Hypercube WTF

This package is the Python surface for HypercubeWTF (import hypercube_wtf). Full API reference: docs/Python_SDK.md. C++ integration guide: docs/CPP_SDK.md. Project home: github.com/dliptak001/HypercubeWTF.

HypercubeWTF processes spatial data of the kind presented to a CNN. It is built from three core classes.

The WTF class wraps the other two and manages training and prediction.

The other two form a pipeline: reservoir → readout.

The Reservoir class is a preprocessing stage that consumes input patterns, drives a short synthetic orbit on a frozen hypercube reservoir, and returns a field with the same dimensions as the input.

The Readout class is a small HypercubeCNN that classifies or regresses that field.

This is reservoir computing, aimed at data that has no time.

The point of this experiment is to see if a preprocessing stage in front of HypercubeCNN outperforms HypercubeCNN by itself. HypercubeEtalon has the same goal; it just does it a slightly different way, using an etalon transit with no time at all, whereas here the preprocessor is a reservoir with synthetic time. The aim is a hypercube preprocessor effective enough that the readout can be a single layer with a single convolutional channel and no pooling. Then training is fast, the memory footprint is small, and little to no architectural engineering is required for the CNN.

---

HypercubeAI ecosystem

HypercubeCascade  ·  HypercubeCNN  ·  HypercubeESN  ·  HypercubeEtalon  ·  HypercubeHopfield  ·  HypercubeLCN  ·  HypercubeWorldModel  ·  HypercubeWTF

📄 Foundational paper: Boolean Hypercubes as a Neural Substrate (D. C. Liptak, 2026)

HypercubeWTF is an experiment in the HypercubeAI project — our quest to systematically re-implement classical neural architectures on a Boolean hypercube topology instead of Euclidean grids or random graphs. The central thesis is “topology-native intelligence”: the hypercube’s algebraic structure (vertex-transitive symmetry, Hamming geometry, bitwise addressing) can serve as a first-class computational substrate.

• A topology you don’t store — the graph is specified: connectivity is implicit in the vertex indices; with a seed and a few config scalars the whole reservoir reconstructs mathematically.
• Perfect homogeneity — every vertex has the same degree and the same local world, so local dynamics mean the same thing everywhere — no structural favorites baked in by a random graph.
• Cheap navigation — each neighbor is a few bit operations on the vertex index, not a pointer chase through a stored edge list, so walks stay arithmetic and cache-friendly.
• Topology-native pairing — the readout consumes the reservoir’s output with zero geometric distortion, and the learned kernels exploit the same locality that generated the dynamics. The data never leaves the hypercube it was born on.

Each product in the family is a different architecture on that same foundation.

---

What does WTF stand for?

The design goal was simple: take the HypercubeESN idea — frozen reservoir, trained head — and aim it at data that has no time. There was no lineage to steal a name from, so the usual naming exercise followed. Nothing stuck. After a few hours of “maybe this?” and “nah.”, the working monologue devolved to what the f*** do we call this project?

So we called it that.

HypercubeWTF

The monologue won — and the brand gained a little personality :-)

---

The Reservoir

HypercubeESN and HypercubeCNN are two examples of how solutions can be built on that substrate. HypercubeESN drives a frozen hypercube reservoir with a stream and reads the state out along the way. HypercubeCNN trains a convolutional stack directly on a static cube field. WTF sits between them: the ESN's reservoir, aimed at the CNN's data.

The Reservoir is a fork of the HypercubeESN core: one neuron per vertex, frozen recurrent weights over the cube's edges, a delay line of M slices, tanh activation, and a frozen initial condition that is reloaded before every sample. Nothing in it is ever trained.

A stream has a next sample. A still field does not. So WTF invents a short stretch of synthetic time: it re-addresses the same fixed field over the cube for T passes and samples the reservoir once at the end. Geometry and weights stay put; only the registration of the field moves. The orbit goes something like this.

Leave the caller's field alone. The drive is built in a scratch
buffer.

Reload the reservoir's frozen initial condition.

LOOP:

    Remap the field by xor with the pass index: vertex v is driven
    by the field value at v xor c.

    Inject that remapping. Step the reservoir: every vertex forms
    the weighted sum of its neighbors and its drive, and writes
    tanh of that sum.

    Increment the pass index.

GOTO LOOP

After T passes, the reservoir's live output is the feature field.
That is what the Readout sees.

Every episode starts from the same frozen initial condition, so the feature field depends on the input field and nothing else. Bulk collection fans independent episodes across worker reservoirs that share the frozen weights.

---

White noise filter

The reservoir preprocessor behaves as a near unity passthrough at low to no white noise levels, and offers a meaningful filtering effect at moderate to high noise levels. The write-up is examples/mnist/WhiteNoiseFilter.md.

---

Training-data quality

The same orbit also softens the blow of a degraded training set. With heavy white noise on the test fields, corrupting the training set costs the pack-only path about 19 points of test accuracy and the reservoir path about 8. On clean test fields both paths lose about a point. The write-up is examples/mnist/TrainingDataQualitySensitivity.md.

Both studies use MNIST on small cubes because it is handy to pack and run, not because we are chasing digit accuracy.

---

Raman baseline extraction (a vibrational spectroscopy application)

The first real-world test is Raman spectra: recover the slow fluorescence background under sharp molecular peaks without lifting the baseline into the bands or cutting trenches beneath them. Polynomials, asymmetric least squares, and ordinary convolutional nets tend to follow the empty stretches well and then fail where it matters, under peaks and peak clusters. Analysts have worked around that for decades with spectrum-specific cleanup, because no method identifies and extracts a true baseline across a broad range of peak intensities and baseline characteristics without occasional, and often frequent, human intervention.

WTF appears to have solved that problem (albeit on synthetic data only so far).

Trained for 60 epochs on the LCOHard set — 10,000 synthetic LiCoO₂ (lithium cobalt oxide) spectra — it scores a validation RMSE of 4.76 counts on 2,000 held-out spectra whose baselines span hundreds of counts.

Below are four held-out validation spectra: grey is the raw spectrum, red the true baseline, blue the extract. For all four shown here, and for each of the remaining 1996 validation spectra not shown, baseline identification is, WITHOUT EXCEPTION, quite remarkable.

And it does this with the thin readout the project aims for: one HypercubeCNN layer, one convolutional channel, no pooling.

In our judgment this at least matches the best of the established techniques on spectra like these, and very likely beats them.

Three hosts, one floor

The Reservoir is the whole preprocessor here: one orbit, then the readout. It is the Cascade's second stage run alone — same seed, same spectral radius, same history depth, same pass count — and on spectra like these it is already enough. The etalon-only sibling (HypercubeEtalon) scores 4.77 on the same split; the two-stage (HypercubeCascade) scores 4.82. Three preprocessors that share no mechanism — a transit, an orbit, and the two in series — carry the same one-layer, one-channel readout to the same floor, and their overlays are indistinguishable.

Real spectra, however, are not nearly this clean. Low laser power, short integration times, and weak scatterers all put noise on the spectrum, and that is where a baseline extractor has to earn its keep.

That is where the hosts should separate. The MNIST white-noise study (examples/mnist/WhiteNoiseFilter.md) found this reservoir a near-unity passthrough on clean fields and a filter that holds accuracy as the noise rises; the Cascade's study found the two-stage path pulling ahead of the etalon alone from moderate noise up.

That is the next experiment.

The overlay and the training profile are in examples/RamanBaselineExtraction/.

Runnable programs live under examples/.

The Raman spectra themselves (about 1 GB) are not in this repository.

---

Installation

Preferred: install a pre-built wheel from PyPI (no compiler).

pip install hypercube-wtf

import hypercube_wtf as hw
print(hw.__version__)

Package name on PyPI: hypercube-wtf. Import name: hypercube_wtf. Main type: hw.WTF.

Wheels target Python 3.10–3.14 on common Windows, Linux, and macOS machines. Runtime dependency: NumPy only.

From source (full repository)

To compile the extension yourself, clone this entire repository (not a minimal source-only download of the python/ folder alone — the C++ core and vendored HypercubeCNN live next to python/). You need Python 3.10+, a C++23 compiler, and CMake ≥ 3.20.

git clone https://github.com/dliptak001/HypercubeWTF.git
cd HypercubeWTF/python
pip install .

On Windows with CLion’s MinGW, put that compiler’s bin folder (and Ninja) on your PATH, then:

pip install . --no-build-isolation --force-reinstall --no-deps

(Exact CLion paths change with the version.) Step-by-step toolchain notes: docs/Python_SDK.md.

---

Quick start

You bring each sample as a length-N float array (N = 2^dim). How you get there — pad an image, reshape a spectrum, invent a layout — is up to you. This package does not pack 784 pixels or 300 bins for you.

Shapes that matter:

Array	Shape	Notes
fields	(count, N)	one length-N field per row
labels (classification)	(count,)	integer class indices
targets (regression)	(count, num_outputs)	float targets

import numpy as np
import hypercube_wtf as hw

dim = 7
N = 2**dim
rng = np.random.default_rng(0)
fields = rng.standard_normal((200, N), dtype=np.float32)
labels = rng.integers(0, 4, size=200)

wtf = hw.WTF(
    dim=dim,
    history_depth=4,
    T=100,
    ic_seed=2,
    readout_num_outputs=4,
    readout_task="classification",
    readout_epochs=80,
)
wtf.fit(fields, labels)  # collect_episodes + train

print(wtf.N, wtf.T, wtf.num_collected)
print(f"train sanity check: {wtf.accuracy_on_collected():.3f}")
print(wtf.predict_class(fields[0]), wtf.predict(fields[0]).shape)

wtf.save("model.pkl")
loaded = hw.WTF.load("model.pkl")

Step by step (same loop, more control)

wtf = hw.WTF(
    dim=7,
    readout_num_outputs=4,
    readout_task="classification",
)
wtf.collect_episodes(fields_train, labels_train)
wtf.train()
logits = wtf.predict(fields_test[0])       # (num_outputs,) float32
cls = wtf.predict_class(fields_test[0])    # int

For regression, set readout_task="regression" and pass float targets instead of class labels. Then use r2_on_collected() the same way as the classification sanity check.

accuracy_on_collected and r2_on_collected only look at the samples you already trained on — they are a quick sanity check, not a test score. For real evaluation, hold some fields out and call predict / predict_class yourself.

---

Features

• One class — hypercube_wtf.WTF is the whole product surface
• Episode loop — collect_episode / collect_episodes → train → predict / predict_class
• fit — clear, collect, and train when your arrays are ready
• dim 5–16 — field length N = 2^dim; orbit length T; end-of-orbit ages readout_slices (B)
• Classification or regression — readout_task fixed at construction
• Bulk collect can parallelize — collect_threads (0 = auto)
• Train-only field noise — train_input_noise_sigma on collect, never on predict
• Skip-the-orbit path — bypass_reservoir=True for pack-only comparisons (needs B = 1)
• Inspect an episode — run_episode(x) then last_features()
• Save / load — save / load (pickle: config + readout weights; collected samples are not stored). Optional save_readout_hcnn_model / load_readout_hcnn_model for portable HCNW + arch JSON
• NumPy float32 — arrays converted for you; prefer contiguous float32

---

Examples

For a first try, paste the Quick start after pip install hypercube-wtf. That is self-contained.

If you want a longer walk-through, the demo scripts on GitHub under python/examples/ are there to open or download — they are not added to your machine by pip.

Script	What it is for
synthetic_classification.py	Multi-class toy fields: fit, then train and test accuracy

# from a clone of HypercubeWTF, after: pip install hypercube-wtf
python python/examples/synthetic_classification.py

These use easy made-up fields so the API is obvious — not scores to publish. More notes: python/examples/README.md.

---

Documentation

Doc	Role
docs/Python_SDK.md	Canonical Python API — every method, layout, pickle, limits
python/examples/README.md	Demo scripts on GitHub
Project README	Product story and C++ demos from the repo root
docs/CPP_SDK.md	Native library guide (same product, C++)
WhiteNoiseFilter.md	Early white-noise study (MNIST as a test bed)
TrainingDataQualitySensitivity.md	Early training-quality study (MNIST as a test bed)

---

Ecosystem

• HypercubeCascade: two frozen stages in series plus a thin readout.
• HypercubeCNN: cube-native conv stack with shared kernels.
• HypercubeESN: echo-state reservoir computing on streams.
• HypercubeEtalon: frozen etalon transit plus a thin readout.
• HypercubeHopfield: Hopfield-style dynamics on the cube.
• HypercubeLCN: the locally connected net, every weight trained.
• HypercubeWorldModel: frozen WTF encoder + trained LCN predictor and decoder.
• HypercubeWTF: frozen reservoir orbit plus a thin readout.

---

License

Apache 2.0. See LICENSE.