因想比較『辨識率』的差異,特用 PC 訓練 Michael Nielsen 先生之『network.py』,得到如下結果︰
Epoch 0: 9098 / 10000 Epoch 1: 9257 / 10000 Epoch 2: 9300 / 10000 Epoch 3: 9296 / 10000 Epoch 4: 9323 / 10000 Epoch 5: 9414 / 10000 Epoch 6: 9431 / 10000 Epoch 7: 9442 / 10000 Epoch 8: 9442 / 10000 Epoch 9: 9447 / 10000 Epoch 10: 9459 / 10000 Epoch 11: 9478 / 10000 Epoch 12: 9473 / 10000 Epoch 13: 9483 / 10000 Epoch 14: 9505 / 10000 Epoch 15: 9516 / 10000 Epoch 16: 9506 / 10000 Epoch 17: 9498 / 10000 Epoch 18: 9507 / 10000 Epoch 19: 9505 / 10000 Epoch 20: 9504 / 10000 Epoch 21: 9506 / 10000 Epoch 22: 9513 / 10000 Epoch 23: 9505 / 10000 Epoch 24: 9517 / 10000 Epoch 25: 9497 / 10000 Epoch 26: 9512 / 10000 Epoch 27: 9507 / 10000 Epoch 28: 9540 / 10000 Epoch 29: 9499 / 10000
再將此『腦』打包後,『移植』到樹莓派 3 上︰
>>> import mnist_loader >>> training_data, validation_data, test_data = \ ... mnist_loader.load_data_wrapper() >>> import network >>> net = network.Network([784, 30, 10]) >>> npzfile = network.np.load("swb.npz") >>> npzfile.files ['s', 'b2', 'w2', 'w1', 'b1'] >>> net.weights[0] = npzfile["w1"] >>> net.weights[1] = npzfile["w2"] >>> net.biases[0] = npzfile["b1"] >>> net.biases[1] = npzfile["b2"] >>> net.evaluate(test_data=test_data) 9499 >>> net.feedforward(training_data[0][0]) array([[ 3.25975452e-04], [ 1.50559437e-07], [ 1.12470024e-08], [ 5.58044192e-02], [ 8.82495098e-09], [ 9.88381262e-01], [ 1.39075949e-09], [ 3.61043544e-08], [ 3.45088788e-09], [ 5.59930823e-08]]) >>> network.np.argmax(net.feedforward(training_data[0][0])) 5 >>>
果在意料之內耶!『辨識率』為 的乎!!
由於好奇那『腦』到底長怎樣的呢?心想何不就看看『weights』之圖象哩?
>>> net.weights[0].shape (30, 784) >>> net.weights[1].shape (10, 30) >>> import matplotlib.pyplot as plt >>> img = net.weights[0] >>> plt.imshow(img) <matplotlib.image.AxesImage object at 0x552a790> >>> plt.show()
>>> img1 = net.weights[1] >>> plt.imshow(img1) <matplotlib.image.AxesImage object at 0x5a71a50> >>> plt.show()
畢卡索名作《格爾尼卡》,1937年,收藏於馬德里索菲亞王后藝術中心
都顯得『太寫實』的嘞??!!
於是心思馳盪,想著『意識』可有『母語』?世界能否『溝通』?所謂『腦波傳意』到底所說何事啊???
PLOS ONE (originally PLoS ONE) is an open access peer-reviewed scientific journal published by the Public Library of Science (PLOS) since 2006. It covers primary research from any discipline within science and medicine. All submissions go through an internal and external pre-publication peer review, but are not excluded on the basis of lack of perceived importance or adherence to a scientific field. The PLOS ONE online platform employs a “publish first, judge later” methodology, with post-publication user discussion and rating features.
腦‧腦長距離意識交流
腦‧腦直接界面
『PLOS ONE』原稱 PLoS ONE 是一個開放使用的『同行評審』 peer-reviewed 之網路日誌,於二零零六年由『科學公眾圖書館』Public Library of Science 所發行。 PLOS ONE 之內容涵蓋『科學』與『醫藥學』的各種領域內的『基礎研究』。對於所有提交的文本,出版前會經過內外的『同行評審』後刊載,但是並不排斥那些覺得缺乏基本重要性或是與特定科學領域無涉的文章。它的線上平台憑藉著『讀者的討論與評等』之『出版后』機制,對文本採取『先出版,後評價』的方法學。
今年八月十九日,PLOS ONE 上發表了一篇《Conscious Brain-to-Brain Communication in Humans Using Non-Invasive Technologies》文章,藉著新進發展的『腦機界面』 BCI brain-computer interfaces,以及剛出現的『非侵入式機腦界面』 nCBI non-invasive computer-brain interfaces,將 BCI 得到的『腦電波編碼』,透過網際網路傳送,而後於另一端『解碼』,探討長距離『腦』對『腦』的『意識傳達』與『意念交流』之『可能性』。據聞『實驗結果』錯誤率僅有百分之十五!其後同一組團隊於十一月五日,又發表了一篇《A Direct Brain-to-Brain Interface in Humans》文章,更進一步的證實了『人與人』之間『腦對腦』的『意念交流』的『可行性』!!
由於作者對於『腦電波』 EEG ElectroEncephaloGraphy 與『腦磁波』 MEG MagnetoEncephaloGraphic 的『原理』與『機制』了解十分有限,故於此無法多作評論。然而『傳遞』的既然是『電磁波』,依其所說又可以用『編碼』和『解碼』遠距離的在人與人之間『傳送』。這樣說來古人所講的至親間之『心電感應』以及《楞嚴經》所言的『甚深禪定』中所生之『六神通』 ── 神足通、天耳通、他心通、宿命通、天眼通、漏盡通 ── 也許就可能是真的了!!
─── 摘自《水的生命!!下上》