
Perceptrons are the building blocks of many theoretical approaches to a wide
range of complex systems, ranging from neural networks and deep learning
machines, to constraint satisfaction problems, glasses and ecosystems. Despite
their applicability and importance, a detailed study of their Langevin dynamics
has never been performed yet. Here we derive the meanfield dynamical equations
that describe the continuous random perceptron in the thermodynamic limit, in a
very general setting with arbitrary noise and friction kernels, not necessarily
related by equilibrium relations. We derive the equations in two ways: via a
dynamical cavity method, and via a pathintegral approach in its supersymmetric
formulation. The end point of both approaches is the reduction of the dynamics
of the system to an effective stochastic process for a representative dynamical
variable. Because the perceptron is formally very close to a system of
interacting particles in a high dimensional space, the methods we develop here
can be transferred to the study of liquid and glasses in high dimensions.
Potentially interesting applications are thus the study of the glass transition
in active matter, the study of the dynamics around the jamming transition, and
the calculation of rheological properties in driven systems.

In this work we discuss possible physical origins for nontrivial exponents
in the athermal rheology of soft materials at low but finite driving rates. A
key ingredient in our scenario is the presence of a selfconsistent mechanical
noise that stems from the spatial superposition of longrange elastic responses
to localized plastically deforming regions. We study analytically a meanfield
model, in which this mechanical noise is accounted for by a stress diffusion
term coupled to the plastic activity. Within this description we show how a
dependence of the shear modulus and/or the local relaxation time on the shear
rate introduces corrections to the usual meanfield prediction, concerning the
HerschelBulkleytype rheological response of exponent 1/2. This feature of the
meanfield picture is then shown to be robust with respect to structural
disorder and partial relaxation of the local stress. We test this prediction
numerically on a mesoscopic lattice model that implements explicitly the
longrange elastic response to localized shear transformations, and we conclude
on how our scenario might be tested in rheological experiments.

Mechanical signaling plays a key role in biological processes like embryo
development and cancer growth. One prominent way to probe mechanical properties
of tissues is to study their response to externally applied forces. Using a
particlebased model featuring random apoptosis and environmentdependent
division rates, we evidence a crossover from linear flow to a shearthinning
regime with increasing shear rate. To rationalize this nonlinear flow we
derive a theoretical meanfield scenario that accounts for the interplay of
mechanical and active noise in local stresses. These noises are respectively
generated by the elastic response of the cell matrix to cell rearrangements and
by the internal activity.

We investigate the stationarystate fluctuations of a growing onedimensional
interface described by the KPZ dynamics with a noise featuring smooth spatial
correlations of characteristic range $\xi$. We employ Nonperturbative
Functional Renormalization Group methods in order to resolve the properties of
the system at all scales. We show that the physics of the standard
(uncorrelated) KPZ equation emerges on large scales independently of $\xi$.
Moreover, the Renormalization Group flow is followed from the initial condition
to the fixed point, that is from the microscopic dynamics to the largedistance
properties. This provides access to the smallscale features (and their
dependence on the details of the noise correlations) as well as to the
universal largescale physics. In particular, we compute the kinetic energy
spectrum of the stationary state as well as its nonuniversal amplitude. The
latter is experimentally accessible by measurements at large scales and retains
a signature of the microscopic noise correlations. Our results are compared to
previous analytical and numerical results from independent approaches. They are
in agreement with direct numerical simulations for the kinetic energy spectrum
as well as with the prediction, obtained with the replica trick by Gaussian
variational method, of a crossover in $\xi$ of the nonuniversal amplitude of
this spectrum.

We study the scaling properties of a onedimensional interface at
equilibrium, at finite temperature and in a disordered environment with a
finite disorder correlation length. We focus our approach on the scalings of
its geometrical fluctuations as a function of its length. At large
lengthscales, the roughness of the interface, defined as the variance of its
endpoint fluctuations, follows a powerlaw behaviour whose exponent
characterises its superdiffusive behaviour. In 1+1 dimensions, the roughness
exponent is known to be the characteristic 2/3 exponent of the
KardarParisiZhang (KPZ) universality class. An important feature of the model
description is that its Flory exponent, obtained by a power counting argument
on its Hamiltonian, is equal to 3/5 and thus does not yield the correct KPZ
roughness exponent. In this work, we review the available powercounting
options, and relate the physical validity of the exponent values that they
predict, to the existence (or not) of welldefined optimal trajectories in a
largesize or lowtemperature asymptotics. We identify the crucial role of the
'cutoff' lengths of the problem (the disorder correlation length and the
system size), which one has to carefully follow throughout the scaling
analysis. To complement the latter, we device a novel Gaussian Variational
Method (GVM) scheme to compute the roughness, taking into account the effect of
a large but finite interface length. Interestingly, such a procedure yields the
correct KPZ roughness exponent, instead of the Flory exponent usually obtained
through the GVM approach for an infinite interface. We explain the physical
origin of this improvement of the GVM procedure and discuss possible extensions
of this work to other disordered systems.

The response of spatially extended systems to a force leading their steady
state out of equilibrium is strongly affected by the presence of disorder. We
focus on the mean velocity induced by a constant force applied on
onedimensional interfaces. In the absence of disorder, the velocity is linear
in the force. In the presence of disorder, it is widely admitted, as well as
experimentally and numerically verified, that the velocity presents a stretched
exponential dependence in the force (the socalled 'creep law'), which is out
of reach of linear response, or more generically of direct perturbative
expansions at small force. In dimension one, there is no exact analytical
derivation of such a law, even from a theoretical physical point of view. We
propose an effective model with two degrees of freedom, constructed from the
full spatially extended model, that captures many aspects of the creep
phenomenology. It provides a justification of the creep law form of the
velocityforce characteristics, in a quasistatic approximation. It allows,
moreover, to capture the nontrivial effects of shortrange correlations in the
disorder, which govern the lowtemperature asymptotics. It enables us to
establish a phase diagram where the creep law manifests itself in the vicinity
of the origin in the forcesystemsizetemperature coordinates. Conjointly,
we characterise the crossover between the creep regime and a linearresponse
regime that arises due to finite system size.

We show that, at least at a meanfield level, the effect of structural
disorder in sheared amorphous media is very dissimilar depending on the thermal
or athermal nature of their underlying dynamics. We first introduce a toy
model, including explicitly two types of noise (thermal versus athermal).
Within this interpretation framework, we argue that meanfield athermal
dynamics can be accounted for by the socalled H{\'e}braudLequeux (HL) model,
in which the mechanical noise stems explicitly from the plastic activity in the
sheared medium. Then, we show that the inclusion of structural disorder, by
means of a distribution of yield energy barriers, has no qualitative effect in
the HL model, while such a disorder is known to be one of the key ingredients
leading kinematically to a finite macroscopic yield stress in other meanfield
descriptions, such as the SoftGlassyRheology model. We conclude that the
statistical mechanisms at play in the emergence of a macroscopic yield stress,
and a complex stationary dynamics at low shear rate, are different in thermal
and athermal amorphous systems.

Experimental realizations of a 1D interface always exhibit a finite
microscopic width $\xi>0$; its influence is erased by thermal fluctuations at
sufficiently high temperatures, but turns out to be a crucial ingredient for
the description of the interface fluctuations below a characteristic
temperature $T_c(\xi)$. Exploiting the exact mapping between the static 1D
interface and a 1+1 Directed Polymer (DP) growing in a continuous space, we
study analytically both the freeenergy and geometrical fluctuations of a DP,
at finite temperature $T$, with a shortrange elasticity and submitted to a
quenched randombond Gaussian disorder of finite correlation length $\xi$.
We derive the exact `time'evolution equations of the disorder freeenergy
$\bar{F}(t,y)$, its derivative $\eta (t,y)$, and their respective twopoint
correlators $\bar{C}(t,y)$ and $\bar{R}(t,y)$. We compute the exact solution of
its linearized evolution $\bar{R}^{lin}(t,y)$, and we combine its qualitative
behavior and the asymptotic properties known for an uncorrelated disorder
($\xi=0$), to construct a `toymodel' leading to a simple description of the DP.
This model is characterized by Brownianlike freeenergy fluctuations,
correlated at small $y<\xi$, of amplitude $\tilde{D}_{\infty}(T,\xi)$. We
present an extended scaling analysis of the roughness predicting
$\tilde{D}_{\infty} \sim 1/T$ at hightemperatures and $\tilde{D}_{\infty} \sim
1/T_c(\xi)$ at lowtemperatures. We identify the connection between the
temperatureinduced crossover and the full replicasymmetry breaking in
previous Gaussian Variational Method computations. Finally we discuss the
consequences of the lowtemperature regime for two experimental realizations of
KPZ interfaces, namely the static and quasistatic behavior of magnetic domain
walls and the highvelocity steadystate dynamics of interfaces in liquid
crystals.

We study numerically the geometrical and freeenergy fluctuations of a static
onedimensional (1D) interface with a shortrange elasticity, submitted to a
quenched randombond Gaussian disorder of finite correlation length $\xi>0$,
and at finite temperature $T$. Using the exact mapping from the static 1D
interface to the 1+1 Directed Polymer (DP) growing in a continuous space, we
focus our analysis on the disorder freeenergy of the DP endpoint, a quantity
which is strictly zero in absence of disorder and whose sampletosample
fluctuations at a fixed growing `time' $t$ inherit the statistical
translationinvariance of the microscopic disorder explored by the DP.
Constructing a new numerical scheme for the integration of the
KardarParisiZhang (KPZ) evolution equation obeyed by the freeenergy, we
address numerically the `time' and temperaturedependence of the disorder
freeenergy fluctuations at fixed finite $\xi$. We examine on one hand the
amplitude $\tilde{D}_{t}$ and effective correlation length $\tilde{\xi}_t$ of
the freeenergy fluctuations, and on the other hand the imprint of the specific
microscopic disorder correlator on the large`time' shape of the freeenergy
twopoint correlator. We observe numerically the crossover to a lowtemperature
regime below a finite characteristic temperature $T_c(\xi)$, as previously
predicted by GaussianVariationalMethod (GVM) computations and scaling
arguments, and extensively investigated analytically in [Phys. Rev. E, 87
042406 (2013)]. Finally we address numerically the `time' and
temperaturedependence of the roughness $B(t)$, which quantifies the DP
endpoint transverse fluctuations, and we show how the amplitude
$\tilde{D}_{\infty}(T,\xi)$ controls the different regimes experienced by
$B(t)$  in agreement with the analytical predictions of a DP `toymodel'
approach.

We study the fluctuations of the directed polymer in 1+1 dimensions in a
Gaussian random environment with a finite correlation length {\xi} and at
finite temperature. We address the correspondence between the geometrical
transverse fluctuations of the directed polymer, described by its roughness,
and the fluctuations of its freeenergy, characterized by its twopoint
correlator. Analytical arguments are provided in favor of a generic scaling law
between those quantities, at finite time, nonvanishing {\xi} and explicit
temperature dependence. Numerical results are in good agreement both for
simulations on the discrete directed polymer and on a continuous directed
polymer (with shortrange correlated disorder). Applications to recent
experiments on liquid crystals are discussed.

We briefly introduce the generic framework of Disordered Elastic Systems
(DES), giving a short `recipe' of a DES modeling and presenting the quantities
of interest in order to probe the static and dynamical disorderinduced
properties of such systems. We then focus on a particular lowdimensional DES,
namely the onedimensional interface in shortranged elasticity and
shortranged quenched disorder. Illustrating different elements given in the
introductory sections, we discuss specifically the consequences of the
interplay between a finite temperature T>0 and a finite interface width \xi>0
on the static geometrical fluctuations at different lengthscales, and the
implications on the quasistatic dynamics.

At finite temperature and in presence of disorder, a onedimensional elastic
interface displays different scaling regimes at small and large lengthscales.
Using a replica approach and a Gaussian Variational Method (GVM), we explore
the consequences of a finite interface width $\xi$ on the smalllengthscale
fluctuations. We compute analytically the static roughness $B(r)$ of the
interface as a function of the distance $r$ between two points on the
interface. We focus on the case of shortrange elasticity and randombond
disorder. We show that for a finite width $\xi$ two temperature regimes exist.
At low temperature, the expected thermal and randommanifold regimes,
respectively for small and large scales, connect via an intermediate `modified'
Larkin regime, that we determine. This regime ends at a temperatureindependent
characteristic `Larkin' length. Above a certain `critical' temperature that we
identify, this intermediate regime disappears. The thermal and randommanifold
regimes connect at a single crossover lengthscale, that we compute. This is
also the expected behavior for zero width. Using a directed polymer
description, we also study via a second GVM procedure and generic scaling
arguments, a modified toy model that provides further insights on this
crossover. We discuss the relevance of the two GVM procedures for the roughness
at large lengthscale in those regimes. In particular we analyze the scaling of
the temperaturedependent prefactor in the roughness $B(r)\sim T^{2
\text{\thorn}} r^{2 \zeta}$ and its corresponding exponent $\text{\thorn}$. We
briefly discuss the consequences of those results for the quasistatic creep law
of a driven interface, in connection with previous experimental and numerical
studies.