@astermind/astermind-elm/dist/tasks/ConfidenceClassifierELM.d.ts

JavaScript Extreme Learning Machine (ELM) library for browser and Node.js.

import { ELM } from '../core/ELM';
import { ELMConfig, Activation } from '../core/ELMConfig';
export type ConfidenceLabel = 'low' | 'high';
export interface ConfidenceClassifierOptions {
    /** Class names to use (default: ['low','high']) */
    categories?: [ConfidenceLabel, ConfidenceLabel] | string[];
    /** Activation for hidden units (default: 'relu') */
    activation?: Activation;
    /** Verbose console logs */
    verbose?: boolean;
    /** Export filename for saved JSON (optional) */
    exportFileName?: string;
}
/**
 * ConfidenceClassifierELM is a lightweight wrapper that classifies whether
 * an upstream model’s prediction is "low" or "high" confidence based on
 * (embedding, metadata) numeric features.
 */
export declare class ConfidenceClassifierELM {
    private baseConfig;
    private elm;
    private categories;
    private activation;
    constructor(baseConfig: ELMConfig, opts?: ConfidenceClassifierOptions);
    /** One-hot helper */
    private oneHot;
    /**
     * Train from numeric (vector, meta) → combined features + labels.
     * `vectors[i]` and `metas[i]` must be aligned with `labels[i]`.
     */
    train(vectors: number[][], metas: number[][], labels: string[]): void;
    /** Predict full distribution for a single (vec, meta). */
    predict(vec: number[], meta: number[], topK?: number): Array<{
        label: string;
        prob: number;
    }>;
    /** Probability the label is "high" (or the second category by default). */
    predictScore(vec: number[], meta: number[], positive?: string): number;
    /** Predicted top-1 label. */
    predictLabel(vec: number[], meta: number[]): string;
    /** Batch prediction (distributions). */
    predictBatch(vectors: number[][], metas: number[][], topK?: number): Array<Array<{
        label: string;
        prob: number;
    }>>;
    /** Compute accuracy and confusion counts for a labeled set. */
    evaluate(vectors: number[][], metas: number[][], labels: string[]): {
        accuracy: number;
        confusion: Record<string, Record<string, number>>;
    };
    loadModelFromJSON(json: string): void;
    saveModelAsJSONFile(filename?: string): void;
    /** Access underlying ELM if needed */
    getELM(): ELM;
    /** Current category ordering used by the model */
    getCategories(): string[];
}