-
Weight-based Decomposition: A Case for Bilinear MLPs
Authors:
Michael T. Pearce,
Thomas Dooms,
Alice Rigg
Abstract:
Gated Linear Units (GLUs) have become a common building block in modern foundation models. Bilinear layers drop the non-linearity in the "gate" but still have comparable performance to other GLUs. An attractive quality of bilinear layers is that they can be fully expressed in terms of a third-order tensor and linear operations. Leveraging this, we develop a method to decompose the bilinear tensor…
▽ More
Gated Linear Units (GLUs) have become a common building block in modern foundation models. Bilinear layers drop the non-linearity in the "gate" but still have comparable performance to other GLUs. An attractive quality of bilinear layers is that they can be fully expressed in terms of a third-order tensor and linear operations. Leveraging this, we develop a method to decompose the bilinear tensor into a set of sparsely interacting eigenvectors that show promising interpretability properties in preliminary experiments for shallow image classifiers (MNIST) and small language models (Tiny Stories). Since the decomposition is fully equivalent to the model's original computations, bilinear layers may be an interpretability-friendly architecture that helps connect features to the model weights. Application of our method may not be limited to pretrained bilinear models since we find that language models such as TinyLlama-1.1B can be finetuned into bilinear variants.
△ Less
Submitted 6 June, 2024;
originally announced June 2024.
-
Musical Features for Automatic Music Transcription Evaluation
Authors:
Adrien Ycart,
Lele Liu,
Emmanouil Benetos,
Marcus T. Pearce
Abstract:
This technical report gives a detailed, formal description of the features introduced in the paper: Adrien Ycart, Lele Liu, Emmanouil Benetos and Marcus T. Pearce. "Investigating the Perceptual Validity of Evaluation Metrics for Automatic Piano Music Transcription", Transactions of the International Society for Music Information Retrieval (TISMIR), Accepted, 2020.
This technical report gives a detailed, formal description of the features introduced in the paper: Adrien Ycart, Lele Liu, Emmanouil Benetos and Marcus T. Pearce. "Investigating the Perceptual Validity of Evaluation Metrics for Automatic Piano Music Transcription", Transactions of the International Society for Music Information Retrieval (TISMIR), Accepted, 2020.
△ Less
Submitted 15 April, 2020;
originally announced April 2020.
-
An energy-based generative sequence model for testing sensory theories of Western harmony
Authors:
Peter M. C. Harrison,
Marcus T. Pearce
Abstract:
The relationship between sensory consonance and Western harmony is an important topic in music theory and psychology. We introduce new methods for analysing this relationship, and apply them to large corpora representing three prominent genres of Western music: classical, popular, and jazz music. These methods centre on a generative sequence model with an exponential-family energy-based form that…
▽ More
The relationship between sensory consonance and Western harmony is an important topic in music theory and psychology. We introduce new methods for analysing this relationship, and apply them to large corpora representing three prominent genres of Western music: classical, popular, and jazz music. These methods centre on a generative sequence model with an exponential-family energy-based form that predicts chord sequences from continuous features. We use this model to investigate one aspect of instantaneous consonance (harmonicity) and two aspects of sequential consonance (spectral distance and voice-leading distance). Applied to our three musical genres, the results generally support the relationship between sensory consonance and harmony, but lead us to question the high importance attributed to spectral distance in the psychological literature. We anticipate that our methods will provide a useful platform for future work linking music psychology to music theory.
△ Less
Submitted 2 July, 2018;
originally announced July 2018.
-
Effects of pitch and timing expectancy on musical emotion
Authors:
Sarah A. Sauvé,
Aminah Sayed,
Roger T. Dean,
Marcus T. Pearce
Abstract:
Pitch and timing information work hand in hand to create a coherent piece of music; but what happens when this information goes against the norm? Relationships between musical expectancy and emotional responses were investigated in a study conducted with 40 participants: 20 musicians and 20 non-musicians. Participants took part in one of two behavioural paradigms measuring continuous expectancy or…
▽ More
Pitch and timing information work hand in hand to create a coherent piece of music; but what happens when this information goes against the norm? Relationships between musical expectancy and emotional responses were investigated in a study conducted with 40 participants: 20 musicians and 20 non-musicians. Participants took part in one of two behavioural paradigms measuring continuous expectancy or emotional responses (arousal and valence) while listening to folk melodies that exhibited either high or low pitch predictability and high or low onset predictability. The causal influence of pitch predictability was investigated in an additional condition where pitch was artificially manipulated and a comparison conducted between original and manipulated forms; the dynamic correlative influence of pitch and timing information and its perception on emotional change during listening was evaluated using cross-sectional time series analysis. The results indicate that pitch and onset predictability are consistent predictors of perceived expectancy and emotional response, with onset carrying more weight than pitch. In addition, musicians and non-musicians do not differ in their responses, possibly due to shared cultural background and knowledge. The results demonstrate in a controlled lab-based setting a precise, quantitative relationship between the predictability of musical structure, expectation and emotional response.
△ Less
Submitted 11 August, 2017;
originally announced August 2017.
-
Attention but not musical training affects auditory streaming
Authors:
Sarah A. Sauvé,
Marcus T. Pearce
Abstract:
While musicians generally perform better than non-musicians in various auditory discrimination tasks, effects of specific instrumental training have received little attention. The effects of instrument-specific musical training on auditory grou** in the context of stream segregation are investigated here in three experiments. In Experiment 1a, participants listened to sequences of ABA tones and…
▽ More
While musicians generally perform better than non-musicians in various auditory discrimination tasks, effects of specific instrumental training have received little attention. The effects of instrument-specific musical training on auditory grou** in the context of stream segregation are investigated here in three experiments. In Experiment 1a, participants listened to sequences of ABA tones and indicated when they heard a change in rhythm. This change is caused by the manipulation of the B tones' timbre and indexes a change in perception from integration to segregation, or vice versa. While it was expected that musicians would detect a change in rhythm earlier when their own instrument was involved, no such pattern was observed. In Experiment 1b, designed to control for potential expectation effects in Experiment 1a, participants heard sequences of static ABA tones and reported their initial perceptions, whether the sequence was integrated or segregated. Results show that participants tend to initially perceive these static sequences as segregated, and that perception is influenced by similarity between the timbres involved. Finally, in Experiment 2 violinists and flautists located mistuned notes in an interleaved melody paradigm containing a violin and a flute melody. Performance did not depend on the instrument the participant played but rather which melody their attention was directed to. Taken together, results from the three experiments suggest that the specific instrument one practices does not have an influence on auditory grou**, but attentional mechanisms are necessary for processing auditory scenes.
△ Less
Submitted 11 August, 2017;
originally announced August 2017.
