Jazz or contact dance, a group of experienced performers explore a space of possibilities through interaction. The setup is based on a common practice from theatre and dance in which two actors create synchronized and dance-like motion together. In the simplified setup two participants mirror each other (with or without a designated leader) by moving handles on two parallel tracks. The mirror game enabled the discovery of a basic mechanism of joint improvisation: the ability of expert pairs to share leadership by mutually predicting each other’s actions (Noy et al., PNAS 2011). 

I will end the talk with some possible follow-up projects which might be of interest to computer science students looking for research projects.

When you hear the term "storage", do you think of a disk? a file? a flash drive? Storage Systems mean much more than that. These are the complex systems that store the world's exabytes of data in the cloud, they are responsible that data will always be available and reliable, and support Big Data Analytics systems.

Triggered by the Data proliferation and Cloud revolution of our era, Storage systems have undergone a major evolution in the way they are designed, built and used. In this talk I will tell the story behind this evolution: from Blocks and Files, to Objects. I will explain what led to it, and will review how cloud object stores are built and used in today's applications.

I'll offer a crash course on quantum computing, which seeks to exploit the strange rules of quantum physics to solve certain problems dramatically faster than we know how to solve them with any existing computer. I promise no hype: just a a sober summary of how a quantum computer would actually work (hint: it's not just by "trying every possible answer in parallel"), for which problems quantum computers are and aren't expected to provide an advantage, and the current status of the worldwide effort to make quantum computing practical—and even more immediately, to achieve the first demonstration of "quantum supremacy," or a clear quantum speedup for some task (which might be a contrived one). I'll also say something about the ultimate physical limits of computation, and about speculative proposals for going beyond even quantum computers.

A dataset with $M$ items has $2^M$ subsets anyone of which may be the one satisfying our objective. With a good data display and interactivity our fantastic pattern-recognition defeats this combinatorial explosion by extracting insights from the visual patterns. This is the core reason for data visualization. With parallel coordinates the search for relations in multivariate data is transformed into a 2-D pattern recognition problem. Together with criteria for good query design, we illustrate this on several real datasets (financial, process control, credit-score, one with hundreds of variables) with stunning results. A geometric classification algorithm yields the classification rule explicitly and visually. The minimal set of variables, features, are found and ordered by their predictive value. A model of a country’s economy reveals sensitivities, impact of constraints, trade-offs and economic sectors unknowingly competing for the same resources. An overview of the methodology provides foundational understanding; learning the patterns corresponding to various multivariate relations. These patterns are robust in the presence of errors and that is good news for the applications. A topology of proximity emerges opening the way for visualization in Big Data.

Recent Internet research has been driven by two facts and their contradictory implications: the current Internet architecture is both inherently flawed (so we should explore radically different alternative designs) and deeply entrenched (so we should restrict ourselves to backwards-compatible and therefore incremental improvements). In this talk, we try to reconcile these two perspectives by first identifying the fundamental omission in the current architecture, which makes architectural evolution hard, and then redressing this omission with a backwards-compatible incrementally deployable architectural framework called Trotsky, in which one can seamlessly deploy radically new designs. We show how this can lead to a permanent revolution in Internet architecture by (i) easing the deployment of new architectures and (ii) allowing multiple coexisting architectures to be used simultaneously by applications. Trotsky thus enables both architectural evolution and diversity.

Joint work with James Murphy McCauley, Barath Raghavan, Brian Kim, Aurojit Panda, and Scott Shenker

The talk will focus on few counter-intuitive results from studies on several different topics, (using different methodologies):

1. Optimal seeding policies in user-generated content networks (UGC) such as YouTube, Facebook and SoundCloud advocate for using high status (e.g., high degree) individuals. However, perhaps surprisingly, under quite common conditions, low status (low degree) individuals facilitate larger reach and more effective dissemination. Previous models ignore the response probability of the individual (seed) at hand. Because response probability is a function of status differences (as will be shown), a seeding policy that prioritizes high status individuals is suboptimal, and perhaps should be revised.
2. We introduce evidence that name stereotypes can be manifested in facial appearance, producing a face-name matching effect, whereby both a social perceiver and a computer (through a machine learning based classifier) are able to accurately match a person’s name to his or her face. In nine studies (not all of which will be presented), we demonstrate this effect, showing that participants examining an unfamiliar face accurately select the person’s true name from a list of several names, significantly above chance level. We replicate the effect in two countries and find that it extends beyond the limits of socioeconomic cues, and even among fraternal twins. We also find the effect using a computer-based paradigm and 94,000 faces. A self-fulfilling prophecy seems to be at work, as initial evidence shows that facial regions controlled by the individual (e.g., hairstyle) are sufficient to produce the effect, and socially using one’s given name is necessary to generate the effect. Together, these studies suggest that facial appearance represents social expectations of how a person with a specific name should look.
3. Finally, if time permits, I will present evidence of how the velocity of the platform we use (e.g., train, bus) influences related decisions such as risk taking, type of search for information, and focus.

In this study, we introduce a new set of features, which are extracted from the pre-cursor sequence only that based on distance between k-mers. The new set of those features is named k-mer distance. The new set are capturing the distance between each k-mer and the rest of the k-mers. The distance is calculated to be the average. The final value is normalized by the length of the sequence. Surprisingly the new set of features works as well as other hundreds of published features.

The number of photos and videos being taken each second is staggering, and keeps rising. Photos and videos are a cornerstone of communication, yet exiting tools such as Photoshop are overwhelming and hard to master, and typically supply low-level interactions that must be combined to achieve high-level goals. New tools are needed: goal-oriented tools which are easy to use, allowing users to reflect unique creative objectives.