BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:PCF-GAN and beyond:  High-Fidelity Generative Models for Synthetic
  Time Series Generation - Hao Ni (University College London)
DTSTART:20240904T123000Z
DTEND:20240904T131500Z
UID:TALK217603@talks.cam.ac.uk
DESCRIPTION:Generating high-fidelity time series data using generative adv
 ersarial networks (GANs) remains a challenging task\, as it is difficult t
 o capture the temporal dependence of joint probability distributions induc
 ed by time-series data. To this end\, a key step is the development of an 
 effective discriminator to distinguish between time series distributions.&
 nbsp\;In this&nbsp\;talk\, I will introduce&nbsp\;the so-called&nbsp\;PCF-
 GAN\, a novel GAN that incorporates the path characteristic function (PCF)
  as the principled representation of time series distribution into the dis
 criminator to enhance its generative performance.&nbsp\; On the one hand\,
  we establish theoretical foundations of the&nbsp\;PCF&nbsp\;distance by p
 roving its characteristicity\, boundedness\, differentiability with respec
 t to generator parameters\, and weak continuity\, which ensure the stabili
 ty and feasibility of training the&nbsp\;PCF-GAN. On the other hand\, we d
 esign efficient initialisation and optimisation schemes for&nbsp\;PCFs to 
 strengthen the discriminative power and accelerate training efficiency. To
  further boost the capabilities of complex time series generation\, we int
 egrate the auto-encoder structure via sequential embedding into the&nbsp\;
 PCF-GAN\, which provides additional reconstruction functionality. Extensiv
 e numerical experiments on various datasets demonstrate the consistently s
 uperior performance of&nbsp\;PCF-GAN over state-of-the-art baselines\, in 
 both generation and reconstruction quality. Lastly\, an application of&nbs
 p\;PCF-GAN to Levy area generation is presented\, which shows its potentia
 l to accelerate the high-order SDE simulation.\nThis&nbsp\;talk&nbsp\;is b
 ased on two papers:&nbsp\;[https://arxiv.org/pdf/2305.12511.pdf] (joint wo
 rk with Siran Li (Shanghai Jiao Tong University) and Hang Lou (UCL) ) and 
 [https://arxiv.org/pdf/2308.02452.pdf] (Joint work with Andraž Jelinčič
  (Oxford)\, Jiajie Tao (UCL)\, William F Turner (Imperial College)\, Thoma
 s Cass (Imperial College)\, James Foster (Oxford)).
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR