UNPKG

@andypai/neuroflow

Version:

simple neural network library inspired by karpathy/micrograd and tfjs

27 lines (24 loc) 728 B
import Sequential from '../sequential.js' import Layer from '../layer.js' import Neuron from '../neuron.js' import Value from '../engine.js' const bootstrapModel = (structure, activation = 'softmax') => { const model = new Sequential({ layers: structure.map((layer, index) => { const isLast = index === structure.length - 1 return new Layer({ neurons: layer.map( ({ weights, bias }) => new Neuron({ weights: weights.map((w) => new Value(w)), bias, activation: isLast ? activation : 'relu', }), ), activation: isLast ? activation : 'relu', }) }), }) return model } export default bootstrapModel