#!/usr/bin/env python3
from seed_common import run_seed

# Generate 500 distinct model descriptions via combinatorics

def gen_math_models(n: int):
    families = [
        "Bayesian", "Frequentist", "Gaussian Process", "Kalman Filter", "Hidden Markov",
        "ARIMA", "GARCH", "Markov Decision Process", "Monte Carlo", "Variational Inference",
        "MCMC", "Principal Component", "Independent Component", "t-SNE", "UMAP",
        "Support Vector Machine", "Random Forest", "Gradient Boosting", "XGBoost", "LightGBM",
        "CatBoost", "Elastic Net", "Lasso", "Ridge", "K-Means",
        "DBSCAN", "Spectral Clustering", "Hierarchical Clustering", "Nonnegative Matrix Factorization",
        "Autoencoder", "Variational Autoencoder", "GAN", "Diffusion Model", "Transformer",
        "Graph Neural Network", "Temporal Convolutional Network", "Long Short-Term Memory",
        "Convolutional Neural Network", "Bayesian Optimization", "Multi-Armed Bandit",
    ]
    tasks = [
        "binary classification", "multiclass classification", "regression", "time-series forecasting",
        "anomaly detection", "causal inference", "counterfactual estimation", "dimensionality reduction",
        "topic modeling", "recommendation", "segmentation", "clustering", "survival analysis",
        "reinforcement learning", "policy optimization", "control", "simulation",
    ]
    qualifiers = [
        "hierarchical", "conjugate prior", "robust", "sparse", "regularized", "nonparametric",
        "semi-supervised", "self-supervised", "transfer-learning", "online", "streaming",
        "batched", "stochastic", "deterministic", "Bayes-optimal", "approximate", "ensemble",
    ]
    domains = [
        "vision", "NLP", "speech", "finance", "healthcare", "operations", "geospatial",
        "energy", "retail", "manufacturing", "security", "recommendation systems",
    ]

    seen = set()
    i = 0
    for f in families:
        for q in qualifiers:
            for t in tasks:
                phrase = f"{f} {q} model for {t}"
                if phrase in seen:
                    continue
                seen.add(phrase)
                yield f"Model: {phrase}"
                i += 1
                if i >= n:
                    return
    # If not enough combinations, add domain variants
    if i < n:
        for f in families:
            for d in domains:
                phrase = f"{f} approach applied to {d}"
                if phrase in seen:
                    continue
                seen.add(phrase)
                yield f"Model: {phrase}"
                i += 1
                if i >= n:
                    return

if __name__ == "__main__":
    run_seed("math_model", gen_math_models)