SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.07688 · 2021

Next Generation Reservoir Computing

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2106.07688.json · badge: /api/ran/2106.07688.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
quantinfo/ng-rc-paper-code canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
find_err Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVAR-NRMSE-vsTraintime.py
code served (permissive licence) · get_code("89839651a9a0f904")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy.integrate import solve_ivp

dt = 0.025

ridge_param = 2.5e-06

def find_err(warmup, traintime):
    lyaptime = 1.104
    testtime = lyaptime
    maxtime = warmup + traintime + testtime
    warmup_pts = round(warmup / dt)
    traintime_pts = round(traintime / dt)
    warmtrain_pts = warmup_pts + traintime_pts
    testtime_pts = round(testtime / dt)
    maxtime_pts = round(maxtime / dt)
    lyaptime_pts = round(lyaptime / dt)
    d = 3
    k = 2
    dlin = k * d
    dnonlin = int(dlin * (dlin + 1) / 2)
    dtot = 1 + dlin + dnonlin
    t_eval = np.linspace(0, maxtime, maxtime_pts + 1)
    sigma = 10
    beta = 8 / 3
    rho = 28

    def lorenz(t, y):
        dy0 = sigma * (y[1] - y[0])
        dy1 = y[0] * (rho - y[2]) - y[1]
        dy2 = y[0] * y[1] - beta * y[2]
        return [dy0, dy1, dy2]
    lorenz_soln = solve_ivp(lorenz, (0, maxtime), [17.67715816276679, 12.931379185960404, 43.91404334248268], t_eval=t_eval, method='RK23')
    lorenz_stats = np.zeros((3, 3))
    for i in range(3):
        lorenz_stats[0, i] = np.mean(lorenz_soln.y[i, warmtrain_pts:maxtime_pts])
        lorenz_stats[1, i] = np.min(lorenz_soln.y[i, warmtrain_pts:maxtime_pts])
        lorenz_stats[2, i] = np.max(lorenz_soln.y[i, warmtrain_pts:maxtime_pts])
    total_var = np.var(lorenz_soln.y[0, :]) + np.var(lorenz_soln.y[1, :]) + np.var(lorenz_soln.y[2, :])
    x = np.zeros((dlin, maxtime_pts))
    for delay in range(k):
        for j in range(delay, maxtime_pts):
            x[d * delay:d * (delay + 1), j] = lorenz_soln.y[:, j - delay]
    out_train = np.ones((dtot, traintime_pts))
    out_train[1:dlin + 1, :] = x[:, warmup_pts - 1:warmtrain_pts - 1]
    cnt = 0
    for row in range(dlin):
        for column in range(row, dlin):
            out_train[dlin + 1 + cnt] = x[row, warmup_pts - 1:warmtrain_pts - 1] * x[column, warmup_pts - 1:warmtrain_pts - 1]
            cnt += 1
    W_out = (x[0:d, warmup_pts:warmtrain_pts] - x[0:d, warmup_pts - 1:warmtrain_pts - 1]) @ out_train[:, :].T @ np.linalg.pinv(out_train[:, :] @ out_train[:, :].T + ridge_param * np.identity(dtot))
    x_predict = x[0:d, warmup_pts - 1:warmtrain_pts - 1] + W_out @ out_train[:, 0:traintime_pts]
    train_nrmse = np.sqrt(np.mean((x[0:d, warmup_pts:warmtrain_pts] - x_predict[:, :]) ** 2) / total_var)
    out_test = np.ones(dtot)
    x_test = np.zeros((dlin, testtime_pts))
    x_test[:, 0] = x[:, warmtrain_pts - 1]
    for j in range(testtime_pts - 1):
        out_test[1:dlin + 1] = x_test[:, j]
        cnt = 0
        for row in range(dlin):
            for column in range(row, dlin):
                out_test[dlin + 1 + cnt] = x_test[row, j] * x_test[column, j]
                cnt += 1
        x_test[d:dlin, j + 1] = x_test[0:dlin - d, j]
        x_test[0:d, j + 1] = x_test[0:d, j] + W_out @ out_test[:]
    test_nrmse = np.sqrt(np.mean((x[0:d, warmtrain_pts - 1:warmtrain_pts + lyaptime_pts - 1] - x_test[0:d, 0:lyaptime_pts]) ** 2) / total_var)
    return test_nrmse

ARGS = {
    "warmup": 0.5,
    "traintime": 1.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
find_err Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVARtimedelayNRMSE-RK23.py
code served (permissive licence) · get_code("d9793b72efea5300")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    "warmup": 5.0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
lorenz Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVAR-Noise.py
code served (permissive licence) · get_code("54019732db0279c6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy.interpolate import interp1d

dt = 0.025
warmup = 5.0
traintime = 10.0
testtime = 120.0
maxtime = warmup + traintime + testtime
maxtime_pts = round(maxtime / dt)

t_eval = np.linspace(0, maxtime, maxtime_pts + 1)

rng = np.random.default_rng(seed=1)
width = 1.0
xran = rng.normal(0.0, width, maxtime_pts + 1)
yran = rng.normal(0.0, width, maxtime_pts + 1)
zran = rng.normal(0.0, width, maxtime_pts + 1)

xran_i = interp1d(t_eval, xran)
yran_i = interp1d(t_eval, yran)
zran_i = interp1d(t_eval, zran)

sigma = 10
beta = 8 / 3
rho = 28

def lorenz(t, y):
    dy0 = sigma * (y[1] - y[0]) + xran_i(t)
    dy1 = y[0] * (rho - y[2]) - y[1] + yran_i(t)
    dy2 = y[0] * y[1] - beta * y[2] + zran_i(t)
    return [dy0, dy1, dy2]

t_val = 10.5
y_val = [1.0, 2.0, 3.0]

ARGS = {
    "t": t_val,
    "y": y_val
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
lorenz Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVARtimedelay-RK23.py
code served (permissive licence) · get_code("8a5225640d787a1b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The lorenz function takes two parameters: t (time) and y (state vector)
# t is a scalar (time value)
# y is a sequence/list of 3 values representing the state [y0, y1, y2]

t = 0.0

y = [1.0, 1.0, 1.0]

ARGS = {
    't': t,
    'y': y
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
lorenz_nn Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVAR-Noise.py
code served (permissive licence) · get_code("0d3396b677068fb6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function: lorenz_nn takes two parameters
# - t: time value (appears to be used as context, not directly accessed)
# - y: a list/array with at least 3 elements that support indexing and arithmetic

# Create minimal valid arguments
t = 0.0  # time parameter as a scalar

# y needs to be indexable with at least 3 elements
# The function accesses y[0], y[1], y[2]
y = [1.0, 1.0, 1.0]  # Initial condition for Lorenz system

ARGS = {
    't': t,
    'y': y
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
return_map_spline Ran
ran on a synthesized input
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVARtimedelayReturnMap-RK23.py
code served (permissive licence) · get_code("256ed8f20cd322cd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (0, 2) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import scipy.interpolate

# Construct a small numpy array that will work with the function
# The function expects an array-like object with len() and indexing support
# It will create a spline with k=4 (cubic spline), so we need at least 5 points
v = np.array([1.0, 2.5, 3.2, 2.8, 1.5, 0.9, 1.2, 2.0], dtype=np.float64)

ARGS = {
    'v': v
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (0, 2) — shape (0, 2) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
do_nvar Not yet run
unverified (harvested, no recorded run)
quantinfo/ng-rc-paper-code/DoubleScrollNVAR-RK23.py
code served (permissive licence) · get_code("a2bf0edbaebd863c")
How it ran

no invocation recorded.

find_err Not yet run
unverified (harvested, no recorded run)
quantinfo/ng-rc-paper-code/LorenzConstLinQuadraticNVARtimedelayReturnMap-RK23.py
code served (permissive licence) · get_code("4beb5179f174f579")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Reservoir computing is a best-in-class machine learning algorithm for processing information generated by dynamical systems using observed time-series data. Importantly, it requires very small training data sets, uses linear optimization, and thus requires minimal computing resources. However, the algorithm uses randomly sampled matrices to define the underlying recurrent neural network and has a multitude of metaparameters that must be optimized. Recent results demonstrate the equivalence of reservoir computing to nonlinear vector autoregression, which requires no random matrices, fewer metaparameters, and provides interpretable results. Here, we demonstrate that nonlinear vector autoregression excels at reservoir computing benchmark tasks and requires even shorter training data sets and training time, heralding the next generation of reservoir computing.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2106.07688")
get_code_for_paper("2106.07688")
have("2106.07688")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2106.07688.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2106.07688.svg)](https://syntology.ai/paper/2106.07688)

Connect an agent — have() is free.