TensorFlow2-实战（一）：回归问题（手写梯度下降，不用任何模型）、步骤：【初始化参数】--＞【循环（①计算Loss；②计算梯度；③利用梯度下降来更新参数）】

古城微笑少年丶 2023-09-29 10:19 6阅读 0赞

import numpy as np
    
    # y = wx + b
    def compute_loss_for_line_given_points(w, b, points):
        totalError = 0
        for i in range(0, len(points)):
            x = points[i, 0]
            y = points[i, 1]
            # computer mean-squared-error
            totalError += (y - (w * x + b)) ** 2
        # average loss for each point
        avg_loss = totalError / float(len(points))
        return avg_loss
    
    
    def step_gradient(step, w_current, b_current, points, learningRate):
        w_gradient = 0
        b_gradient = 0
        N = float(len(points))
        for i in range(0, len(points)):
            x = points[i, 0]
            y = points[i, 1]
            # 三、计算梯度
            w_gradient += (2 / N) * x * ((w_current * x + b_current) - y)  # grad_w = 2(wx+b-y)*x
            b_gradient += (2 / N) * ((w_current * x + b_current) - y)  # grad_b = 2(wx+b-y)
        # 四、利用梯度下降来更新参数
        new_w = w_current - (learningRate * w_gradient)  # update w'
        new_b = b_current - (learningRate * b_gradient)  # update b'
        # 计算 loss
        curr_loss = compute_loss_for_line_given_points(new_w, new_b, points)
        print("\tstep = {0}----当前位置: w_current = {1}, b_current = {2}----计算w、b在当前位置的梯度: w' = {3}, b' = {4}----利用梯度下降更新参数后: new_w = {5}, new_b = {6}----参数更新后的loss: curr_loss = {7}".format(step, w_current, b_current, w_gradient, b_gradient, new_w, new_b, curr_loss))
        return new_w, new_b
    
    
    def gradient_descent_runner(points, initial_w, initial_b, learning_rate, num_iterations):
        w = initial_w
        b = initial_b
        # update for several times
        for step_idx in range(num_iterations):
            w, b = step_gradient(step_idx, w, b, np.array(points), learning_rate)
        return w, b
    
    
    def run():
        points = np.genfromtxt("data.csv", delimiter=",")
        learning_rate = 0.0001
    
        # 一、初始化参数
        initial_w = 0  # initial slope guess
        initial_b = 0  # initial y-intercept guess
        num_iterations = 100
        print("\n初始化参数：initial_w = {0}; initial_b = {1}".format(initial_w, initial_b))
    
        # 二、计算初始 loss
        initial_loss = compute_loss_for_line_given_points(initial_w, initial_b, points)
        print("\n初始时的loss: initial_loss = {0}".format(initial_loss))
    
        # 三、计算梯度 & 四、更新参数
        final_w, final_b = gradient_descent_runner(points, initial_w, initial_b, learning_rate, num_iterations)
    
        # 五、计算最终 loss
        final_loss = compute_loss_for_line_given_points(final_w, final_b, points)
        print("最终时的loss: final_loss = {0}".format(final_loss))
    
        print("\n经过{0}轮迭代后: w = {1}, b = {2}, final_loss = {3}".format(num_iterations, final_w, final_b, final_loss))
    
    
    if __name__ == '__main__':
        run()

打印结果：

初始化参数：initial_w = 0; initial_b = 0
    
    初始时的loss: initial_loss = 5565.107834483211
    	step = 0----当前位置: w_current = 0, b_current = 0----计算w、b在当前位置的梯度: w' = -7370.702973591052, b' = -145.47010110737295----利用梯度下降更新参数后: new_w = 0.7370702973591052, new_b = 0.014547010110737297----参数更新后的loss: curr_loss = 1484.586557408649
    	step = 1----当前位置: w_current = 0.7370702973591052, b_current = 0.014547010110737297----计算w、b在当前位置的梯度: w' = -3697.2515699241044, b' = -73.26952848859113----利用梯度下降更新参数后: new_w = 1.1067954543515157, new_b = 0.02187396295959641----参数更新后的loss: curr_loss = 457.8542575737673
    	step = 2----当前位置: w_current = 1.1067954543515157, b_current = 0.02187396295959641----计算w、b在当前位置的梯度: w' = -1854.5921056159586, b' = -37.05261361696725----利用梯度下降更新参数后: new_w = 1.2922546649131115, new_b = 0.025579224321293136----参数更新后的loss: curr_loss = 199.50998572553894
    	step = 3----当前位置: w_current = 1.2922546649131115, b_current = 0.025579224321293136----计算w、b在当前位置的梯度: w' = -930.2859073813355, b' = -18.885652378512194----利用梯度下降更新参数后: new_w = 1.385283255651245, new_b = 0.027467789559144355----参数更新后的loss: curr_loss = 134.50591058200533
    	step = 4----当前位置: w_current = 1.385283255651245, b_current = 0.027467789559144355----计算w、b在当前位置的梯度: w' = -466.6397673307541, b' = -9.77282422594607----利用梯度下降更新参数后: new_w = 1.4319472323843205, new_b = 0.028445071981738963----参数更新后的loss: curr_loss = 118.14969342239947
    	step = 5----当前位置: w_current = 1.4319472323843205, b_current = 0.028445071981738963----计算w、b在当前位置的梯度: w' = -234.0677651372032, b' = -5.20168784739655----利用梯度下降更新参数后: new_w = 1.4553540088980408, new_b = 0.02896524076647862----参数更新后的loss: curr_loss = 114.0341490603815
    	step = 6----当前位置: w_current = 1.4553540088980408, b_current = 0.02896524076647862----计算w、b在当前位置的梯度: w' = -117.40608822094528, b' = -2.9087335956808102----利用梯度下降更新参数后: new_w = 1.4670946177201354, new_b = 0.0292561141260467----参数更新后的loss: curr_loss = 112.99857731713661
    	step = 7----当前位置: w_current = 1.4670946177201354, b_current = 0.0292561141260467----计算w、b在当前位置的梯度: w' = -58.886805042409094, b' = -1.7585503776042795----利用梯度下降更新参数后: new_w = 1.4729832982243762, new_b = 0.02943196916380713----参数更新后的loss: curr_loss = 112.7379818756847
    	step = 8----当前位置: w_current = 1.4729832982243762, b_current = 0.02943196916380713----计算w、b在当前位置的梯度: w' = -29.532636718523086, b' = -1.1815986057594474----利用梯度下降更新参数后: new_w = 1.4759365618962286, new_b = 0.029550129024383073----参数更新后的loss: curr_loss = 112.67238435909097
    	step = 9----当前位置: w_current = 1.4759365618962286, b_current = 0.029550129024383073----计算w、b在当前位置的梯度: w' = -14.808136521512168, b' = -0.8921885034931599----利用梯度下降更新参数后: new_w = 1.4774173755483797, new_b = 0.02963934787473239----参数更新后的loss: curr_loss = 112.65585181499746
    	step = 10----当前位置: w_current = 1.4774173755483797, b_current = 0.02963934787473239----计算w、b在当前位置的梯度: w' = -7.422101836095167, b' = -0.7470137049465564----利用梯度下降更新参数后: new_w = 1.4781595857319891, new_b = 0.029714049245227046----参数更新后的loss: curr_loss = 112.65166489759584
    	step = 11----当前位置: w_current = 1.4781595857319891, b_current = 0.029714049245227046----计算w、b在当前位置的梯度: w' = -3.7171538026644315, b' = -0.674189542994775----利用梯度下降更新参数后: new_w = 1.4785313011122556, new_b = 0.029781468199526522----参数更新后的loss: curr_loss = 112.65058436150113
    	step = 12----当前位置: w_current = 1.4785313011122556, b_current = 0.029781468199526522----计算w、b在当前位置的梯度: w' = -1.85869519103733, b' = -0.6376575680589751----利用梯度下降更新参数后: new_w = 1.4787171706313593, new_b = 0.02984523395633242----参数更新后的loss: curr_loss = 112.65028544701505
    	step = 13----当前位置: w_current = 1.4787171706313593, b_current = 0.02984523395633242----计算w、b在当前位置的梯度: w' = -0.9264639430062118, b' = -0.6193303097782138----利用梯度下降更新参数后: new_w = 1.4788098170256598, new_b = 0.02990716698731024----参数更新后的loss: curr_loss = 112.6501832029397
    	step = 14----当前位置: w_current = 1.4788098170256598, b_current = 0.02990716698731024----计算w、b在当前位置的梯度: w' = -0.45884252102788725, b' = -0.6101348161015641----利用梯度下降更新参数后: new_w = 1.4788557012777626, new_b = 0.029968180468920396----参数更新后的loss: curr_loss = 112.65013044507197
    	step = 15----当前位置: w_current = 1.4788557012777626, b_current = 0.029968180468920396----计算w、b在当前位置的梯度: w' = -0.22427650093916096, b' = -0.6055199553740327----利用梯度下降更新参数后: new_w = 1.4788781289278565, new_b = 0.0300287324644578----参数更新后的loss: curr_loss = 112.65009013922885
    	step = 16----当前位置: w_current = 1.4788781289278565, b_current = 0.0300287324644578----计算w、b在当前位置的梯度: w' = -0.10661463849945285, b' = -0.603202810280004----利用梯度下降更新参数后: new_w = 1.4788887903917065, new_b = 0.030089052745485803----参数更新后的loss: curr_loss = 112.65005296694635
    	step = 17----当前位置: w_current = 1.4788887903917065, b_current = 0.030089052745485803----计算w、b在当前位置的梯度: w' = -0.04759369133031832, b' = -0.6020382345426587----利用梯度下降更新参数后: new_w = 1.4788935497608395, new_b = 0.030149256568940068----参数更新后的loss: curr_loss = 112.65001658353181
    	step = 18----当前位置: w_current = 1.4788935497608395, b_current = 0.030149256568940068----计算w、b在当前位置的梯度: w' = -0.01798792504891189, b' = -0.6014518052574411----利用梯度下降更新参数后: new_w = 1.4788953485533445, new_b = 0.030209401749465813----参数更新后的loss: curr_loss = 112.64998039901862
    	step = 19----当前位置: w_current = 1.4788953485533445, b_current = 0.030209401749465813----计算w、b在当前位置的梯度: w' = -0.0031372638438682543, b' = -0.6011553831010357----利用梯度下降更新参数后: new_w = 1.478895662279729, new_b = 0.030269517287775917----参数更新后的loss: curr_loss = 112.64994426496074
    	step = 20----当前位置: w_current = 1.478895662279729, b_current = 0.030269517287775917----计算w、b在当前位置的梯度: w' = 0.004312010755272766, b' = -0.6010044329774986----利用梯度下降更新参数后: new_w = 1.4788952310786534, new_b = 0.030329617731073667----参数更新后的loss: curr_loss = 112.64990814400619
    	step = 21----当前位置: w_current = 1.4788952310786534, b_current = 0.030329617731073667----计算w、b在当前位置的梯度: w' = 0.008048636517292707, b' = -0.6009264538698962----利用梯度下降更新参数后: new_w = 1.4788944262150017, new_b = 0.030389710376460655----参数更新后的loss: curr_loss = 112.64987202675677
    	step = 22----当前位置: w_current = 1.4788944262150017, b_current = 0.030389710376460655----计算w、b在当前位置的梯度: w' = 0.009922940783500422, b' = -0.6008850781581111----利用梯度下降更新参数后: new_w = 1.4788934339209234, new_b = 0.030449798884276465----参数更新后的loss: curr_loss = 112.64983591084761
    	step = 23----当前位置: w_current = 1.4788934339209234, b_current = 0.030449798884276465----计算w、b在当前位置的梯度: w' = 0.010863076077107792, b' = -0.6008620632870942----利用梯度下降更新参数后: new_w = 1.4788923476133158, new_b = 0.030509885090605173----参数更新后的loss: curr_loss = 112.6497997956837
    	step = 24----当前位置: w_current = 1.4788923476133158, b_current = 0.030509885090605173----计算w、b在当前位置的梯度: w' = 0.011334617910296885, b' = -0.6008482585120101----利用梯度下降更新参数后: new_w = 1.4788912141515247, new_b = 0.030569969916456375----参数更新后的loss: curr_loss = 112.64976368111527
    	step = 25----当前位置: w_current = 1.4788912141515247, b_current = 0.030569969916456375----计算w、b在当前位置的梯度: w' = 0.011571106089602878, b' = -0.6008390736791128----利用梯度下降更新参数后: new_w = 1.4788900570409158, new_b = 0.030630053823824287----参数更新后的loss: curr_loss = 112.64972756710466
    	step = 26----当前位置: w_current = 1.4788900570409158, b_current = 0.030630053823824287----计算w、b在当前位置的梯度: w' = 0.011689687759941947, b' = -0.600832206296969----利用梯度下降更新参数后: new_w = 1.4788888880721398, new_b = 0.030690137044453985----参数更新后的loss: curr_loss = 112.64969145364233
    	step = 27----当前位置: w_current = 1.4788888880721398, b_current = 0.030690137044453985----计算w、b在当前位置的梯度: w' = 0.011749125718937492, b' = -0.6008265014006962----利用梯度下降更新参数后: new_w = 1.478887713159568, new_b = 0.030750219694594055----参数更新后的loss: curr_loss = 112.6496553407261
    	step = 28----当前位置: w_current = 1.478887713159568, b_current = 0.030750219694594055----计算w、b在当前位置的梯度: w' = 0.011778896287395924, b' = -0.600821379642184----利用梯度下降更新参数后: new_w = 1.4788865352699392, new_b = 0.030810301832558274----参数更新后的loss: curr_loss = 112.64961922835515
    	step = 29----当前位置: w_current = 1.4788865352699392, b_current = 0.030810301832558274----计算w、b在当前位置的梯度: w' = 0.01179378523940322, b' = -0.6008165504115521----利用梯度下降更新参数后: new_w = 1.4788853558914152, new_b = 0.03087038348759943----参数更新后的loss: curr_loss = 112.64958311652931
    	step = 30----当前位置: w_current = 1.4788853558914152, b_current = 0.03087038348759943----计算w、b在当前位置的梯度: w' = 0.011801209344453056, b' = -0.6008118679344496----利用梯度下降更新参数后: new_w = 1.4788841757704807, new_b = 0.030930464674392876----参数更新后的loss: curr_loss = 112.64954700524882
    	step = 31----当前位置: w_current = 1.4788841757704807, b_current = 0.030930464674392876----计算w、b在当前位置的梯度: w' = 0.011804888969175309, b' = -0.6008072590882101----利用梯度下降更新参数后: new_w = 1.4788829952815838, new_b = 0.030990545400301696----参数更新后的loss: curr_loss = 112.64951089451318
    	step = 32----当前位置: w_current = 1.4788829952815838, b_current = 0.030990545400301696----计算w、b在当前位置的梯度: w' = 0.011806690304899448, b' = -0.6008026871934107----利用梯度下降更新参数后: new_w = 1.4788818146125533, new_b = 0.03105062566902104----参数更新后的loss: curr_loss = 112.64947478432276
    	step = 33----当前位置: w_current = 1.4788818146125533, b_current = 0.03105062566902104----计算w、b在当前位置的梯度: w' = 0.011807549462947264, b' = -0.6007981338510664----利用梯度下降更新参数后: new_w = 1.478880633857607, new_b = 0.031110705482406145----参数更新后的loss: curr_loss = 112.64943867467751
    	step = 34----当前位置: w_current = 1.478880633857607, b_current = 0.031110705482406145----计算w、b在当前位置的梯度: w' = 0.011807936009969211, b' = -0.6007935898319836----利用梯度下降更新参数后: new_w = 1.4788794530640061, new_b = 0.031170784841389344----参数更新后的loss: curr_loss = 112.64940256557722
    	step = 35----当前位置: w_current = 1.4788794530640061, b_current = 0.031170784841389344----计算w、b在当前位置的梯度: w' = 0.011808085489754205, b' = -0.6007890505066291----利用梯度下降更新参数后: new_w = 1.4788782722554572, new_b = 0.031230863746440008----参数更新后的loss: curr_loss = 112.64936645702217
    	step = 36----当前位置: w_current = 1.4788782722554572, b_current = 0.031230863746440008----计算w、b在当前位置的梯度: w' = 0.011808116051717832, b' = -0.6007845135528076----利用梯度下降更新参数后: new_w = 1.478877091443852, new_b = 0.03129094219779529----参数更新后的loss: curr_loss = 112.64933034901209
    	step = 37----当前位置: w_current = 1.478877091443852, b_current = 0.03129094219779529----计算w、b在当前位置的梯度: w' = 0.011808086963425168, b' = -0.600779977805631----利用梯度下降更新参数后: new_w = 1.4788759106351557, new_b = 0.031351020195575854----参数更新后的loss: curr_loss = 112.64929424154703
    	step = 38----当前位置: w_current = 1.4788759106351557, b_current = 0.031351020195575854----计算w、b在当前位置的梯度: w' = 0.011808027953513545, b' = -0.6007754426807939----利用梯度下降更新参数后: new_w = 1.4788747298323603, new_b = 0.031411097739843935----参数更新后的loss: curr_loss = 112.64925813462713
    	step = 39----当前位置: w_current = 1.4788747298323603, b_current = 0.031411097739843935----计算w、b在当前位置的梯度: w' = 0.01180795393672085, b' = -0.6007709078851536----利用梯度下降更新参数后: new_w = 1.4788735490369667, new_b = 0.03147117483063245----参数更新后的loss: curr_loss = 112.64922202825228
    	step = 40----当前位置: w_current = 1.4788735490369667, b_current = 0.03147117483063245----计算w、b在当前位置的梯度: w' = 0.01180787238996217, b' = -0.6007663732717456----利用梯度下降更新参数后: new_w = 1.4788723682497276, new_b = 0.031531251467959626----参数更新后的loss: curr_loss = 112.6491859224223
    	step = 41----当前位置: w_current = 1.4788723682497276, b_current = 0.031531251467959626----计算w、b在当前位置的梯度: w' = 0.011807787067777298, b' = -0.600761838766785----利用梯度下降更新参数后: new_w = 1.4788711874710208, new_b = 0.0315913276518363----参数更新后的loss: curr_loss = 112.64914981713753
    	step = 42----当前位置: w_current = 1.4788711874710208, b_current = 0.0315913276518363----计算w、b在当前位置的梯度: w' = 0.011807699853244813, b' = -0.600757304333255----利用梯度下降更新参数后: new_w = 1.4788700067010354, new_b = 0.03165140338226963----参数更新后的loss: curr_loss = 112.64911371239778
    	step = 43----当前位置: w_current = 1.4788700067010354, b_current = 0.03165140338226963----计算w、b在当前位置的梯度: w' = 0.011807611687812525, b' = -0.6007527699526529----利用梯度下降更新参数后: new_w = 1.4788688259398666, new_b = 0.031711478659264894----参数更新后的loss: curr_loss = 112.64907760820297
    	step = 44----当前位置: w_current = 1.4788688259398666, b_current = 0.031711478659264894----计算w、b在当前位置的梯度: w' = 0.011807523046359236, b' = -0.6007482356156437----利用梯度下降更新参数后: new_w = 1.478867645187562, new_b = 0.03177155348282646----参数更新后的loss: curr_loss = 112.64904150455321
    	step = 45----当前位置: w_current = 1.478867645187562, b_current = 0.03177155348282646----计算w、b在当前位置的梯度: w' = 0.01180743416760599, b' = -0.6007437013175376----利用梯度下降更新参数后: new_w = 1.4788664644441452, new_b = 0.03183162785295821----参数更新后的loss: curr_loss = 112.64900540144839
    	step = 46----当前位置: w_current = 1.4788664644441452, b_current = 0.03183162785295821----计算w、b在当前位置的梯度: w' = 0.0118073451680214, b' = -0.6007391670560387----利用梯度下降更新参数后: new_w = 1.4788652837096283, new_b = 0.03189170176966381----参数更新后的loss: curr_loss = 112.64896929888867
    	step = 47----当前位置: w_current = 1.4788652837096283, b_current = 0.03189170176966381----计算w、b在当前位置的梯度: w' = 0.011807256110735409, b' = -0.6007346328299159----利用梯度下降更新参数后: new_w = 1.4788641029840173, new_b = 0.0319517752329468----参数更新后的loss: curr_loss = 112.64893319687393
    	step = 48----当前位置: w_current = 1.4788641029840173, b_current = 0.0319517752329468----计算w、b在当前位置的梯度: w' = 0.011807167022086063, b' = -0.6007300986386414----利用梯度下降更新参数后: new_w = 1.478862922267315, new_b = 0.03201184824281067----参数更新后的loss: curr_loss = 112.64889709540407
    	step = 49----当前位置: w_current = 1.478862922267315, b_current = 0.03201184824281067----计算w、b在当前位置的梯度: w' = 0.011807077921291764, b' = -0.6007255644818454----利用梯度下降更新参数后: new_w = 1.4788617415595229, new_b = 0.03207192079925886----参数更新后的loss: curr_loss = 112.64886099447925
    	step = 50----当前位置: w_current = 1.4788617415595229, b_current = 0.03207192079925886----计算w、b在当前位置的梯度: w' = 0.01180698881047526, b' = -0.6007210303594792----利用梯度下降更新参数后: new_w = 1.4788605608606418, new_b = 0.032131992902294806----参数更新后的loss: curr_loss = 112.64882489409928
    	step = 51----当前位置: w_current = 1.4788605608606418, b_current = 0.032131992902294806----计算w、b在当前位置的梯度: w' = 0.011806899698548534, b' = -0.6007164962713737----利用梯度下降更新参数后: new_w = 1.478859380170672, new_b = 0.032192064551921945----参数更新后的loss: curr_loss = 112.64878879426435
    	step = 52----当前位置: w_current = 1.478859380170672, b_current = 0.032192064551921945----计算w、b在当前位置的梯度: w' = 0.011806810585664351, b' = -0.6007119622175199----利用梯度下降更新参数后: new_w = 1.4788581994896135, new_b = 0.032252135748143694----参数更新后的loss: curr_loss = 112.6487526949743
    	step = 53----当前位置: w_current = 1.4788581994896135, b_current = 0.032252135748143694----计算w、b在当前位置的梯度: w' = 0.011806721471952386, b' = -0.6007074281979161----利用梯度下降更新参数后: new_w = 1.4788570188174663, new_b = 0.03231220649096349----参数更新后的loss: curr_loss = 112.64871659622933
    	step = 54----当前位置: w_current = 1.4788570188174663, b_current = 0.03231220649096349----计算w、b在当前位置的梯度: w' = 0.011806632358242197, b' = -0.6007028942125507----利用梯度下降更新参数后: new_w = 1.4788558381542305, new_b = 0.03237227678038474----参数更新后的loss: curr_loss = 112.6486804980292
    	step = 55----当前位置: w_current = 1.4788558381542305, b_current = 0.03237227678038474----计算w、b在当前位置的梯度: w' = 0.011806543244956558, b' = -0.6006983602614114----利用梯度下降更新参数后: new_w = 1.478854657499906, new_b = 0.03243234661641088----参数更新后的loss: curr_loss = 112.64864440037395
    	step = 56----当前位置: w_current = 1.478854657499906, b_current = 0.03243234661641088----计算w、b在当前位置的梯度: w' = 0.01180645413296233, b' = -0.6006938263444779----利用梯度下降更新参数后: new_w = 1.4788534768544928, new_b = 0.03249241599904533----参数更新后的loss: curr_loss = 112.64860830326369
    	step = 57----当前位置: w_current = 1.4788534768544928, b_current = 0.03249241599904533----计算w、b在当前位置的梯度: w' = 0.011806365020568421, b' = -0.6006892924617877----利用梯度下降更新参数后: new_w = 1.4788522962179906, new_b = 0.032552484928291506----参数更新后的loss: curr_loss = 112.64857220669826
    	step = 58----当前位置: w_current = 1.4788522962179906, b_current = 0.032552484928291506----计算w、b在当前位置的梯度: w' = 0.01180627590845873, b' = -0.6006847586133264----利用梯度下降更新参数后: new_w = 1.4788511155903998, new_b = 0.03261255340415284----参数更新后的loss: curr_loss = 112.64853611067775
    	step = 59----当前位置: w_current = 1.4788511155903998, b_current = 0.03261255340415284----计算w、b在当前位置的梯度: w' = 0.011806186798125395, b' = -0.6006802247990624----利用梯度下降更新参数后: new_w = 1.4788499349717201, new_b = 0.03267262142663274----参数更新后的loss: curr_loss = 112.64850001520203
    	step = 60----当前位置: w_current = 1.4788499349717201, b_current = 0.03267262142663274----计算w、b在当前位置的梯度: w' = 0.01180609768855767, b' = -0.6006756910190185----利用梯度下降更新参数后: new_w = 1.4788487543619513, new_b = 0.03273268899573464----参数更新后的loss: curr_loss = 112.64846392027134
    	step = 61----当前位置: w_current = 1.4788487543619513, b_current = 0.03273268899573464----计算w、b在当前位置的梯度: w' = 0.011806008579545946, b' = -0.6006711572731961----利用梯度下降更新参数后: new_w = 1.4788475737610933, new_b = 0.032792756111461964----参数更新后的loss: curr_loss = 112.64842782588546
    	step = 62----当前位置: w_current = 1.4788475737610933, b_current = 0.032792756111461964----计算w、b在当前位置的梯度: w' = 0.011805919470077697, b' = -0.600666623561613----利用梯度下降更新参数后: new_w = 1.4788463931691462, new_b = 0.032852822773818124----参数更新后的loss: curr_loss = 112.64839173204442
    	step = 63----当前位置: w_current = 1.4788463931691462, b_current = 0.032852822773818124----计算w、b在当前位置的梯度: w' = 0.011805830361772962, b' = -0.6006620898842412----利用梯度下降更新参数后: new_w = 1.47884521258611, new_b = 0.032912888982806546----参数更新后的loss: curr_loss = 112.64835563874827
    	step = 64----当前位置: w_current = 1.47884521258611, b_current = 0.032912888982806546----计算w、b在当前位置的梯度: w' = 0.011805741254191204, b' = -0.6006575562410894----利用梯度下降更新参数后: new_w = 1.4788440320119847, new_b = 0.032972954738430656----参数更新后的loss: curr_loss = 112.64831954599687
    	step = 65----当前位置: w_current = 1.4788440320119847, b_current = 0.032972954738430656----计算w、b在当前位置的梯度: w' = 0.011805652147730328, b' = -0.6006530226321438----利用梯度下降更新参数后: new_w = 1.47884285144677, new_b = 0.03303302004069387----参数更新后的loss: curr_loss = 112.64828345379041
    	step = 66----当前位置: w_current = 1.47884285144677, b_current = 0.03303302004069387----计算w、b在当前位置的梯度: w' = 0.01180556304211855, b' = -0.6006484890574175----利用梯度下降更新参数后: new_w = 1.4788416708904657, new_b = 0.03309308488959961----参数更新后的loss: curr_loss = 112.6482473621288
    	step = 67----当前位置: w_current = 1.4788416708904657, b_current = 0.03309308488959961----计算w、b在当前位置的梯度: w' = 0.011805473936737698, b' = -0.6006439555169159----利用梯度下降更新参数后: new_w = 1.478840490343072, new_b = 0.033153149285151305----参数更新后的loss: curr_loss = 112.64821127101195
    	step = 68----当前位置: w_current = 1.478840490343072, b_current = 0.033153149285151305----计算w、b在当前位置的梯度: w' = 0.011805384832294763, b' = -0.6006394220106258----利用梯度下降更新参数后: new_w = 1.4788393098045889, new_b = 0.033213213227352364----参数更新后的loss: curr_loss = 112.64817518043988
    	step = 69----当前位置: w_current = 1.4788393098045889, b_current = 0.033213213227352364----计算w、b在当前位置的梯度: w' = 0.011805295728413157, b' = -0.6006348885385582----利用梯度下降更新参数后: new_w = 1.4788381292750161, new_b = 0.033273276716206224----参数更新后的loss: curr_loss = 112.64813909041263
    	step = 70----当前位置: w_current = 1.4788381292750161, b_current = 0.033273276716206224----计算w、b在当前位置的梯度: w' = 0.011805206624387665, b' = -0.600630355100721----利用梯度下降更新参数后: new_w = 1.4788369487543538, new_b = 0.0333333397517163----参数更新后的loss: curr_loss = 112.64810300093023
    	step = 71----当前位置: w_current = 1.4788369487543538, b_current = 0.0333333397517163----计算w、b在当前位置的梯度: w' = 0.011805117522879272, b' = -0.60062582169707----利用梯度下降更新参数后: new_w = 1.4788357682426014, new_b = 0.033393402333886005----参数更新后的loss: curr_loss = 112.64806691199259
    	step = 72----当前位置: w_current = 1.4788357682426014, b_current = 0.033393402333886005----计算w、b在当前位置的梯度: w' = 0.011805028419654917, b' = -0.6006212883276769----利用梯度下降更新参数后: new_w = 1.4788345877397595, new_b = 0.033453464462718775----参数更新后的loss: curr_loss = 112.64803082359973
    	step = 73----当前位置: w_current = 1.4788345877397595, b_current = 0.033453464462718775----计算w、b在当前位置的梯度: w' = 0.011804939319056018, b' = -0.6006167549924678----利用梯度下降更新参数后: new_w = 1.4788334072458276, new_b = 0.03351352613821802----参数更新后的loss: curr_loss = 112.6479947357516
    	step = 74----当前位置: w_current = 1.4788334072458276, b_current = 0.03351352613821802----计算w、b在当前位置的梯度: w' = 0.01180485021769151, b' = -0.6006122216914986----利用梯度下降更新参数后: new_w = 1.4788322267608058, new_b = 0.03357358736038717----参数更新后的loss: curr_loss = 112.64795864844824
    	step = 75----当前位置: w_current = 1.4788322267608058, b_current = 0.03357358736038717----计算w、b在当前位置的梯度: w' = 0.011804761118309415, b' = -0.6006076884247233----利用梯度下降更新参数后: new_w = 1.478831046284694, new_b = 0.03363364812922964----参数更新后的loss: curr_loss = 112.64792256168965
    	step = 76----当前位置: w_current = 1.478831046284694, b_current = 0.03363364812922964----计算w、b在当前位置的梯度: w' = 0.011804672018252305, b' = -0.6006031551921853----利用梯度下降更新参数后: new_w = 1.4788298658174923, new_b = 0.03369370844474886----参数更新后的loss: curr_loss = 112.64788647547576
    	step = 77----当前位置: w_current = 1.4788298658174923, b_current = 0.03369370844474886----计算w、b在当前位置的梯度: w' = 0.011804582919999973, b' = -0.600598621993844----利用梯度下降更新参数后: new_w = 1.4788286853592003, new_b = 0.03375376830694824----参数更新后的loss: curr_loss = 112.64785038980665
    	step = 78----当前位置: w_current = 1.4788286853592003, b_current = 0.03375376830694824----计算w、b在当前位置的梯度: w' = 0.011804493822664242, b' = -0.6005940888297149----利用梯度下降更新参数后: new_w = 1.478827504909818, new_b = 0.03381382771583121----参数更新后的loss: curr_loss = 112.64781430468229
    	step = 79----当前位置: w_current = 1.478827504909818, b_current = 0.03381382771583121----计算w、b在当前位置的梯度: w' = 0.011804404724978568, b' = -0.6005895556998206----利用梯度下降更新参数后: new_w = 1.4788263244693456, new_b = 0.03387388667140119----参数更新后的loss: curr_loss = 112.64777822010262
    	step = 80----当前位置: w_current = 1.4788263244693456, b_current = 0.03387388667140119----计算w、b在当前位置的梯度: w' = 0.011804315628070938, b' = -0.600585022604137----利用梯度下降更新参数后: new_w = 1.4788251440377829, new_b = 0.033933945173661606----参数更新后的loss: curr_loss = 112.64774213606768
    	step = 81----当前位置: w_current = 1.4788251440377829, b_current = 0.033933945173661606----计算w、b在当前位置的梯度: w' = 0.011804226532328599, b' = -0.6005804895426585----利用梯度下降更新参数后: new_w = 1.4788239636151297, new_b = 0.03399400322261587----参数更新后的loss: curr_loss = 112.64770605257742
    	step = 82----当前位置: w_current = 1.4788239636151297, b_current = 0.03399400322261587----计算w、b在当前位置的梯度: w' = 0.011804137437495754, b' = -0.6005759565153928----利用梯度下降更新参数后: new_w = 1.478822783201386, new_b = 0.03405406081826741----参数更新后的loss: curr_loss = 112.64766996963193
    	step = 83----当前位置: w_current = 1.478822783201386, b_current = 0.03405406081826741----计算w、b在当前位置的梯度: w' = 0.011804048342812123, b' = -0.6005714235223479----利用梯度下降更新参数后: new_w = 1.4788216027965517, new_b = 0.034114117960619646----参数更新后的loss: curr_loss = 112.64763388723102
    	step = 84----当前位置: w_current = 1.4788216027965517, b_current = 0.034114117960619646----计算w、b在当前位置的梯度: w' = 0.011803959248018359, b' = -0.6005668905635348----利用梯度下降更新参数后: new_w = 1.478820422400627, new_b = 0.034174174649676----参数更新后的loss: curr_loss = 112.6475978053749
    	step = 85----当前位置: w_current = 1.478820422400627, b_current = 0.034174174649676----计算w、b在当前位置的梯度: w' = 0.011803870155858931, b' = -0.600562357638897----利用梯度下降更新参数后: new_w = 1.4788192420136113, new_b = 0.03423423088543989----参数更新后的loss: curr_loss = 112.64756172406332
    	step = 86----当前位置: w_current = 1.4788192420136113, b_current = 0.03423423088543989----计算w、b在当前位置的梯度: w' = 0.011803781062576846, b' = -0.6005578247485082----利用梯度下降更新参数后: new_w = 1.478818061635505, new_b = 0.034294286667914745----参数更新后的loss: curr_loss = 112.64752564329652
    	step = 87----当前位置: w_current = 1.478818061635505, b_current = 0.034294286667914745----计算w、b在当前位置的梯度: w' = 0.011803691970751373, b' = -0.600553291892314----利用梯度下降更新参数后: new_w = 1.478816881266308, new_b = 0.03435434199710398----参数更新后的loss: curr_loss = 112.6474895630743
    	step = 88----当前位置: w_current = 1.478816881266308, b_current = 0.03435434199710398----计算w、b在当前位置的梯度: w' = 0.011803602879250974, b' = -0.6005487590703438----利用梯度下降更新参数后: new_w = 1.47881570090602, new_b = 0.03441439687301101----参数更新后的loss: curr_loss = 112.6474534833968
    	step = 89----当前位置: w_current = 1.47881570090602, b_current = 0.03441439687301101----计算w、b在当前位置的梯度: w' = 0.01180351378936173, b' = -0.6005442262825647----利用梯度下降更新参数后: new_w = 1.4788145205546412, new_b = 0.034474451295639266----参数更新后的loss: curr_loss = 112.64741740426382
    	step = 90----当前位置: w_current = 1.4788145205546412, b_current = 0.034474451295639266----计算w、b在当前位置的梯度: w' = 0.011803424698806353, b' = -0.6005396935290223----利用梯度下降更新参数后: new_w = 1.4788133402121713, new_b = 0.03453450526499217----参数更新后的loss: curr_loss = 112.64738132567561
    	step = 91----当前位置: w_current = 1.4788133402121713, b_current = 0.03453450526499217----计算w、b在当前位置的梯度: w' = 0.011803335608854937, b' = -0.6005351608096938----利用梯度下降更新参数后: new_w = 1.4788121598786104, new_b = 0.034594558781073134----参数更新后的loss: curr_loss = 112.64734524763193
    	step = 92----当前位置: w_current = 1.4788121598786104, b_current = 0.034594558781073134----计算w、b在当前位置的梯度: w' = 0.011803246520578625, b' = -0.6005306281245604----利用梯度下降更新参数后: new_w = 1.4788109795539583, new_b = 0.03465461184388559----参数更新后的loss: curr_loss = 112.64730917013289
    	step = 93----当前位置: w_current = 1.4788109795539583, b_current = 0.03465461184388559----计算w、b在当前位置的梯度: w' = 0.011803157434105316, b' = -0.600526095473618----利用梯度下降更新参数后: new_w = 1.478809799238215, new_b = 0.03471466445343295----参数更新后的loss: curr_loss = 112.64727309317841
    	step = 94----当前位置: w_current = 1.478809799238215, b_current = 0.03471466445343295----计算w、b在当前位置的梯度: w' = 0.011803068346218026, b' = -0.6005215628569214----利用梯度下降更新参数后: new_w = 1.4788086189313803, new_b = 0.03477471660971864----参数更新后的loss: curr_loss = 112.64723701676857
    	step = 95----当前位置: w_current = 1.4788086189313803, b_current = 0.03477471660971864----计算w、b在当前位置的梯度: w' = 0.011802979259174506, b' = -0.6005170302744368----利用梯度下降更新参数后: new_w = 1.4788074386334544, new_b = 0.03483476831274608----参数更新后的loss: curr_loss = 112.6472009409034
    	step = 96----当前位置: w_current = 1.4788074386334544, b_current = 0.03483476831274608----计算w、b在当前位置的梯度: w' = 0.011802890174212877, b' = -0.6005124977261351----利用梯度下降更新参数后: new_w = 1.478806258344437, new_b = 0.03489481956251869----参数更新后的loss: curr_loss = 112.64716486558272
    	step = 97----当前位置: w_current = 1.478806258344437, b_current = 0.03489481956251869----计算w、b在当前位置的梯度: w' = 0.011802801088938608, b' = -0.6005079652120627----利用梯度下降更新参数后: new_w = 1.478805078064328, new_b = 0.0349548703590399----参数更新后的loss: curr_loss = 112.64712879080662
    	step = 98----当前位置: w_current = 1.478805078064328, b_current = 0.0349548703590399----计算w、b在当前位置的梯度: w' = 0.011802712003603943, b' = -0.6005034327322155----利用梯度下降更新参数后: new_w = 1.4788038977931277, new_b = 0.03501492070231312----参数更新后的loss: curr_loss = 112.64709271657507
    	step = 99----当前位置: w_current = 1.4788038977931277, b_current = 0.03501492070231312----计算w、b在当前位置的梯度: w' = 0.011802622919578454, b' = -0.6004989002865637----利用梯度下降更新参数后: new_w = 1.4788027175308358, new_b = 0.03507497059234178----参数更新后的loss: curr_loss = 112.64705664288809
    最终时的loss: final_loss = 112.64705664288809
    
    经过100轮迭代后: w = 1.4788027175308358, b = 0.03507497059234178, final_loss = 112.64705664288809
    
    Process finished with exit code 0

data.csv