-
$Γ$-VAE: Curvature regularized variational autoencoders for uncovering emergent low dimensional geometric structure in high dimensional data
Authors:
Jason Z. Kim,
Nicolas Perrin-Gilbert,
Erkan Narmanli,
Paul Klein,
Christopher R. Myers,
Itai Cohen,
Joshua J. Waterfall,
James P. Sethna
Abstract:
Natural systems with emergent behaviors often organize along low-dimensional subsets of high-dimensional spaces. For example, despite the tens of thousands of genes in the human genome, the principled study of genomics is fruitful because biological processes rely on coordinated organization that results in lower dimensional phenotypes. To uncover this organization, many nonlinear dimensionality r…
▽ More
Natural systems with emergent behaviors often organize along low-dimensional subsets of high-dimensional spaces. For example, despite the tens of thousands of genes in the human genome, the principled study of genomics is fruitful because biological processes rely on coordinated organization that results in lower dimensional phenotypes. To uncover this organization, many nonlinear dimensionality reduction techniques have successfully embedded high-dimensional data into low-dimensional spaces by preserving local similarities between data points. However, the nonlinearities in these methods allow for too much curvature to preserve general trends across multiple non-neighboring data clusters, thereby limiting their interpretability and generalizability to out-of-distribution data. Here, we address both of these limitations by regularizing the curvature of manifolds generated by variational autoencoders, a process we coin ``$Γ$-VAE''. We demonstrate its utility using two example data sets: bulk RNA-seq from the The Cancer Genome Atlas (TCGA) and the Genotype Tissue Expression (GTEx); and single cell RNA-seq from a lineage tracing experiment in hematopoietic stem cell differentiation. We find that the resulting regularized manifolds identify mesoscale structure associated with different cancer cell types, and accurately re-embed tissues from completely unseen, out-of distribution cancers as if they were originally trained on them. Finally, we show that preserving long-range relationships to differentiated cells separates undifferentiated cells -- which have not yet specialized -- according to their eventual fate. Broadly, we anticipate that regularizing the curvature of generative models will enable more consistent, predictive, and generalizable models in any high-dimensional system with emergent low-dimensional behavior.
△ Less
Submitted 1 March, 2024;
originally announced March 2024.
-
Enumeration of corner polyhedra and 3-connected Schnyder labelings
Authors:
Éric Fusy,
Erkan Narmanli,
Gilles Schaeffer
Abstract:
We show that corner polyhedra and 3-connected Schnyder labelings join the growing list of planar structures that can be set in exact correspondence with (weighted) models of quadrant walks via a bijection due to Kenyon, Miller, Sheffield and Wilson.
Our approach leads to a first polynomial time algorithm to count these structures, and to the determination of their exact asymptotic growth constan…
▽ More
We show that corner polyhedra and 3-connected Schnyder labelings join the growing list of planar structures that can be set in exact correspondence with (weighted) models of quadrant walks via a bijection due to Kenyon, Miller, Sheffield and Wilson.
Our approach leads to a first polynomial time algorithm to count these structures, and to the determination of their exact asymptotic growth constants: the number $p_n$ of corner polyhedra and $s_n$ of 3-connected Schnyder labelings of size $n$ respectively satisfy $(p_n)^{1/n}\to 9/2$ and $(s_n)^{1/n}\to 16/3$ as $n$ goes to infinity.
While the growth rates are rational, like in the case of previously known instances of such correspondences, the exponent of the asymptotic polynomial correction to the exponential growth does not appear to follow from the now standard Denisov-Wachtel approach, due to a bimodal behavior of the step set of the underlying tandem walk. However a heuristic argument suggests that these exponents are $-1-π/\arccos(9/16)\approx -4.23$ for $p_n$ and $-1-π/\arccos(22/27)\approx -6.08$ for $s_n$, which would imply that the associated series are not D-finite.
△ Less
Submitted 29 October, 2023; v1 submitted 18 February, 2022;
originally announced February 2022.
-
On the enumeration of plane bipolar posets and transversal structures
Authors:
Éric Fusy,
Erkan Narmanli,
Gilles Schaeffer
Abstract:
We show that plane bipolar posets (i.e., plane bipolar orientations with no transitive edge) and transversal structures can be set in correspondence to certain (weighted) models of quadrant walks, via suitable specializations of a bijection due to Kenyon, Miller, Sheffield and Wilson. We then derive exact and asymptotic counting results. In particular we prove (computationally and then bijectively…
▽ More
We show that plane bipolar posets (i.e., plane bipolar orientations with no transitive edge) and transversal structures can be set in correspondence to certain (weighted) models of quadrant walks, via suitable specializations of a bijection due to Kenyon, Miller, Sheffield and Wilson. We then derive exact and asymptotic counting results. In particular we prove (computationally and then bijectively) that the number of plane bipolar posets on $n+2$ vertices equals the number of plane permutations of size $n$. Regarding transversal structures, for each $v\geq 0$ we consider $t_n(v)$ the number of such structures with $n+4$ vertices and weight $v$ per quadrangular inner face (the case $v=0$ corresponds to having only triangular inner faces). We obtain a recurrence to compute $t_n(v)$, and an asymptotic formula that for $v=0$ gives $t_n(0)\sim c\ \!(27/2)^nn^{-1-π/\mathrm{arccos}(7/8)}$ for some $c>0$, which also ensures that the associated generating function is not D-finite.
△ Less
Submitted 30 September, 2023; v1 submitted 14 May, 2021;
originally announced May 2021.
