ML gortex example

simplest_optimization.go

package main

import (
	g "github.com/vseledkin/gortex"
	"fmt"
)

func main() {
	X := float32(1)
	Y := float32(1)
	X1 := float32(0)
	Y1 := float32(2)

	Y_vector := g.Mat(1, 1)
	Y_vector.W[0] = Y

	X_vector := g.Mat(1, 1)
	X_vector.W[0] = X

	Y_vector1 := g.Mat(1, 1)
	Y_vector1.W[0] = Y1

	X_vector1 := g.Mat(1, 1)
	X_vector1.W[0] = X1

	parameter := g.Mat(1, 1)
	parameter.W[0] = 0.5

	optimizer := g.NewOptimizer(g.OpOp{Method: g.SGD, LearningRate: 0.01, Momentum: 0.95})
	step := 0
	for true {
		cost := float32(0)
		graph := &g.Graph{NeedsBackprop: true}
		result := graph.Mul(parameter, X_vector)
		cost += graph.MSE(result, Y_vector)
		graph.Backward()

		graph1 := &g.Graph{NeedsBackprop: true}
		result1 := graph1.Mul(parameter, X_vector1)
		cost += graph1.MSE(result1, Y_vector1)

		graph1.Backward()

		optimizer.Step(map[string]*g.Matrix{"parameters": parameter})
		fmt.Printf("%d Current a*x=y %f*%f=%f abs error=%f\n", step, parameter.W[0], X_vector.W[0], result.W[0], cost)
		step++
		if cost/2 < 1e-4 { // /2 because we have two samples
			break
		}
	}
}