2.26 美國科學人人工智慧選段精讀
PREDICTION MACHINE
可以進行預測的機器
關鍵詞:
prediction:
預測
文中還有forecast
computational neuroscience:
計算神經學
neuroscience是一個合成詞,將neural和science組合
computer algorithms:
計算機演算法(雲演算法就可以用這個詞)
人工智慧中,演算法用algorithm,計算用compute
選句精讀
The nerve cells in the eye process basic features of an image before transferring this information to higher-level regions thatinterpretthe overall meaning of a scene.
interpret:
說明,解釋
eg. The whole speech might well be interpreted as a coded message to the Americans (整個演講完全可以看成是在向美國人傳遞編碼信息)
句式結構:
The nerve cells process basic features.....
注意:
1 指代:this 一般就近指代,因此this information指代的是前文basic features of an image
2 before後跟隨的獨立主格結構,獨立主格結構使用時需要該從句和主句主語一致,因此原句還原 before the nerve cells transfer....
意思:
眼睛裡的神經細胞處理基本的視覺影像,然後將這些影像信息傳輸給更高一級的區域解讀這些影像的總體含義。
Crucially, the downward signals from the higher levels of the brain continuallyinteract withthe 「upward」 signals from the senses, generating a prediction error: the difference between what we expect and what we experience.
interact:
互動
inter-:前綴,表示相互之間的意思
eg.Internet表示網之間,international表示國家之間
句式結構:
the downward signals...interact with.....
注意:
1 該句定語修飾較多,要分清楚
I from the higher levels 修飾downward signals
IIfrom the senses修飾upward signals
2 generate a prediction error是狀語從句修飾前文
意思:
重要的是:從腦中高層級發出的向下信號持續與從感官發出的向上信號互動,會產生一個錯誤的預判:我們所期待的和我們所經歷的不一致。
In this computational experiment, published in 1999 in Nature Neuroscience, the researcherssimulatedneuronal connections in the visual cortex—complete with(包括) downward connections carrying forecasts and upward connections bringing sensory signals from the outside world.
complete with:
包括=include
simulate:
模仿,模擬
句式結構:
researchers simulated neuronal connections....
注意:
1 破折號後面是對前面neuronal connections in the visual cortex 的補充說明
意思:
這個計算實驗發表在1999年自然神經科學雜誌上,在這個實驗中,研究者們模擬視覺皮質的神經鏈接----包括帶有預測信息的向下的鏈接和帶有外部世界感官信息的向上的鏈接。
難度:
★★★☆
(句子較長,修飾較多,專業性強)
1 Our brains are constantly trying to predict the future—and updating their expectations to match reality. Say you encounter your neighbor』s cat for the first time. Knowing your owngregarious(社交的)puppy, you expect that the cat will also enjoy your caresses. When you reach over to pet the creature, however, it scratches you. You update your theory aboutcuddly-looking(看起來可愛)animals—surmising(猜測), perhaps, that the kitty will be friendlier if you bring it a treat. With goodies in hand, the cat indeed lets you stroke it without inflicting wounds. Next time you encounter a furryfeline(貓科動物), you offer atuna(金槍魚)tidbit(佳肴)before trying to touch it.
我們的腦子一直都在嘗試預測未來---並且為了與事實匹配而更新腦中的預期。比如說,你第一次遇見你鄰居家的貓。想到你自家的小狗狗,你預期這隻貓也會享受你的照顧。然而,當你伸手去摸時,這隻貓撓了你。你更新有關看起來可愛的動物的理論,你猜測,如果你先給它一點好處,小貓咪也許會更加友善。事實上,只要有好東西在手,貓咪會讓你碰它,而且不傷害你。下次,當你遇到一隻毛絨絨的貓咪時,你會先給它美味的金槍魚,然後嘗試觸碰它。
2 In this manner, the higher processing centers in the brain continually refine their internal models according to the signals received from the sensory organs. Take our visual systems, which are highly complex.The nerve cells in the eye process basic features of an image before transferring this information to higher-level regions that interpret the overall meaning of a scene.Intriguingly(有趣的), neural connections also run in the other direction: from high-level processing centers, such as areas in theparietal(頭頂骨的)ortemporal cortices(顳皮層), to low-level ones such as the primaryvisual cortex(視覺皮質)and thelateral geniculate nucleus(外側膝狀體核). Some neuroscientists believe that these 「downward」 connections carry the brain』s predictions to lower levels, influencing what we see.
在這個動作中,腦中更高一級的處理中心會根據感官器官接收到的信號,持續改善內在的模式。以我們視覺系統為例。視覺系統是一個極度複雜的系統。眼睛裡的神經細胞處理基本的視覺影像,然後將這些影像信息傳輸給更高一級的區域解讀這些影像的總體含義。有趣的是,神經鏈接在其他區域也運行:從高一級的處理中心,例如頭頂骨區域或者顳皮層,到低一級的處理中心,例如視覺皮質和外側膝狀體核。有些神經科學家相信:這些向下的鏈接將腦中的預測帶向低層級的控制中心,最後影響我們看到的東西。
3 Crucially,the downward signals from the higher levels of the brain continually interact with the 「upward」 signals from the senses, generating a prediction error: the difference between what we expect and what we experience.A signal conveying this discrepancy returns to the higher levels, helping to refine internal models and generating fresh guesses, in an unending loop. 「The prediction error signal drives the system toward estimates of what』s really out there,」 says Rajesh P. N. Rao, a computational neuroscientist at the University of Washington.
重要的是:從腦中高層級發出的向下信號持續與從感官發出的向上信號互動,會產生一個錯誤的預判:我們所期待的和我們所經歷的不一致。帶有這些不一致信息的信號回到高層級,幫助我們改進內在機制模式並且產生新的猜想,這是一個無限的循環。華盛頓大學計算神經科學家Rajesh P. N. Rao說:這樣錯誤預判的信號會讓系統趨向於評估我們之外真正存在了什麼?
4 While Rao was a doctoral student at the University of Rochester, he and his supervisor, computational neuroscientist Dana H. Ballard, now at the University of Texas at Austin, became the first to test such predictive coding in an artificial neural network. (A class of computer algorithms modeled on the human brain, neural networksincrementally(遞增)adapt internalparameters(參數)to generate the required output from a given input.)In this computational experiment, published in 1999 in Nature Neuroscience, the researchers simulated neuronal connections in the visual cortex—complete with(包括) downward connections carrying forecasts and upward connections bringing sensory signals from the outside world.After training the network using pictures of nature, they found that it could learn to recognize key features of an image, such as a zebra』s stripes.
當Roa還是羅切斯特大學的博士生時,他和他的導師:計算神經學家Dana H. Ballard(現在德州州立大學任教)成為第一個通過人工神經網路測試這一預測編碼的人。(有關人腦的一種計算機演算法模式,神經網路逐漸適應內在參數,從已給定的輸入得到需要的輸出)。這個計算實驗發表在1999年自然神經科學雜誌上,在這個實驗中,研究者們模擬視覺皮質的神經鏈接----包括帶有預測信息的向下的鏈接和帶有外部世界感官信息的向上的鏈接。通過實用自然圖片訓練網路後,研究者們發現,網路可以學會認知圖片中的關鍵特徵,比如斑馬的條紋。
文章框架:
第一段:通過切身例子引出身體機制如何進行預判
第二段:預判信息的運作流程
第三段:預判信息存在的不一致性
第四段:將預判機制運用到試驗中,發現網路可以被訓練
知識補充:
顳皮層:
下顳皮質是顳葉的下面部分。調節視覺分辨能力。
外側膝狀體:
視覺系統的中繼核。位於丘腦枕腹面、大腦腳外側和內側膝狀體 的外上方。由較大的背側核及較小的腹側核兩部分組成。外膝體是從視網膜等通過接收信息,直接投射到視覺皮層,對視覺信息加工信息。
複習:
gregarious
cuddly-looking
surmise
feline
tuna
tidbit
Intriguingly
parietal
temporal cortices
visual cortex
lateral geniculate nucleus
incrementally
parameters
interpret
interact with
simulate
complete with
TAG:英語學志 |