BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Drupal iCal API//EN
X-WR-CALNAME:Events items teaser
BEGIN:VEVENT
UID:647b686788419
DTSTART;TZID=America/Toronto:20201005T093000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20201005T093000
SUMMARY:PhD Defence: Computational Mechanisms of Language Understanding and
 \nUse in the Brain and Behaviour
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS PHD DEFENCE WILL BE GIVEN ONLINE.
 \n\nIVANA KAJIĆ\, PHD CANDIDATE\n_David R. Cheriton School of Computer S
 cience_\n\nSUPERVISOR: Professor Chris Eliasmith\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678d12d
DTSTART;TZID=America/Toronto:20200923T130000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200923T130000
SUMMARY:Master’s Thesis Presentation: Sentiment Lexicon Induction and\nIn
 terpretable Multiple-instance Learning in Financial Markets
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nCHENGYAO FU\, MASTER’S CANDIDATE\n_David R. Cher
 iton School of Computer Science_\n\nSUPERVISORS: Professors Alan Huang and
  Yuying Li\n\nSentiment analysis has been widely used in the domain of fin
 ance.\nThere are two most common textual sentiment analysis methods in\nfi
 nance: \\textit{dictionary-based approach} and \\textit{machine\nlearning 
 approach}.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678da91
DTSTART;TZID=America/Toronto:20200923T110000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200923T110000
SUMMARY:Master’s Thesis Presentation: Affective and Human-Like Virtual\nA
 gents
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nNEIL BUDNARAIN\, MASTER’S CANDIDATE\n_David R. C
 heriton School of Computer Science_\n\nSUPERVISOR: Professor Jesse Hoey\n
  
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678e169
DTSTART;TZID=America/Toronto:20200917T160000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200917T160000
SUMMARY:PhD Seminar: Problems and Opportunities in Training Deep Learning\n
 Software Systems: An Analysis of Variance
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS PHD SEMINAR WILL BE GIVEN ONLINE.
 \n\nHUNG PHAM\, PHD CANDIDATE\n_David R. Cheriton School of Computer Scie
 nce_\n\nSUPERVISORS: Professors Lin Tan and Yaoliang Yu\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678e7d6
DTSTART;TZID=America/Toronto:20200909T100000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200909T100000
SUMMARY:Master’s Thesis Presentation: Disentangled Syntax and Semantics f
 or\nStylized Text Generation
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nYAO LU\, MASTER’S CANDIDATE\n_David R. Cheriton 
 School of Computer Science_\n\nSUPERVISOR: Professor Olga Vechtomova\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678ee0d
DTSTART;TZID=America/Toronto:20200805T123000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200805T123000
SUMMARY:Master’s Thesis Presentation: Variational Inference for Text\nGen
 eration: Improving the Posterior
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nVIKASH BALASUBRAMANIAN\, MASTER’S CANDIDATE\n_Da
 vid R. Cheriton School of Computer Science_\n\nLearning useful representat
 ions of data is a crucial task in machine\nlearning with wide ranging appl
 ications. In this thesis we explore\nimproving representations of models b
 ased on variational inference by\nimproving the posterior.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678f4d0
DTSTART;TZID=America/Toronto:20200805T100000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200805T100000
SUMMARY:Master’s Thesis Presentation: Dynamic Fusion Techniques for\nEffe
 ctive Multimodal Deep Learning
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nGAURAV SAHU\, MASTER’S CANDIDATE\n_David R. Cher
 iton School of Computer Science_\n\nEffective fusion of data from multiple
  modalities\, such as video\,\nspeech\, and text\, is a challenging task d
 ue to the heterogeneous\nnature of multimodal data. In this work\, we prop
 ose fusion techniques\nthat aim to model context from different modalities
  effectively.\nInstead of defining a deterministic fusion operation\, such
  as\nconcatenation\, for the network\, we let the network decide how to\nc
 ombine given multimodal features more effectively.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b68678fbf4
DTSTART;TZID=America/Toronto:20200804T090000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200804T090000
SUMMARY:Master’s Thesis Presentation: Data Augmentation for Text\nClassif
 ication Tasks
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nDANIEL TAMMING\, MASTER’S CANDIDATE\n_David R. C
 heriton School of Computer Science_\n\nThanks to increases in computing po
 wer and the growing availability of\nlarge datasets\, neural networks have
  achieved state of the art results\nin many natural language processing (N
 LP) and computer vision (CV)\ntasks. These models require a large number o
 f training examples that\nare balanced between classes\, but in many appli
 cation areas they rely\non training sets that are either small or imbalanc
 ed\, or both. To\naddress this\, data augmentation has become standard pra
 ctice in CV.\nThis research is motivated by the observation that\, relativ
 e to CV\,\ndata augmentation is underused and understudied in NLP.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b6867902a4
DTSTART;TZID=America/Toronto:20200730T130000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200730T130000
SUMMARY:Master’s Thesis Presentation: Decay Makes Supervised Predictive\n
 Coding Generative
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nWEI SUN\, MASTER’S CANDIDATE\n_David R. Cheriton
  School of Computer Science_\n\nPredictive Coding is a hierarchical model 
 of neural computation that\napproximates backpropagation using only local 
 computations and local\nlearning rules. An important aspect of Predictive 
 Coding is the\npresence of feedback connections between layers. These feed
 back\nconnections allow Predictive Coding networks to potentially be\ngene
 rative as well as discriminative. However\, Predictive Coding\nnetworks tr
 ained on supervised classification tasks cannot generate\naccurate input s
 amples close to the training inputs from the class\nvectors alone.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
BEGIN:VEVENT
UID:647b686790bf3
DTSTART;TZID=America/Toronto:20200526T100000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200526T100000
SUMMARY:Master’s Thesis Presentation: Analysis of Textual and Non-Textual
 \nSources of Sentiment in GitHub
CLASS:PUBLIC
DESCRIPTION:Summary \n\nPLEASE NOTE: THIS MASTER’S THESIS PRESENTATION WI
 LL BE GIVEN ONLINE.\n\nNALIN DE ZOYSA\, MASTER’S CANDIDATE\n_David R. C
 heriton School of Computer Science_\n\nGitHub is a collaborative platform 
 that is used primarily for the\ndevelopment of software. In order to gain 
 more insight into how teams\nwork on GitHub\, we wish to analyze the senti
 ment content available via\ncommunication on the platform.\n 
DTSTAMP:20230603T162055Z
END:VEVENT
END:VCALENDAR