School of Informatics
The University of Edinburgh Skip Navigation The University of Edinburgh School of Informatics

Pavlos Andreadis' blog

March 2, 2017

March 2, IPAB Workshop talk – Efficient Preference Elicitation with Coarse Preferences

IPAB workshop – 02/03/2017

Original Post here, Mar 02 2017 12.45 13.45

Pavlos Andreadis

Title:

“Efficient Preference Elicitation with Coarse Preferences”

Abstract:

“We present a new model for the representation of preferences that are based on coarse criteria, leading to better recommendations in problems where user behaviour is consistent with them making decisions  by  aggregating  many  individual  options. When  this  is  the  case,  our  ‘coarse  preference’ model enables an elicitation procedure that outperforms  the  state-of-the-art  utility  elicitation  procedure  that  is  based  on  the  assumption  of  additive independence  of  variables.   We  demonstrate  this through experiments performed with a real-world data  set  derived  from  a  mobile  clothes  shopping app.   We show that our proposed model achieves double the recommendation acceptance rate and a significant  increase  in  platform  profit,  due  to  the use of a decomposition that is better suited to the true nature of the users’ indifference regions.  Furthermore,  our  model  is  of  lower  complexity  than similar  additive  independence  models,  with  computation  time  rising  at  a  slower  rate  with  respect to dimensionality of item descriptions – achieving a halving of computation time in our experiment. Our results have insights for recommendation systems designers,  potentially enabling new ways of representing  user  utility  when  secondary  criteria need to be maximised without compromising performance.”

 

Paul Henderson

Title

”End-to-end training of object class detectors for mean average precision”

 

Abstract

I will present a method for training CNN-based object class detectors directly using mean average precision (mAP) as the training loss, in a truly end-to-end fashion that includes non-maximum suppresion (NMS) at training time. mAP following NMS forms a piecewise-constant structured loss over thousands of windows, with gradients that do not convey useful information for gradient descent. Hence, we define new, general gradient-like quantities for piecewise constant functions, which have wide applicability. We describe how to calculate these efficiently for mAP following NMS, enabling us to train a detector based on Fast R-CNN directly for mAP.

March 8, 2016

March 10, IPAB Workshop talk – Efficient Utility Elicitation using a Model of Coarse Preferences

Filed under: Talks — by s1371409 @ 10:54 pm
Tags: , , , ,

This Thursday 12:45 , March the 10th, I will be presenting my recent experiments on Coarse Preferences at room 4.31/4.33 in the Informatics Forum, University of Edinburgh. You can find the full list of speakers and abstracts below. (original post)

What
  • IPAB Workshop
When Mar 10, 2016
from 12:45 PM to 01:30 PM
Where 4.31/33
Add event to calendar vCal
iCal

 

Speaker: Pavlos Andreadis

Talk title: Efficient Utility Elicitation using a Model of Coarse Preferences

Abstract:  Preference elicitation is the process of learning a decision maker’s model of choice under non-strategic uncertainty. Starting with the hypothesis that the hypothesis space of real human choice makers admits varying levels of coarseness, we wish to construct learning and inference algorithms that can achieve preference elicitation more quickly and efficiently than with techniques that do not model this. A key technical issue is to identify the user’s own representation of the solution space, which we approach by reasoning about possible partitions over the original outcome space. We argue that learning a model in this way and making inferences using beliefs over the user’s individual representation of the solution space can yield significant benefits in performance, as measured for instance by the number of queries needed to make recommendations. I will present an instantiation of this methodology, including learning a model of coarse preferences through decision tree based representations. I will also present preliminary results on a synthetic data set from a ride-sharing scenario, and with real human choice data in the Sushi Dataset, showing that preference elicitation using our model incorporating coarseness is faster and no less accurate than the baseline which ignores this structure.

 

Speaker: Adam Erskine

Talk Title: Parameter selection in Particle Swarm optimisation

Abstract: The particle swarm optimisation (PSO) algorithm is a popular metaheuristic used to solve search and optimisation type problems. Originally inspired by bird flocking it is simple to implement and continues to be used and studied. It requires parameter tuning to obtain good solutions to a given problem. The parameters control the balance between exploration  and exploitation of a problem space. Analysis of PSO as a random dynamical system predicts that there exists a locus of parameter values that lead to optimum performance. Here, we outline this approach and provide supporting empirical evidence.

 

Speaker: Tom Stone

Talk Title: Skyline-based Localisation for Aggressively Manoeuvring Robots using UV sensors and Spherical Harmonics

Abstract:  Place recognition is a key capability for navigating robots. While significant advances have been achieved on large, stable platforms such as robot cars, achieving robust performance on rapidly manoeuvring platforms in outdoor natural conditions remains a challenge, with few systems able to deal with both variable conditions and large tilt variations caused by rough terrain.  Taking inspiration from biology, we propose a novel combination of sensory modality and image processing to obtain a significant improvement in the robustness of sequence-based image matching for place recognition. We use a UV-sensitive fisheye lens camera to segment sky from ground, providing illumination invariance, and encode the resulting binary images using spherical harmonics to enable rotation-invariant image matching. In combination, these methods also produce substantial pitch and roll invariance, as the spherical harmonics for the sky shape are minimally affected, providing the sky remains visible.

Our system demonstrated an improved condition- and tilt-invariance, enabling robust place recognition during aggressive zigzag manoeuvring along bumpy trails and at tilt angles of up to 60 degrees. I will present our methods and also the resulting performance compared to a leading appearance-invariant technique (SeqSLAM) and a leading viewpoint-invariant technique (FAB-MAP 2.0) on three new outdoor datasets encompassing variable robot heading, tilt, and lighting conditions in both forested and urban environments.