[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)

本文標(biāo)題：從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)，如今留學(xué)的人越來越多，不論高中生、大學(xué)生還是讀研的學(xué)生，都想早日去留學(xué)接受好的教育，很多同學(xué)對(duì)美國名?？蒲?美國留學(xué)中介,美國留學(xué)條件,美國留學(xué)網(wǎng),美國留學(xué)申請(qǐng)，美國研究生留學(xué)的相關(guān)問題有所疑問，下面澳際小編整理了《[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)》，歡迎閱讀，如有疑問歡迎聯(lián)系我們的在線老師，進(jìn)行一對(duì)一答疑。

　　帝國理工學(xué)院, 1907年建立于英國倫敦，在國際學(xué)術(shù)界有著頂級(jí)聲望，在各類權(quán)威榜單中排名穩(wěn)居世界前十。又與劍橋大學(xué)、牛津大學(xué)、倫敦大學(xué)學(xué)院、倫敦政治經(jīng)濟(jì)學(xué)院并稱為“G5超級(jí)精英大學(xué)”，研究水平被公認(rèn)為英國大學(xué)的前五強(qiáng)之列，尤其以工程專業(yè)而著名。在帝國的相關(guān)人物中，共有14位諾貝爾獎(jiǎng)獲得者和3位菲爾茲獎(jiǎng)獲得者。

　　本文作者H同學(xué)，品學(xué)兼優(yōu)，目前是帝國理工學(xué)院，二年級(jí)學(xué)生。

　　H同學(xué)有更高的夢(mèng)想，希望本科畢業(yè)后能夠到美國最好的大學(xué)讀研究生。為此，參與了名?？蒲许?xiàng)目，增加學(xué)術(shù)背景，開拓視野，獲得真知。

　　本文是學(xué)生在美國大學(xué)科研學(xué)習(xí)結(jié)束后所寫的感受，供學(xué)生及家長(zhǎng)參考。

　　Final Report for the Summer Research

　　注：中文譯文是編輯進(jìn)行的整理，英語能力較好的學(xué)生和家長(zhǎng)建議直接閱讀英文。

　　In this summer, fortunately, I have a chance to do summer research in US. The lab which I have joined is called computer science artificial intelligence lab, which is the top research center for machine learning in the world.

　　幸運(yùn)的是，在這個(gè)夏天，我有機(jī)會(huì)在美國做暑期研究。我加入的實(shí)驗(yàn)室被稱為計(jì)算機(jī)科學(xué)人工智能實(shí)驗(yàn)室，它是世界上機(jī)器學(xué)習(xí)的頂尖研究中心。

　　In the first week, I have searched a large amount of the information in order to locate the topic which might attract my interests. Finally, I have decided to make the artificial intelligence for automatically playing the Texas Hold’em poker (like the famous alpha go) one of the difficulties is that how to choose the suitable models.

　　在確定主題之后，我首先要做的是使用OpenCV庫來進(jìn)行撲克檢測(cè)，但是，我?guī)缀鯖]有關(guān)于Python的編碼體驗(yàn)，所以我必須看YouTube上關(guān)于在Python中使用OpenCV的教程。經(jīng)過幾天的斗爭(zhēng)，我用特征檢測(cè)來捕捉撲克的數(shù)量和類型，修正檢測(cè)的準(zhǔn)確性相當(dāng)高。缺點(diǎn)之一是檢測(cè)真正需要黑色背景。

　　After the topic has been determined, first thing which I have to do is using the OpenCV library to make the poker detection, however, I have almost no coding experience on python, so I have to watch the tutorial on YouTube about using OpenCV in python. After few days of struggle, I used the feature detection to capture the number and type of pokers, the accuracy of correcting detection is pretty high. One of the disadvantages is the detection truly need the black background.

　　在第一周，我搜索了大量的信息，以便找到可能吸引我興趣的話題。最后，我決定讓人工智能自動(dòng)玩德克薩斯HORD撲克(像著名的Alpha Go)，其中一個(gè)難點(diǎn)是如何選擇合適的模型。

　　After the detection has been finished, I discussed with the Prof A, and, she gives me the advice of using the random forest or the decision trees to build the models. Then the major problems become the data collecting. Due to the data on the internet are hard to decode, and useful data for the whole battle were too less to use. This situation might lead to a large bias problem for the deep learning. There is one alternating method to find the data, which were building the game and played by the professional player and machine, like the development method of alpha go. This method seems possible, however, there were two difficulties which might not be overcome during the short time. One is there is no professional player who I knew at that time. The other is that I also had no experience of making the Texas Hold’em poker game, and there is almost no possible for me to build the game alone for three weeks.

　　經(jīng)檢測(cè)已經(jīng)完成，我所討論的A教授，而且，她給我用隨機(jī)森林或決策樹構(gòu)建模型的建議。主要問題是數(shù)據(jù)采集。由于互聯(lián)網(wǎng)上的數(shù)據(jù)很難解讀，為整個(gè)戰(zhàn)役中有用的數(shù)據(jù)太少用。這種情況可能為深入學(xué)習(xí)造成較大偏差問題。有一個(gè)交替的方法來找到數(shù)據(jù)，這是建筑的游戲，由職業(yè)球員和機(jī)器玩，如α方法去發(fā)展。這種方法似乎是可行的，但是，有兩個(gè)困難不可克服的短時(shí)間內(nèi)。一是沒有專業(yè)的球員，我知道在那個(gè)時(shí)間。另外，我也沒有讓德克薩斯撲克游戲體驗(yàn)，幾乎沒有可能對(duì)我來說3周建立單獨(dú)的游戲。

　　Unfortunately, I chose to give up the project of Texas Hold’em poker. Then I had tried to change my interests in the computer vision. At that time, one of the competitions on the Kaggle has attracted my sights. The competition is about the VR object recognition and relationships between objects. The previous one which has been building the mature techniques to get the right position and recognition by using the convolution neural networks, nevertheless, the relationships are quite interesting for me to do the research on this aspect.

　　不幸的是，我選擇放棄德克薩斯撲克項(xiàng)目。然后我試著改變我對(duì)計(jì)算機(jī)視覺的興趣。那時(shí)候，一個(gè)比賽的比賽吸引了我的目光。競(jìng)爭(zhēng)是關(guān)于VR對(duì)象識(shí)別和對(duì)象之間的關(guān)系。前一種是利用卷積神經(jīng)網(wǎng)絡(luò)建立成熟的技術(shù)來獲得正確的位置和識(shí)別，然而，這些關(guān)系對(duì)我來說是非常有趣的。

　　So, first I have to use a week to build the neural network and change the parameters of the model and spent almost other weeks to decide the models for relationship parts. The random forest might be quite suitable for this project. Finally, I have written the frame of the training models.

　　所以，首先，我必須用一個(gè)星期來建立神經(jīng)網(wǎng)絡(luò)，并改變模型的參數(shù)，并花了幾乎其他星期來決定模型的關(guān)系部分。隨機(jī)森林可能相當(dāng)適合這個(gè)項(xiàng)目。最后，我編寫了培訓(xùn)模式的框架。

　　To be conclude, one of the most significant things which I have learnt from the research experience is the mental of science, Prof. A has taught me a lot about not only about the academic knowledges but also the mental of hard working and thinking differently, braving creating is also another mental which I would always remind myself in the latter life.

　　總而言之，我從研究經(jīng)驗(yàn)中學(xué)到的最重要的東西之一是科學(xué)的精神，A教授不僅僅教會(huì)了我很多關(guān)于學(xué)術(shù)知識(shí)的知識(shí)，也教會(huì)了我努力工作和思考的精神，勇于創(chuàng)造也是另一回事。我在后一輩子總是會(huì)提醒自己的。

此外，如果您對(duì)移民、置業(yè)相關(guān)服務(wù)感興趣，您可手機(jī)訪問：澳際移民置業(yè)官網(wǎng)http://immi.aoji.cn/

通過上述內(nèi)容講解，希望能幫助同學(xué)們更好的了解[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)最新動(dòng)態(tài)。想要咨詢更多最新[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)留學(xué)相關(guān)問題，可隨時(shí)咨詢我們專業(yè)的出國留學(xué)高級(jí)顧問老師。溫馨提示：了解最新出國留學(xué)動(dòng)態(tài)，也可以掃描下方二維碼直接添加海外顧問老師微信。

以上就是有關(guān)[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)的相關(guān)內(nèi)容介紹，希望對(duì)您有所幫助，對(duì)此如果還有什么想要了解的，可以關(guān)注澳際留學(xué)網(wǎng)相關(guān)資訊或咨詢專業(yè)的顧問老師。

[背景提升]從“帝國理工”到美國名校，學(xué)霸的科研體會(huì)內(nèi)容來自互聯(lián)網(wǎng)不代表本網(wǎng)觀點(diǎn)，如果本網(wǎng)轉(zhuǎn)載的稿件涉及您的版權(quán)請(qǐng)發(fā)郵件至jinglanghua@aoji.cn，我們將第一時(shí)間本網(wǎng)將依照國家相關(guān)法律法規(guī)盡快妥善處理