NEWS

Data for Refugees
Türk Telekom, TÜBİTAK and Boğaziçi University initiated the "D4R – Data for Read more...
Senior Projects Poster Session
Our senior students have completed their CMPE 491 Graduation project Read more...
EU Funding for Full-time Msc/Phd Positions in Cognitive Robotics and Robot Learning
Project name: IMAGINE: Robots Understanding Their Actions by Imagining Their Read more...
Special 6-week training course organized with Havelsan: "Introduction to Machine Learning and Data Analysis"

CmpE Events

Tuesday, May 29th

  1. PhD Thesis Defense: Reception Modeling and Achievable Rate Analysis of Sphere-to-Sphere Molecular Communication via Diffusion by Gaye Genç
    • Start time: 10:00am, Tuesday, May 29th
    • End time: 11:00am, Tuesday, May 29th
    • Where: AVS Conference Room, BM
    • Abstract:
      As a widely acknowledged information transfer method in the nanonetworking domain, Molecular Communication via Diffusion (MCvD) presents many advantages as well as challenges. In order to assess the capabilities and restrictions of MCvD, a thorough understanding of the reception process and the achievable rate holds utmost importance. With a reflective spherical transmitter and a fully absorbing spherical receiver, the network setup becomes more realistic, but analytical derivations become increasingly difficult. In this thesis, we propose two novel heavy-tail distributions to statistically model the distribution of the first passage time of messenger molecules (MM), conduct Kolmogorov-Smirnov goodness of fit tests for model validation, and examine the modeling performance under diverse deployment parameters. We also investigate MM absorption probability, Signal-to-Interference Ratio, and the advantages of using a reflective transmitter. Since the heavy-tailed signal causes Inter-Symbol Interference (ISI), the MCvD channel has memory, and Shannon's capacity formula for memoryless channels is inapplicable. To this end, we propose an accurate ISI-aware model of demodulation and bit error probabilities for Binary Concentration Shift Keying modulated MCvD, carry out goodness of fit tests, and prove that the literature's assumption of a single symbol duration memory is overly optimistic. Furthermore, we adapt the general formulation of the achievable rate for ergodic finite state ISI channels to MCvD and investigate the effect of deployment parameters, demodulation threshold, and input distribution on the achievable rate. We also present preliminary findings on estimating the achievable rate with a neural network. Finally, we apply the biological concept of protrusions to MCvD, in order to physically reduce ISI and study the effect of protrusion deployment parameters on the achievable rate.

    • View this event in Google Calendar

Friday, June 1st

  1. PhD Thesis Defense: Crowd Labelling for Continuous Valued Annotations by Yunus Emre Kara
    • Start time: 02:00pm, Friday, June 1st
    • End time: 03:00pm, Friday, June 1st
    • Where: AVS Conference Room, BM
    • Abstract:

      As machine learning gained immense popularity across a wide variety of domains
      in the last decade, it has become more important than ever to have fast and inexpensive ways to annotate vast amounts of data. With the emergence of crowdsourcing services, the research direction has gravitated toward putting `the wisdom of crowds' to use. We call the process of crowdsourcing based label collection crowd-labeling. In this thesis, we focus on crowd consensus estimation of continuous-valued labels. Unfortunately, spammers and inattentive annotators pose a threat to the quality and trustworthiness of the consensus. Thus, we develop Bayesian models taking different annotator behaviours into account and introduce two crowd-labeled datasets for evaluating our models. High quality consensus estimation requires a meticulous choice of the candidate annotator and the sample in need of a new annotation. Due to time and budget limitations, it is beneficial to make this choice while collecting the annotations. To this end, we propose an active crowd-labeling approach for actively estimating consensus from continuous valued crowd annotations. Our method is based on annotator models with unknown parameters, and Bayesian inference is employed to reach a consensus in the form of ordinal, binary, or continuous values. We introduce ranking functions for choosing the candidate annotator and sample pair for requesting an annotation. In addition, we propose a penalizing method for preventing annotator domination, investigate the explore-exploit trade-o for incorporating new annotators into the system, and study the effects of inducing a stopping criterion based on consensus quality. Experimental results on the benchmark datasets suggest that our method provides a budget and
      time-sensitive solution to the crowd-labeling problem. Finally, we introduce a multivariate model incorporating cross attribute correlations in multivariate annotations and present preliminary observations.

    • View this event in Google Calendar

Contact us

Department of Computer Engineering, Boğaziçi University,
34342 Bebek, Istanbul, Turkey

  • Phone: +90 212 359 45 23/24
  • Fax: +90 212 2872461
 

Connect with us

We're on Social Networks. Follow us & get in touch.