SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1612.05231 · 2016

Tunable Efficient Unitary Neural Networks (EUNN) and their application to RNNs

arXiv · PDF · Open in the Atlas

Code that ran

11 of 22 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, 1 violated it and 9 have no contract verdict; 1 failure of our instrument; 11 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/1612.05231.json · badge: /api/ran/1612.05231.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
iguanaus/EUNN-theano canonical (classifier label)
classifier basis: census
2 ran · 3 no recorded run
solgaardlab/neurophox community (classifier label)
classifier basis not recorded on the link
7 ran · 8 no recorded run
flaport/torch_eunn reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
checkerboard_to_param Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/decompositions.py
code served (permissive licence) · get_code("2f5ba97abdeb258c")
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 (4, 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

# The function checkerboard_to_param takes:
# - checkerboard: np.ndarray
# - units: int

# Looking at the function body:
# - checkerboard is indexed with .T (transpose) and sliced
# - checkerboard.T[::2, :-1:2] or checkerboard.T[::2, ::2] or checkerboard.T[1::2, 1::2]
# - param is (units, units // 2)
# - For checkerboard.T to be indexed, checkerboard must be 2D
# - checkerboard.T has shape where first dim is even rows and second dim has columns

# Let's work backwards:
# If units = 4, then param.shape = (4, 2)
# param[::2, :] gets rows 0, 2 -> shape (2, 2)
# param[1::2, :] gets rows 1, 3 -> shape (2, 2)
# 
# checkerboard.T[::2, ::2] should give (2, 2) for the even case
# checkerboard.T[1::2, 1::2] should give (2, 2)
# 
# So if checkerboard.T has shape (4, 4), then:
# checkerboard.T[::2, ::2] is (2, 2) ✓
# checkerboard.T[1::2, 1::2] is (2, 2) ✓
# checkerboard shape would be (4, 4)

# Let's verify with units=4 (even case):
# param shape = (4, 2)
# param[::2, :] shape = (2, 2) -> needs checkerboard.T[::2, ::2] to be (2, 2)
# param[1::2, :] shape = (2, 2) -> needs checkerboard.T[1::2, 1::2] to be (2, 2)
# So checkerboard.T needs to be at least (4, 4), meaning checkerboard is (4, 4)

# Let's try units=4 with checkerboard of shape (4, 4)
units = 4
checkerboard = np.arange(16).reshape(4, 4).astype(float)

ARGS = {
    "checkerboard": checkerboard,
    "units": units
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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 (4, 2) — shape (4, 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
cmm Ran
ran, violated the contract we drafted
flaport/torch_eunn/torch_eunn.py
code served (permissive licence) · get_code("2d824fe3cb24d6a2")
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 Tensor, ndim 4, dtype family float, shape (2, 3, 4, 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 torch

torch.manual_seed(0)

# Create two complex tensors with shape (2, 3, 4, 2)
# where the last dimension stores [real, imaginary] parts
x = torch.randn(2, 3, 4, 2)
y = torch.randn(2, 3, 4, 2)

ARGS = {
    "x": x,
    "y": y,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 4, 2) — shape (2, 3, 4, 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
complex_mse Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/ml/linear.py
code served (permissive licence) · get_code("a7c7c0f896248555")
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 EagerTensor, ndim 1, dtype family float, shape (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 tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)
np.random.seed(0)

# Create small complex tensors
batch_size = 2
num_features = 4

# Create y_true as a complex tensor (B x N)
y_true_real = tf.constant(np.random.randn(batch_size, num_features), dtype=tf.float32)
y_true_imag = tf.constant(np.random.randn(batch_size, num_features), dtype=tf.float32)
y_true = tf.cast(y_true_real, tf.complex64) + 1j * tf.cast(y_true_imag, tf.complex64)

# Create y_pred as a complex tensor (B x N)
y_pred_real = tf.constant(np.random.randn(batch_size, num_features), dtype=tf.float32)
y_pred_imag = tf.constant(np.random.randn(batch_size, num_features), dtype=tf.float32)
y_pred = tf.cast(y_pred_real, tf.complex64) + 1j * tf.cast(y_pred_imag, tf.complex64)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 1, dtype family float, shape (2,) — shape (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
eunn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
flaport/torch_eunn/torch_eunn.py
code served (permissive licence) · get_code("6f86da6493673239")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 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 torch

torch.manual_seed(0)

# Construct angles tensor: shape (hidden_size, capacity) where hidden_size is even
hidden_size = 4
capacity = 3
angles = torch.randn(hidden_size, capacity, dtype=torch.float32)

# Construct x tensor: shape (batch_size, hidden_size, 2) for complex numbers
batch_size = 2
x = torch.randn(batch_size, hidden_size, 2, dtype=torch.float32)

ARGS = {
    "angles": angles,
    "x": x,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 4, 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
get_haar_diagonal_sequence Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/helpers.py
code served (permissive licence) · get_code("3268829a2be85b3d")
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 1, dtype family int, shape (5,) 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

# Construct arguments for get_haar_diagonal_sequence
ARGS = {
    "diagonal_length": 5,
    "parity_odd": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (5,) — shape (5,) 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
get_mzi_transfer_matrix Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/components/mzi.py
code served (permissive licence) · get_code("58bc45ee4257a1b3")
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 (2, 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

# Define arguments for get_mzi_transfer_matrix
ARGS = {
    "internal_upper": 0.5,
    "internal_lower": 0.3,
    "external_upper": 0.2,
    "external_lower": 0.1,
    "hadamard": True,
    "epsilon": (0.01, 0.02),
    "dtype": np.complex128,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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 (2, 2) — shape (2, 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
get_tdc_transfer_matrix Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/components/mzi.py
code served (permissive licence) · get_code("eda08eb52c0cc327")
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 (2, 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

# Define the arguments dictionary for get_tdc_transfer_matrix
ARGS = {
    "kappa": 0.5,
    "delta": 0.3,
    "external_upper": 0.1,
    "external_lower": 0.2,
    "dtype": np.complex128
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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 (2, 2) — shape (2, 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
gradient_descent Ran
ran on a synthesized input
iguanaus/EUNN-theano/optimizations.py
code served (permissive licence) · get_code("f0abdf7c350a7acc")
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. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for parameters
param1 = torch.randn(4, 3, requires_grad=True)
param2 = torch.randn(2, requires_grad=True)
parameters = [param1, param2]

# Create corresponding gradients
grad1 = torch.randn(4, 3)
grad2 = torch.randn(2)
gradients = [grad1, grad2]

# Set learning rate
learning_rate = 0.01

ARGS = {
    "learning_rate": learning_rate,
    "parameters": parameters,
    "gradients": gradients,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list 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
normalized_fidelity Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/ml/linear.py
code served (permissive licence) · get_code("9a1cc22fc4bebe7b")
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 EagerTensor, ndim 0, dtype family float, shape () 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 tensorflow as tf
import numpy as np

# Create a small unitary matrix (2x2)
# Using a simple unitary: [[1, 0], [0, 1]] scaled appropriately
u_real = np.array([[1.0, 0.0], [0.0, 1.0]], dtype=np.complex64)
u = tf.constant(u_real, dtype=tf.complex64)

# Create an estimated unitary (not necessarily unitary)
# A 2x2 complex matrix that approximates the true unitary
u_hat_real = np.array([[0.95 + 0.1j, 0.05], [0.05, 0.95]], dtype=np.complex64)
u_hat = tf.constant(u_hat_real, dtype=tf.complex64)

ARGS = {
    "u": u,
    "u_hat": u_hat,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float, shape () — shape () 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
permute_approx Ran
ran on a synthesized input
iguanaus/EUNN-theano/widget.py
code served (permissive licence) · get_code("b89a8b25a57a4ce1")
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, ndim 3, dtype family int; 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

ARGS = {
    's': 8
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, dtype family int — a tuple 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
to_absolute_theta Ran
ran on a synthesized input
solgaardlab/neurophox/neurophox/helpers.py
code served (permissive licence) · get_code("f69ddd5f6086c06f")
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 1, dtype family float, shape (6,) 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

# Construct a small numpy array of angles (some less than pi, some greater)
theta = np.array([0.5, 2.0, 4.0, 5.5, 7.0, 8.5])

ARGS = {
    "theta": theta
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (6,) — shape (6,) 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
bptrs Not yet run
unverified (harvested, no recorded run)
iguanaus/EUNN-theano/fftconv.py
code served (permissive licence) · get_code("0f9745926384036b")
How it ran

no invocation recorded.

cnorm Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/ml/nonlinearities.py
code served (permissive licence) · get_code("b03f07ab4d8a044b")
How it ran

no invocation recorded.

cnormsq Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/ml/nonlinearities.py
code served (permissive licence) · get_code("b0024b3700fafa76")
How it ran

no invocation recorded.

generate_keras_batch Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/ml/linear.py
code served (permissive licence) · get_code("2e25840573fd6123")
How it ran

no invocation recorded.

get_haar_theta Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/initializers.py
code served (permissive licence) · get_code("fe084038a9330dac")
How it ran

no invocation recorded.

get_initializer Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/initializers.py
code served (permissive licence) · get_code("d958ae5b8ac823b4")
How it ran

no invocation recorded.

get_ortho_haar_theta Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/initializers.py
code served (permissive licence) · get_code("2947f55793f40d77")
How it ran

no invocation recorded.

permute_list12 Not yet run
unverified (harvested, no recorded run)
iguanaus/EUNN-theano/widget.py
code served (permissive licence) · get_code("e48f79c125c0bb57")
How it ran

no invocation recorded.

reck_decomposition Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/decompositions.py
code served (permissive licence) · get_code("7f653a85312e6df6")
How it ran

no invocation recorded.

to_complex_gpuarray Not yet run
unverified (harvested, no recorded run)
iguanaus/EUNN-theano/fftconv.py
code served (permissive licence) · get_code("71587416e135f94f")
How it ran

no invocation recorded.

to_stripe_array Not yet run
unverified (harvested, no recorded run)
solgaardlab/neurophox/neurophox/helpers.py
code served (permissive licence) · get_code("c9c896b7297dde56")
How it ran

no invocation recorded.

Also present in other papers' code

Of 22 samples lifted from this paper's linked repositories, 22 carry a function identity and 0 do not. Of the 22 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 22 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

Using unitary (instead of general) matrices in artificial neural networks (ANNs) is a promising way to solve the gradient explosion/vanishing problem, as well as to enable ANNs to learn long-term correlations in the data. This approach appears particularly promising for Recurrent Neural Networks (RNNs). In this work, we present a new architecture for implementing an Efficient Unitary Neural Network (EUNNs); its main advantages can be summarized as follows. Firstly, the representation capacity of the unitary space in an EUNN is fully tunable, ranging from a subspace of SU(N) to the entire unitary space. Secondly, the computational complexity for training an EUNN is merely $\mathcal{O}(1)$ per parameter. Finally, we test the performance of EUNNs on the standard copying task, the pixel-permuted MNIST digit recognition benchmark as well as the Speech Prediction Test (TIMIT). We find that our architecture significantly outperforms both other state-of-the-art unitary RNNs and the LSTM architecture, in terms of the final performance and/or the wall-clock training speed. EUNNs are thus promising alternatives to RNNs and LSTMs for a wide variety of applications.

For agents

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

get_harvested_code_for_paper("1612.05231")
get_code_for_paper("1612.05231")
have("1612.05231")

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

curl https://syntology.ai/api/ran/1612.05231.json

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

[![Syntology run record](https://syntology.ai/api/ran/1612.05231.svg)](https://syntology.ai/paper/1612.05231)

Connect an agent — have() is free.