Authors: Daniel Wolff and Tillman Weyde

Abstract:
Predicting user’s tastes on music has become crucial for a competitive music recommendation systems, and perceived similarity plays an influential role in this. MIR currently turns towards making recommendation systems adaptive to user preferences and context. Here, we consider the particular task of adapting music similarity measures to user voting data. This work builds on and responds to previous publications based on the MagnaTagATune dataset. We have reproduced the similarity dataset presented by Stober and Nürnberger at AMR 2011 to enable a comparison of approaches. On this dataset, we compare their two-level approach, defning similarity measures on individual facets and combining them in a linear model, to the Metric Learning to Rank (MLR) algorithm. MLR adapts a similarity measure that operates directly on low-level features to the user data. We compare the different algorithms, features and parameter spaces with regards to minimising constraint violations. Furthermore, the effectiveness of the MLR algorithm in generalising to unknown data is evaluated on this
dataset. We also explore the effects of feature choice. Here, we find that the binary genre data shows little correlation with the similarity data, but combined with audio features it clearly improves generalisation.

As probaby a lot of people who use to carry their compact digital cameras around with them, to places the engineers of that device probably never dreamt about (e.g. a beach), I came onto this error a couple of times. Usually, shaking or slightly hitting the camera worked best to resolve any sand-related problems.

Today, though, my camera wouldn’t start with exactly the same error message. This time though there was no much motor sound audible at all from the zoom lens drivers. I used the fabulous guide available here to start off the disassembly of my camera.

The disassembled ixus 870, note that the top button board is still on the camera, but should be removed before removing the flash component

Generally this description, removing the flash component to get onto the zoom gears worked fine. Its wise to use proper x- shape – screwdrivers for this as the camera is assembled tightly. Furthermore care for a couple of extra-screws not mentioned in the manual above. Moreover be CAREFULwith the flash condenser. It is not only wise, but anyone with the interest in his/her own health or a slight interest in fireworks should discharge the flash condenser with a PROPERLY INSULATED bridge or even better:resistor, in order to avoid any unplanned dangerous electric shock. The condenser’s connections are reachable when the front and back-panels are removed.

When removing and reassembling the flash device, carefully open the flash connector located under the top button-panel of the camera by removing the button-panel first.

Flash Connector and disassembled button board

In my case, the removal of the flash device was not necessary at all, as the motor itself was stuck (hence no turning sounds). Good luck with your work, and please let me know if this was helpful to you!

Authors: Daniel Wolff and Tillman Weyde

Abstract: Understanding how we relate and compare pieces of music has been a topic of great interest in musicology as well as for business applications, such as music recommender systems. The way music is compared seems to vary among both individuals and cultures. Adapting a generic model to user ratings is useful for personalisation and can help to better understand such differences. This paper presents an approach to use machine learning techniques for analysing user data that speci?es song similarity. We explore the potential for learning generalisable similarity measures with two state of-the-art algorithms for learning metrics. We use the audio clips and user ratings in the MagnaTagATune dataset, enriched with genre annotations from the Magnatune label.

PDF: Poster

Detecting Bird Sounds via Periodic Structures

A Robust Pattern Recognition Approach to Unsupervised Animal Monitoring

Authors: Daniel Wolff and Tillman Weyde

Abstract: We present an approach to  automatically adapting a computational model for music similarity  to users’ cultural contexts, indicated by parameters like location and age. Using user ratings of perceived similarity, we aim to model variations in  these ratings  associated with culture-based variables of the users.  The resulting similarity measure  meets its useful application in music information retrieval, particularly personalised music recommendation,  but  relations of cultural variables and music perception,  which are represented in the adapted similarity models,  are also of generic musicological interest. This paper  outlines  the general framework envisaged for this task and first experimental results on the feasibility of learning musical similarity from relative user ratings.

PDF: Presentation Slides

Authors: Daniel Wolff and Tillman Weyde

Abstract: In this paper, we compare the effectiveness of basic acoustic features and genre annotations when adapting a music similarity model to user ratings. We use the Metric Learning to Rank algorithm to learn a Mahalanobis metric from comparative similarity ratings in in the MagnaTagATune database. Using common formats for feature data, our approach can easily be transferred to other existing databases. Our results show a notable correlation between songs’ genres and associated similarity ratings, but learning on a combined feature set clearly outperforms either individual approach.