Reminder PIDSIA Seminar tomorrow (room has changed)
by Announcements of talks＠IDSIA
[Please, notice that the room has changed (now: room G1-201, just in front of the usual G1-204)]
REMINDER
*Title:
Policy Optimization via Importance Sampling
*Speaker:
Francesco Faccio (IDSIA)
*Abstract:
Policy optimization is an effective Reinforcement Learning approach to solve continuous control tasks. Recent achievements have shown that alternating online and offline optimization is a successful choice for efficient trajectory reuse. However, deciding when to stop optimizing and collect new trajectories is non-trivial, as it requires to account for the variance of the objective function estimate. In this talk, we propose a novel, model-free, policy search algorithm, POIS, applicable in both action-based and parameter-based settings. We first derive a high-confidence bound for importance sampling estimation; then we define a surrogate objective function, which is optimized offline whenever a new batch of trajectories is collected. Finally, the algorithm is tested on a selection of continuous control tasks, with both linear and deep policies, and compared with state-of-the-art policy optimization methods.
Joint work with Alberto Maria Metelli, Matteo Papini and Marcello Restelli from Politecnico di Milano. To appear at the 32nd Conference on Neural Information Processing Systems (NIPS 2018). Selected for an oral presentation.
*When: Tuesday, 27th of November 2018, 12:00-13:00
*Location: Manno, Galleria 1, 2nd floor, room G1-20
*Registration: Pizza (or alternative food) and drinks will be offered at the end of the talk. If you plan to attend, please register in a timely fashion at the following link so that we will have no shortage of food:
https://doodle.com/poll/nhf3c2mbgea657vc
*Bio:
Francesco Faccio is a Master Student in Mathematical Engineering at Politecnico di Milano. He is currently working as an intern at IDSIA, where he completed his Master's thesis. His main research interests include Reinforcement Learning, Recurrent Neural Networks and Bayesian Statistics.
6 years, 1 month
PIDSIA Seminar by Alessandro Antonucci [15.11]
by Announcements of talks＠IDSIA
*Title: PSAT: Algorithms and Applications of Probabilistic Satisfiability.
*Speaker: Alessandro Antonucci (IDSIA)
*Abstract: We discuss PSAT, a probabilistic extension of the classical satisfiability (SAT) problem. This is achieved by assigning weights to the clauses of a SAT instance. The PSAT instance is satisfiable if and
only if a probability mass function over the literals and consistent with the weights exists . We present two algorithms for PSAT based, respectively, on column generation and integer linear programming,
both showing evidence of phase transition. PSAT solves inferences in a recently proposed probabilistic logic (CCL, in [Antonucci & Facchini, 2018]). This allows to perform machine learning with logical
constraints under relaxed independence assumptions over probabilistic facts. As an application, we consider label ranking and show that existing solvers can be used to solve practical ranking tasks.
*When: Thursday, 15th of November 2018, 12:00-13:00
*Location: Manno, Galleria 1, 2nd floor, room G1-204
*Registration: Pizza (or alternative food) and drinks will be offered at the end of the talk. If you plan to attend, please register in a timely fashion at the following link so that we will have no shortage of food:
https://doodle.com/poll/2vv42mzewuv5hk6a
6 years, 1 month
Monday 12 Nov, h 10:30: "Forecasting of Photovoltaic Generation for Microgrid Applications"
by Announcements of talks＠IDSIA
::Title: "Forecasting of Photovoltaic Generation for Microgrid
Applications"
::IDSIA meeting room
::Speaker
Enrica Scolari
PhD student at EPFL
::Abstract
The penetration of stochastic renewable generation in modern power
systems requires reconsidering conventional practices to ensure the
reliable operation of the electrical network. In order to cope with the
uncertainties of the renewables, predictive schemes that leverage on
forecast of renewable generation recently came into prominence. In
particular, small-scale photovoltaic (PV) systems are expected to
represent most of the future available capacity, and consequently, solar
resource assessment and forecasting are of fundamental importance. The
presentation focuses on forecasting methods and generation models to
support the integration of PV systems in the grid, considering
short-term temporal horizons (below one hour) and limited spatial
resolution (single site installations). In this respect, nonparametric
tools to deliver prediction intervals from sub-second to intra-hour
horizons are proposed and benchmarked. They forecast the AC power and/or
the solar irradiance by extracting selected endogenous influential
variables from historical time-series. The integration of features from
all-sky images into the prediction chain will also be discussed. The
proposed methods are experimentally validated in a real microgrid by
considering possible applications in modern power systems.
6 years, 1 month