Neural Networks: How They Think?

By Larry Hardesty via MIT News Office

By Larry Hardesty via MIT News Office

Image Attribute: An illustrative example of an artificial neural network showing nodes and the links between them. / Source: Electronics360

Image Attribute: An illustrative example of an artificial neural network showing nodes and the links between them. / Source: Electronics360

Artificial-intelligence research has been transformed by machine-learning systems called neural networks, which learn how to perform tasks by analyzing huge volumes of training data.

During training, a neural net continually readjusts thousands of internal parameters until it can reliably perform some task, such as identifying objects in digital images or translating text from one language to another. But on their own, the final values of those parameters say very little about how the neural net does what it does.

Understanding what neural networks are doing can help researchers improve their performance and transfer their insights to other applications, and computer scientists have recently developed some clever techniques for divining the computations of particular neural networks.

But, at the 2017 Conference on Empirical Methods on Natural Language Processing starting this week, researchers from MIT’s Computer Science and Artificial Intelligence Laboratory are presenting a new general-purpose technique for making sense of neural networks that are trained to perform natural-language-processing tasks, in which computers attempt to interpret freeform texts written in ordinary, or “natural,” language (as opposed to a structured language, such as a database-query language).

The technique applies to any system that takes text as input and produces strings of symbols as output, such as an automatic translator. And because its analysis results from varying inputs and examining the effects on outputs, it can work with online natural-language-processing services, without access to the underlying software.

In fact, the technique works with any black-box text-processing system, regardless of its internal machinery. In their experiments, the researchers show that the technique can identify idiosyncrasies in the work of human translators, too.

Theme and variations


The technique is analogous to one that has been used to analyze neural networks trained to perform computer vision tasks, such as object recognition. Software that systematically perturbs — or varies — different parts of an image and resubmits the image to an object recognizer can identify which image features lead to which classifications. But adapting that approach to natural language processing isn’t straightforward.

“What does it even mean to perturb a sentence semantically?” asks Tommi Jaakkola, the Thomas Siebel Professor of Electrical Engineering and Computer Science at MIT and one of the new paper’s two authors. “I can’t just do a simple randomization. And what you are predicting is now a more complex object, like a sentence, so what does it mean to give an explanation?”

Somewhat ironically, to generate test sentences to feed to black-box neural nets, Jaakkola and David Alvarez-Melis, an MIT graduate student in electrical engineering and computer science and first author on the new paper, use a black-box neural net.

They begin by training a network to both compress and decompress natural sentences — to create some intermediate, compact digital representation of the sentence and then try to re-expand it into its original form. During training, the encoder and decoder are evaluated simultaneously, according to how faithfully the decoder’s output matches the encoder’s input.

Neural nets are intrinsically probabilistic: An object-recognition system fed an image of a small dog, for instance, might conclude that the image has a 70 percent probability of representing a dog and a 25 percent probability of representing a cat. Similarly, Jaakkola and Alvarez-Melis’ sentence-compressing network supplies alternatives for each word in a decoded sentence, along with the probabilities that each alternative is correct.

Because the network naturally uses the co-occurrence of words to increase its decoding accuracy, its output probabilities define a cluster of semantically related sentences. For instance, if the encoded sentence is “She gasped in surprise,” the system might assign the alternatives “She squealed in surprise” or “She gasped in horror” as fairly high probabilities, but it would assign much lower probabilities to “She swam in surprise” or “She gasped in coffee.”

For any sentence, then, the system can generate a list of closely related sentences, which Jaakkola and Alvarez-Melis feed to a black-box natural-language processor. The result is a long list of input-output pairs, which the researchers’ algorithms can analyze to determine which changes to which inputs cause which changes to which outputs.

Test cases


The researchers applied their technique to three different set types of natural-language-processing system. One was a system that inferred words’ pronunciation; another was a set of translators, two automated and one human; and the third was a simple computer dialogue system, which attempts to supply plausible responses to arbitrary remarks or questions.

As might be expected, the analysis of the translation systems demonstrated strong dependencies between individual words in the input and output sequences. One of the more intriguing results of that analysis, however, was the identification of gender biases in the texts on which the machine translations systems were trained.

For instance, the nongendered English word “dancer” has two gendered translations in French, “danseur” and “danseuse.” The system translated the sentence “The dancer is charming” using the feminine: “la danseuse est charmante.” But the researchers’ analysis showed that the choice of the word “danseuse” was as heavily influenced by the word “charming” as it was by the word “dancer.” A different adjective might have resulted in a different translation of “dancer.”

The dialogue system, which was trained on pairs of lines from Hollywood movies, was intentionally underpowered. Although the training set was large, the network itself was too small to take advantage of it. 

“The other experiment we do is in flawed systems,” Alvarez-Melis explains. “If you have a black-box model that is not doing a good job, can you first use this kind of approach to identify the problems? A motivating application of this kind of interpretability is to fix systems, to improve systems, by understanding what they’re getting wrong and why.”

In this case, the researchers’ analyses showed that the dialogue system was frequently keying in on just a few words in an input phrase, which it was using to select a stock response — answering “I don’t know” to any sentence that began with a query word such as “who” or “what,” for example. 

Reprinted with permission of  MIT News
Name

-51,1,3D Technology,2,5G,9,Abkhazia,2,Academics,10,Accidents,20,Activism,1,Adani Group,4,ADB,12,ADIZ,1,Adults,1,Advertising,31,Advisory,2,Aerial Reconnaissance,13,Aerial Warfare,35,Aerospace,5,Afghanistan,88,Africa,111,Agile Methodology,2,Agriculture,20,AI Policy,1,Air Crash,10,Air Defence Identification Zone,1,Air Defense,6,Air Force,29,Air Pollution,1,Airbus,5,Aircraft Carriers,5,Aircraft Systems,5,Al Nusra,1,Al Qaida,4,Al Shabab,1,Alaska,1,ALBA,1,Albania,2,Algeria,3,Alibaba,1,American History,4,AmritaJash,10,Antarctic,1,Antarctica,1,Anthropology,7,Anti Narcotics,12,Anti Tank,1,Anti-Corruption,4,Anti-dumping,1,Anti-Piracy,2,Anti-Submarine,1,Anti-Terrorism Legislation,1,Antitrust,2,APEC,1,Apple,2,Applied Sciences,2,AQAP,2,Arab League,3,Architecture,2,Arctic,6,Argentina,7,Armenia,30,Army,3,Art,3,Artificial Intelligence,81,Artillery,2,Arunachal Pradesh,2,ASEAN,12,Asia,70,Asia Pacific,23,Assassination,2,Asset Management,1,Astrophysics,2,ATGM,1,Atmospheric Science,1,Atomic.Atom,1,Augmented Reality,7,Australia,56,Austria,1,Automation,13,Automotive,129,Autonomous Flight,2,Autonomous Vehicle,3,Aviation,63,AWACS,2,Awards,17,Azerbaijan,16,Azeri,1,B2B,1,Bahrain,9,Balance of Payments,2,Balance of Trade,3,Balkan,10,Balochistan,2,Baltic,3,Baluchistan,8,Bangladesh,28,Banking,52,Bankruptcy,2,Basel,1,Bashar Al Asad,1,Battery Technology,2,Bay of Bengal,5,BBC,2,Beijing,1,Belarus,3,Belgium,1,Belt Road Initiative,3,Beto O'Rourke,1,BFSI,1,Bhutan,13,Big Data,30,Big Tech,1,Bilateral Cooperation,19,BIMSTEC,1,Biography,1,Biotechnology,3,Birth,1,BISA,1,Bitcoin,9,Black Lives Matter,1,Black Money,3,Black Sea,2,Blockchain,32,Blood Diamonds,1,Bloomberg,1,Boeing,21,Boko Haram,7,Bolivia,6,Bomb,3,Bond Market,2,Book,11,Book Review,24,Border Conflicts,11,Border Control and Surveillance,7,Bosnia,1,Brand Management,14,Brazil,104,Brexit,22,BRI,5,BRICS,20,British,3,Broadcasting,16,Brunei,3,Brussels,1,Buddhism,1,Budget,4,Build Back Better,1,Bulgaria,1,Burma,2,Business & Economy,1192,C-UAS,1,California,5,Call for Proposals,1,Cambodia,7,Cameroon,1,Canada,54,Canadian Security Intelligence Service (CSIS),1,Carbon Economy,9,CAREC,1,Caribbean,9,CARICOM,1,Caspian Sea,2,Catalan,3,Catholic Church,1,Caucasus,9,CBRN,1,Cement,1,Central African Republic,1,Central Asia,81,Central Asian,3,Central Eastern Europe,47,Certification,1,Chad,2,Chanakya,1,Charity,2,Chatbots,2,Chemicals,7,Child Labor,1,Child Marriage,1,Children,4,Chile,10,China,575,Christianity,1,CIA,1,CIS,5,Citizenship,2,Civil Engineering,2,Civil Liberties,5,Civil Rights,2,Civil Society,5,Civil Unrest,1,Civilization,1,Clean Energy,5,Climate,66,Climate Change,22,Climate Finance,2,Clinical Research,3,Clinton,1,Cloud Computing,44,Coal,6,Coast Guard,3,Cocoa,1,Cognitive Computing,12,Cold War,5,Colombia,15,Commodities,4,Communication,11,Communism,3,Compliance,1,Computers,40,Computing,1,Conferences,1,Conflict,106,Conflict Diamonds,1,Conflict Resolution,48,Conflict Resources,1,Congo,1,Construction,5,Consumer Behavior,4,Consumer Price Index,4,COP26,4,COP28,1,Copper,2,Coronavirus,107,Corporate Communication,1,Corporate Governance,4,Corporate Social Responsibility,4,Corruption,4,Costa Rica,2,Counter Intelligence,15,Counter Terrorism,81,COVID,9,COVID Vaccine,6,CPEC,8,CPG,3,Credit,2,Credit Rating,1,Credit Score,1,Crimea,4,CRM,1,Croatia,2,Crypto Currency,16,Cryptography,1,CSTO,1,Cuba,7,Culture,5,Currency,8,Customer Relationship Management,1,Cyber Attack,7,Cyber Crime,2,Cyber Security & Warfare,115,Cybernetics,5,Cyberwarfare,16,Cyclone,1,Cyprus,5,Czech Republic,3,DACA,1,DARPA,3,Data,9,Data Analytics,36,Data Center,2,Data Science,2,Database,3,Daughter.Leslee,1,Davos,1,DEA,1,DeBeers,1,Debt,12,Decision Support System,5,Defense,12,Defense Deals,8,Deforestation,2,Deloitte,1,Democracy,22,Democrats,2,Demographic Studies,1,Demonetization,6,Denmark. F-35,1,Denuclearization,1,Diamonds,1,Digital,39,Digital Currency,1,Digital Economy,10,Digital Marketing,6,Digital Transformation,11,Diplomacy,14,Diplomatic Row,2,Disaster Management,4,Disinformation,2,Diversity & Inclusion,1,Djibouti,2,Documentary,3,Doklam,2,Dokolam,1,Dominica,2,Donald Trump,48,Donetsk,2,Dossier,2,Drones,14,E-Government,2,E-International Relations,1,Earning Reports,3,Earth Science,1,Earthquake,7,East Africa,2,East China Sea,9,eBook,1,ECB,1,eCommerce,11,Econometrics,2,Economic Justice,1,Economics,43,Economy,108,ECOWAS,2,Ecuador,3,Edge Computing,2,Editor's Opinion,51,Education,65,EFTA,1,Egypt,27,Election Disinformation,1,Elections,42,Electric Vehicle,14,Electricity,7,Electronics,8,Emerging Markets,1,Employment,19,Energy,316,Energy Policy,28,Energy Politics,27,Engineering,24,England,2,Enterprise Software Solutions,8,Entrepreneurship,15,Environment,47,ePayments,13,Epidemic,6,ESA,1,Ethiopia,3,Eulogy,4,Eurasia,3,Euro,6,Europe,13,European Union,229,EuroZone,5,Exchange-traded Funds,1,Exclusive,2,Exhibitions,2,Explosives,1,Export Import,6,F-35,6,Facebook,9,Fake News,3,Fallen,1,FARC,2,Farnborough. United Kingdom,2,FATF,1,FDI,5,Featured,1351,Federal Reserve,2,Fidel Castro,1,FIFA World Cup,1,Fiji,1,Finance,18,Financial Markets,58,Financial Planning,1,Financial Statement,2,Finland,5,Fintech,14,Fiscal Policy,14,Fishery,3,Five Eyes,1,Food Security,27,Forces,1,Forecasting,2,Foreign Policy,13,Forex,4,France,33,Free Market,1,Free Syrian Army,4,Free Trade Agreement,1,Freedom,3,Freedom of Press,1,Freedom of Speech,2,Frigate,1,FTC,1,Fujairah,97,Fund Management,1,Funding,22,Future,1,G20,10,G24,1,G7,4,Gaddafi,1,Gambia,2,Gaming,1,Garissa Attack,1,Gas Price,23,GATT,1,Gaza,13,GCC,11,GDP,13,GDPR,1,Gender Studies,2,Geneal Management,1,General Management,1,Generative AI,7,Genetics,1,Geo Politics,105,Geography,2,Geoint,14,Geopolitics,8,Georgia,11,Georgian,1,geospatial,9,Geothermal,2,Germany,67,Ghana,3,Gibratar,1,Gig economy,1,Global Perception,1,Global Trade,95,Global Warming,1,Global Water Crisis,11,Globalization,3,Gold,2,Google,20,Gorkhaland,1,Government,128,Government Analytics,1,GPS,1,Greater Asia,170,Greece,13,Green Bonds,1,Green Energy,3,Greenland,1,Gross Domestic Product,1,GST,1,Gujarat,6,Gulf of Tonkin,1,Gun Control,4,Hacking,4,Haiti,2,Hamas,10,Hasan,1,Health,8,Healthcare,72,Heatwave,1,Helicopter,12,Heliport,1,Hezbollah,3,High Altitude Warfare,1,High Speed Railway System,1,Hillary 2016,1,Hillary Clinton,1,Himalaya,1,Hinduism,2,Hindutva,4,History,10,Home Security,1,Honduras,2,Hong Kong,7,Horn of Africa,5,Housing,16,Houthi,12,Howitzer,1,Human Development,32,Human Resource Management,5,Human Rights,6,Humanitarian,3,Hungary,3,Hunger,3,Hydrocarbon,3,Hydrogen,4,IAEA,2,ICBM,1,Iceland,2,ICO,1,Identification,2,IDF,1,Imaging,2,IMEEC,2,IMF,76,Immigration,19,Impeachment,1,Imran Khan,1,Independent Media,72,India,653,India's,1,Indian Air Force,19,Indian Army,7,Indian Nationalism,1,Indian Navy,27,Indian Ocean,24,Indices,1,Indigenous rights,1,Indo-Pacific,4,Indonesia,19,IndraStra,1,Industrial Accidents,3,Industrial Automation,2,Industrial Safety,4,Inflation,10,Infographic,1,Information Leaks,1,Infrastructure,3,Innovations,22,Insider Trading,1,Insurance,3,Intellectual Property,3,Intelligence,5,Intelligence Analysis,8,Interest Rate,3,International Business,13,International Law,11,International Relations,9,Internet,53,Internet of Things,34,Interview,8,Intra-Government,5,Investigative Journalism,4,Investment,32,Investor Relations,1,iPhone,1,IPO,4,Iran,202,Iraq,54,IRGC,1,Iron & Steel,4,ISAF,1,ISIL,9,ISIS,33,Islam,12,Islamic Banking,1,Islamic State,86,Israel,142,ISRO,1,IT ITeS,136,Italy,10,Ivory Coast,1,Jabhat al-Nusra,1,Jack Ma,1,Jamaica,3,Japan,88,JASDF,1,Jihad,1,JMSDF,1,Joe Biden,8,Joint Strike Fighter,5,Jordan,7,Journalism,6,Judicial,4,Justice System,3,Kanchin,1,Kashmir,8,Kaspersky,1,Kazakhstan,25,Kenya,5,Khalistan,2,Kiev,1,Kindle,700,Knowledge Management,4,Korean Conflict,1,Kosovo,2,Kubernetes,1,Kurdistan,8,Kurds,10,Kuwait,7,Kyrgyzstan,9,Labor Laws,10,Labor Market,4,Land Reforms,3,Land Warfare,21,Languages,1,Laos,2,Large language models,1,Laser Defense Systems,1,Latin America,80,Law,6,Leadership,3,Lebanon,10,Legal,11,LGBTQ,2,Li Keqiang,1,Liberalism,1,Library Science,1,Libya,14,Liechtenstein,1,Lifestyle,1,Light Battle Tank,1,Linkedin,1,Littoral Warfare,2,Livelihood,3,Loans,9,Lockdown,1,Lone Wolf Attacks,2,Lugansk,2,Macedonia,1,Machine Learning,7,Madagascar,1,Mahmoud,1,Main Battle Tank,3,Malaysia,12,Maldives,13,Mali,7,Malware,2,Management Consulting,6,Manpower,1,Manto,1,Manufacturing,15,Marijuana,1,Marine Engineering,3,Maritime,50,Market Research,2,Marketing,38,Mars,2,Martech,10,Mass Media,29,Mass Shooting,1,Material Science,2,Mauritania,1,Mauritius,2,MDGs,1,Mechatronics,2,Media War,1,MediaWiki,1,Medicare,1,Mediterranean,12,MENA,6,Mental Health,4,Mercosur,2,Mergers and Acquisitions,18,Meta,2,Metadata,2,Metals,3,Mexico,13,Micro-finance,4,Microsoft,12,Migration,19,Mike Pence,1,Military,110,Military Exercise,10,Military-Industrial Complex,3,Mining,15,Missile Launching Facilities,6,Missile Systems,56,Mobile Apps,3,Mobile Communications,11,Mobility,4,Modi,8,Moldova,1,Monaco,1,Monetary Policy,6,Money Market,2,Mongolia,11,Monkeypox,1,Monsoon,1,Montreux Convention,1,Moon,4,Morocco,2,Morsi,1,Mortgage,3,Moscow,2,Motivation,1,Mozambique,1,Mubarak,1,Multilateralism,2,Mumbai,1,Muslim Brotherhood,2,Mutual Funds,1,Myanmar,30,NAFTA,3,NAM,2,Namibia,1,Nanotechnology,4,Narendra Modi,2,NASA,13,National Identification Card,1,National Security,5,Nationalism,2,NATO,33,Natural Disasters,14,Natural Gas,33,Natural Language Processing,1,Nauru,1,Naval Base,5,Naval Engineering,23,Naval Intelligence,2,Naval Postgraduate School,2,Naval Warfare,50,Navigation,2,Navy,23,NBC Warfare,2,NDC,1,Nearshoring,1,Negotiations,2,Nepal,12,Netflix,1,Neurosciences,7,New Delhi,4,New Normal,1,New York,5,New Zealand,7,News,1245,News Publishers,1,Newspaper,1,NFT,1,NGO,1,Nicaragua,1,Niger,3,Nigeria,10,Nikki Haley,1,Nirbhaya,1,Non Aligned Movement,1,Non Government Organization,4,Nonproliferation,2,North Africa,23,North America,53,North Korea,58,Norway,5,NSA,1,NSG,2,Nuclear,41,Nuclear Agreement,32,Nuclear Doctrine,2,Nuclear Energy,4,Nuclear Fussion,1,Nuclear Propulsion,2,Nuclear Security,47,Nuclear Submarine,1,NYSE,1,Obama,3,ObamaCare,2,OBOR,15,Ocean Engineering,1,Oceania,2,OECD,5,OFID,5,Oil & Gas,382,Oil Gas,7,Oil Price,73,Olympics,2,Oman,25,Omicron,1,Oncology,1,Online Education,5,Online Reputation Management,1,OPEC,129,Open Access,1,Open Journal Systems,1,Open Letter,1,Open Source,4,OpenAI,2,Operation Unified Protector,1,Operational Research,4,Opinion,687,Opinon Poll,1,Optical Communications,1,Pacific,5,Pakistan,181,Pakistan Air Force,3,Pakistan Army,1,Pakistan Navy,3,Palestine,24,Palm Oil,1,Pandemic,84,Papal,1,Paper,3,Papers,110,Papua New Guinea,2,Paracels,1,Partition,1,Partnership,1,Party Congress,1,Passport,1,Patents,2,PATRIOT Act,1,Peace Deal,6,Peacekeeping Mission,1,Pension,1,People Management,1,Persian Gulf,19,Peru,5,Petrochemicals,1,Petroleum,19,Pharmaceuticals,14,Philippines,17,Philosophy,2,Photos,3,Physics,1,Pipelines,5,PLA,2,PLAN,4,Plastic Industry,2,Poland,8,Polar,1,Policing,1,Policy,8,Policy Brief,6,Political Studies,1,Politics,52,Polynesia,3,Pope,1,Population,6,Portugal,1,Poverty,8,Power Transmission,6,President APJ Abdul Kalam,2,Presidential Election,30,Press Release,158,Prison System,1,Privacy,18,Private Equity,2,Private Military Contractors,2,Privatization,1,Programming,1,Project Management,4,Propaganda,5,Protests,11,Psychology,3,Public Policy,55,Public Relations,1,Public Safety,7,Publications,1,Publishing,7,Purchasing Managers' Index,1,Putin,7,Q&A,1,Qatar,114,QC/QA,1,Qods Force,1,Quad,1,Quantum Computing,3,Quantum Physics,4,Quarter Results,2,Racial Justice,2,RADAR,2,Rahul Guhathakurta,4,Railway,9,Raj,1,Ranking,4,Rape,1,RBI,1,RCEP,2,Real Estate,6,Recall,4,Recession,2,Red Sea,5,Referendum,5,Reforms,18,Refugee,23,Regional,4,Regulations,2,Rehabilitation,1,Religion & Spirituality,9,Renewable,17,Report,4,Reports,47,Repository,1,Republicans,3,Rescue Operation,1,Research,5,Research and Development,24,Restructuring,1,Retail,36,Revenue Management,1,Rice,1,Risk Management,5,Robotics,8,Rohingya,5,Romania,2,Royal Canadian Air Force,1,Rupee,1,Russia,310,Russian Navy,5,Saab,1,Saadat,1,SAARC,6,Safety,1,SAFTA,1,SAM,2,Samoa,1,Sanctions,5,SAR,1,SAT,1,Satellite,14,Saudi Arabia,129,Scandinavia,6,Science & Technology,392,Science Fiction,1,SCO,5,Scotland,6,Scud Missile,1,Sea Lanes of Communications,4,SEBI,3,Securities,2,Security,6,Semiconductor,18,Senate,4,Senegal,1,SEO,5,Serbia,4,Services Sector,1,Seychelles,2,SEZ,1,Shadow Bank,1,Shale Gas,4,Shanghai,1,Sharjah,12,Shia,6,Shinzo Abe,1,Shipping,10,Shutdown,2,Siachen,1,Sierra Leone,1,Signal Intelligence,1,Sikkim,5,Silicon Valley,1,Silk Route,6,Simulations,2,Sinai,1,Singapore,16,Situational Awareness,20,Small Modular Nuclear Reactors,1,Smart Cities,7,Social Media,1,Social Media Intelligence,40,Social Policy,40,Social Science,1,Social Security,1,Socialism,1,Soft Power,1,Software,7,Solar Energy,14,Somalia,5,South Africa,20,South America,46,South Asia,467,South China Sea,35,South East Asia,74,South Korea,58,South Sudan,4,Sovereign Wealth Funds,1,Soviet,2,Soviet Union,9,Space,46,Space Station,2,Spain,9,Special Forces,1,Sports,3,Sports Diplomacy,1,Spratlys,1,Sri Lanka,23,Stablecoin,1,Stamps,1,Startups,43,State of the Union,1,Statistics,1,STEM,1,Stephen Harper,1,Stock Markets,23,Storm,2,Strategy Games,5,Strike,1,Sub-Sahara,3,Submarine,16,Sudan,5,Sunni,6,Super computing,1,Supply Chain Management,47,Surveillance,13,Survey,5,Sustainable Development,18,Swami Vivekananda,1,Sweden,4,Switzerland,6,Syria,112,Taiwan,30,Tajikistan,12,Taliban,17,Tamar Gas Fields,1,Tamil,1,Tanzania,4,Tariff,4,Tata,3,Taxation,25,Tech Fest,1,Technology,13,Tel-Aviv,1,Telecom,24,Telematics,1,Territorial Disputes,1,Terrorism,77,Testing,2,Texas,3,Thailand,11,The Middle East,649,Think Tank,316,Tibet,3,TikTok,1,Tobacco,1,Tonga,1,Total Quality Management,2,Town Planning,2,TPP,2,Trade Agreements,14,Trade War,10,Trademarks,1,Trainging and Development,1,Transcaucasus,19,Transcript,4,Transpacific,2,Transportation,47,Travel and Tourism,13,Tsar,1,Tunisia,7,Turkey,74,Turkmenistan,10,U.S. Air Force,3,U.S. Dollar,2,UAE,139,UAV,23,UCAV,1,Udwains,1,Uganda,1,Ukraine,109,Ukraine War,21,Ummah,1,UNCLOS,7,Unemployment,1,UNESCO,1,UNHCR,1,UNIDO,2,United Kingdom,82,United Nations,28,United States,752,University and Colleges,4,Uranium,2,Urban Planning,10,US Army,12,US Army Aviation,1,US Congress,1,US FDA,1,US Navy,18,US Postal Service,1,US Senate,1,US Space Force,2,USA,16,USAF,21,USV,1,UUV,1,Uyghur,3,Uzbekistan,13,Valuation,1,Vatican,3,Vedant,1,Venezuela,19,Venture Capital,4,Vibrant Gujarat,1,Victim,1,Videogames,1,Vietnam,24,Virtual Reality,7,Vision 2030,1,VPN,1,Wahhabism,3,War,1,War Games,1,Warfare,1,Water,17,Water Politics,7,Weapons,11,Wearable,2,Weather,2,Webinar,1,WeChat,1,WEF,3,Welfare,1,West,2,West Africa,19,West Bengal,2,Western Sahara,2,White House,1,Whitepaper,2,WHO,3,Wholesale Price Index,1,Wikileaks,1,Wikipedia,2,Wildfire,1,Wildlife,3,Wind Energy,1,Windows,1,Wireless Security,1,Wisconsin,1,Women,10,Women's Right,11,Workers Union,1,Workshop,1,World Bank,35,World Economy,32,World Peace,10,World War I,1,World War II,3,WTO,6,Wyoming,1,Xi Jinping,9,Xinjiang,2,Yemen,28,Yevgeny Prigozhin,1,Zbigniew Brzezinski,1,Zimbabwe,2,
ltr
item
IndraStra Global: Neural Networks: How They Think?
Neural Networks: How They Think?
By Larry Hardesty via MIT News Office
https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjwi3Abp5y7jYUgd1bgSjkLNv8umcuyhWXUrCPZErsjrhsAG-PkhJrg2iY_jq-lA_unrg0LbXHQ0dVtr2FXTJEQ1MA3fTsDcZeybQ38iEpmlxnqiYM1HVmXu7dGakq_8T-RnFF4T9lA6ApF/s640/neural_network-fullsize.png.jpg
https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjwi3Abp5y7jYUgd1bgSjkLNv8umcuyhWXUrCPZErsjrhsAG-PkhJrg2iY_jq-lA_unrg0LbXHQ0dVtr2FXTJEQ1MA3fTsDcZeybQ38iEpmlxnqiYM1HVmXu7dGakq_8T-RnFF4T9lA6ApF/s72-c/neural_network-fullsize.png.jpg
IndraStra Global
https://www.indrastra.com/2017/10/Neural-Networks-How-they-think-003-10-2017-0031.html
https://www.indrastra.com/
https://www.indrastra.com/
https://www.indrastra.com/2017/10/Neural-Networks-How-they-think-003-10-2017-0031.html
true
1461303524738926686
UTF-8
Loaded All Posts Not found any posts VIEW ALL Readmore Reply Cancel reply Delete By Home PAGES POSTS View All RECOMMENDED FOR YOU LABEL ARCHIVE SEARCH ALL POSTS Not found any post match with your request Back Home Sunday Monday Tuesday Wednesday Thursday Friday Saturday Sun Mon Tue Wed Thu Fri Sat January February March April May June July August September October November December Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec just now 1 minute ago $$1$$ minutes ago 1 hour ago $$1$$ hours ago Yesterday $$1$$ days ago $$1$$ weeks ago more than 5 weeks ago Followers Follow THIS PREMIUM CONTENT IS LOCKED STEP 1: Share to a social network STEP 2: Click the link on your social network Copy All Code Select All Code All codes were copied to your clipboard Can not copy the codes / texts, please press [CTRL]+[C] (or CMD+C with Mac) to copy Table of Content