v0.1.7New: Bayesian models + AutoPipeline

Train. Evaluate.
Improve.

An integral machine learning library: supervised, unsupervised, reinforcement and neural models behind one unified API, with automatic pipelines, evaluation and tuning. The real library runs in your browser on the playground: no install required.

$pip install synaptix

Try the Playground Get Started GitHub

12Modules

24+Models

5Datasets

v0.1.7Version

AutoMLAutoPipeline

Built-in Modules

Twelve modules covering the full ML workflow — from raw data to a tuned, evaluated and exported model.

Pipelines

AutoPipeline: automatic ML end-to-end — detects the task, cleans, compares models, tunes and evaluates. Plus declarative step-based Pipelines.

autopipelineautomltune

View documentation

Supervised Models

8 regressors and 7 classifiers (random forest, boosting, SVM, KNN...) with one unified fit/predict/evaluate API.

classificationregressionsklearn

View documentation

Unsupervised

KMeans (elbow + silhouette), DBSCAN, hierarchical, Gaussian mixtures, PCA and TSNE for clustering and dimensionality reduction.

clusteringpcadbscan

View documentation

Bayesian Models

Bayesian ridge, ARD, Gaussian processes, Naive Bayes variants and full PyMC models with prediction intervals.

bayesianpymcuncertainty

View documentation

Neural Networks

MLP, CNN and LSTM networks over Keras with training history plots and forecasting (requires synaptix[dl]).

mlpcnnlstm

View documentation

Reinforcement

Q-Learning, SARSA and DQN agents over the GridWorld environment — tabular and deep RL in one API.

qlearningsarsadqn

View documentation

Metrics

Regression (MAE/RMSE/R2/MAPE), classification (accuracy/precision/recall/F1) and clustering metrics with full reports.

evaluatef1r2

View documentation

Model Selection

k-fold cross-validation, GridSearch tuning and compare_models — AutoML-lite across 7 classifiers or 7 regressors.

cvgridsearchcompare

View documentation

Preprocessing

DataCleaner, Scaler, Encoder, Imputer, outlier detection/removal and train_test_split — transformers keep column names.

cleanscaleencode

View documentation

Datasets

Bundled datasets (iris, penguins, titanic, S&P 500, course completion) with the same loader signature as StatsLibX and ViewX.

loadpolarscsv

View documentation

Ecosystem

Export results to ViewX dashboards and run statistical diagnostics of models with StatsLibX — each library also works standalone.

viewxstatslibxpayload

View documentation

Quick Start

One line from raw DataFrame to a trained, evaluated model.

example.py

python

import synaptix as sx

df = sx.load_dataset("iris")

# Automatic ML: detect task -> clean -> compare -> tune -> evaluate
auto = sx.AutoPipeline(task="classification")
auto.fit(df, target="species")
print(auto.evaluate())
print(auto.best_model_)

# Or manual: train, evaluate, diagnose
model = sx.supervised.RandomForestClassifier(n_estimators=100)
model.fit(*sx.train_test_split(df.drop(columns="species"), df["species"])[:2])
model.evaluate(df.drop(columns="species"), df["species"], plot=True)
model.diagnostics(df.drop(columns="species"), df["species"])   # via statslibx

Built-in Datasets

Ready-to-use datasets for immediate training and evaluation.

iris.csv

penguins.csv

titanic.csv

sp500_companies.csv

course_completion.csv

Ready to Train?

The playground loads the real synaptix (with scikit-learn) in your browser — train models and evaluate them, zero installation.

Open Playground

Train. Evaluate.Improve.

Built-in Modules

Pipelines

Supervised Models

Unsupervised

Bayesian Models

Neural Networks

Reinforcement

Metrics

Model Selection

Preprocessing

Datasets

Ecosystem

Quick Start

Built-in Datasets

Ready to Train?

Train. Evaluate.
Improve.