Long short-term memory neural network equilibria computation and analysis

Massinissa Amrouche; Shankar Deka; Aleksandra Lekic; Vicenc Rubio Royo; Elaina Teresa Chai; Dušan M. Stipanović; Boris Murmann; Claire J. Tomlin

Long short-term memory neural network equilibria computation and analysis

Massinissa Amrouche, Shankar Deka, Aleksandra Lekic, Vicenc Rubio Royo, Elaina Teresa Chai, Dušan M. Stipanović, Boris Murmann, Claire J. Tomlin

Ei julkaistu Aalto-yliopiston affiliaatiolla

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussa › Conference article in proceedings › Scientific › vertaisarvioitu

Abstrakti

This paper presents a comprehensive approach for computing nontrivial equilibria of autonomous Long Short-Term Memory neural networks using a homotopy formulation. Through simulations, it is shown that the eigenvalues of the linearized models around these nontrivial equilibria tend to move closer to the unit circle as the complexity of the training data increases. This provides insights into the dynamical properties of the LSTM neural networks.

Alkuperäiskieli	Englanti
Otsikko	Spatiotemporal Workshop, 32nd Annual Conference on Neural Information Processing Systems (NeurIPS)
Sivumäärä	5
Tila	Julkaistu - 30 syysk. 2018
OKM-julkaisutyyppi	A4 Artikkeli konferenssijulkaisussa

Siteeraa tätä

@inproceedings{5ad16c2137f14c388ff0e60d0ba519a1,

title = "Long short-term memory neural network equilibria computation and analysis",

abstract = "This paper presents a comprehensive approach for computing nontrivial equilibria of autonomous Long Short-Term Memory neural networks using a homotopy formulation. Through simulations, it is shown that the eigenvalues of the linearized models around these nontrivial equilibria tend to move closer to the unit circle as the complexity of the training data increases. This provides insights into the dynamical properties of the LSTM neural networks.",

author = "Massinissa Amrouche and Shankar Deka and Aleksandra Lekic and {Rubio Royo}, Vicenc and Chai, {Elaina Teresa} and Stipanovi{\'c}, {Du{\v s}an M.} and Boris Murmann and Tomlin, {Claire J.}",

year = "2018",

month = sep,

day = "30",

language = "English",

booktitle = "Spatiotemporal Workshop, 32nd Annual Conference on Neural Information Processing Systems (NeurIPS)",

}

Long short-term memory neural network equilibria computation and analysis. / Amrouche, Massinissa; Deka, Shankar; Lekic, Aleksandra et al.
Spatiotemporal Workshop, 32nd Annual Conference on Neural Information Processing Systems (NeurIPS). 2018.

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussa › Conference article in proceedings › Scientific › vertaisarvioitu

TY - GEN

T1 - Long short-term memory neural network equilibria computation and analysis

AU - Amrouche, Massinissa

AU - Deka, Shankar

AU - Lekic, Aleksandra

AU - Rubio Royo, Vicenc

AU - Chai, Elaina Teresa

AU - Stipanović, Dušan M.

AU - Murmann, Boris

AU - Tomlin, Claire J.

PY - 2018/9/30

Y1 - 2018/9/30

N2 - This paper presents a comprehensive approach for computing nontrivial equilibria of autonomous Long Short-Term Memory neural networks using a homotopy formulation. Through simulations, it is shown that the eigenvalues of the linearized models around these nontrivial equilibria tend to move closer to the unit circle as the complexity of the training data increases. This provides insights into the dynamical properties of the LSTM neural networks.

AB - This paper presents a comprehensive approach for computing nontrivial equilibria of autonomous Long Short-Term Memory neural networks using a homotopy formulation. Through simulations, it is shown that the eigenvalues of the linearized models around these nontrivial equilibria tend to move closer to the unit circle as the complexity of the training data increases. This provides insights into the dynamical properties of the LSTM neural networks.

M3 - Conference article in proceedings

BT - Spatiotemporal Workshop, 32nd Annual Conference on Neural Information Processing Systems (NeurIPS)

ER -

Long short-term memory neural network equilibria computation and analysis

Abstrakti

Sormenjälki

Siteeraa tätä