Search Results for author:
Found 6 papers, 1 papers with code
A universal niche geometry governs the response of ecosystems to environmental perturbations
How ecosystems respond to environmental perturbations is a fundamental question in ecology, made especially challenging due to the strong coupling between species and their environment.
Double Descent Demystified: Identifying, Interpreting & Ablating the Sources of a Deep Learning Puzzle
Double descent is a surprising phenomenon in machine learning, in which as the number of model parameters grows relative to the number of data, test error drops as models grow ever larger into the highly overparameterized (data undersampled) regime.
Emergent competition shapes the ecological properties of multi-trophic ecosystems
Ecosystems are commonly organized into trophic levels -- organisms that occupy the same level in a food chain (e. g., plants, herbivores, carnivores).
Bias-variance decomposition of overparameterized regression with random linear features
We show that the linear random features model exhibits three phase transitions: two different transitions to an interpolation regime where the training error is zero, along with an additional transition between regimes with large bias and minimal bias.
The Geometry of Over-parameterized Regression and Adversarial Perturbations
Classical regression has a simple geometric description in terms of a projection of the training labels onto the column space of the design matrix.
Memorizing without overfitting: Bias, variance, and interpolation in over-parameterized models
In both models, increasing the number of fit parameters leads to a phase transition where the training error goes to zero and the test error diverges as a result of the variance (while the bias remains finite).
