Mager's tips on instructional objectives This is a very simple page that consists of black and white text without any graphics. As is, the text on the page is rather small and difficult (for me, anyway) to read, so one may wish to enlarge it. The process of creating instructional objectives in this format is explained in a clear and straightforward way. Rating 5/5

Students should be taught the material but should also be set free in order to collaborate with peers as well as technology to “tinker” and figure out the answer to the problem on their own which promotes a student centered approach to learning

NEURAL READING COMPREHENSION AND BEYOND

DEEP LEARNING WITH CONVOLUTIONAL NEURAL NETWORKS FOR IMAGE RECOGNITION: STEP-BY-STEP PROCESS FROM PREPARATION TO GENERALIZATION

Deep Learning for Chatbots

EFFICIENT METHODS AND HARDWARE FOR DEEP LEARNING

Deep Compression" can reduce the model sizeby 18?to 49?without hurting the prediction accuracy. We also discovered that pruning and thesparsity constraint not only applies to model compression but also applies to regularization, andwe proposed dense-sparse-dense training (DSD), which can improve the prediction accuracy for awide range of deep learning models. To efficiently implement "Deep Compression" in hardware,we developed EIE, the "Efficient Inference Engine", a domain-specific hardware accelerator thatperforms inference directly on the compressed model which significantly saves memory bandwidth.Taking advantage of the compressed model, and being able to deal with the irregular computationpattern efficiently, EIE improves the speed by 13?and energy efficiency by 3,400?over GPU

深度强化学习综述

深度文本匹配综述

深度强化学习综述

all kinds of text classification models and more with deep learning

reinforcement-learning code and paper tutorials

A Sensitivity Analysis of (and Practitioners’ Guide to) Convolutional Neural Networks for Sentence Classification

Differentiated Attentive Representation Learning for Sentence Classification

A Gentle Tutorial of Recurrent Neural Network with ErrorBackpropagation

A Tutorial on Deep Latent Variable Models of Natural Language

目前并没人来用attention来做机器翻译

LSTM Derivations

博客很赞！

batch normalization的出现是为了解决神经元的输入和当前计算值交互的高度依赖的问题。因为要计算期望值，所以需要拿到所有样本然后进行计算，显然不太现实。因此将取样范围和训练时的mini-batch保持一致。但是这就把局限转移到mini-batch的大小上了，很难应用到RNN。因此需要LayerNormalization.

Layer Normalization

NEURAL MACHINE TRANSLATION BY JOINTLY LEARNING TO ALIGN AND TRANSLATE

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

A STRUCTURED SELF-ATTENTIVE SENTENCE EMBEDDING

Contextual Word Representations: A Contextual Introduction

Deep Learning for Dialogue Systems

策略学习也是对话过程很重要的一环。 最近的策略学习过程有用基于有监督的预训练然后线上强化学习再训练的来提高学习的方案。但是这种方案有个潜在的毛病，在离线的数据中受限于数据量，线上一旦碰到了不常见的情况，容易直接恢复不来。(这个问题应该只是推断吧？有什么实证么？)

所以本文其实想说的是用一种方法来减轻线上和离线的差距。

Wait, this already happened?!

In education, learning data and analytics?

This reminds me of "personalized learning."

Do we capture this data? Can we?

How is this defined?

These apply to social annotation as well.

Are we capturing the right data/how can Hypothesis contribute to this profile?

Can H itself capture some of this data? Through the LMS?

Indeed. Just as it requires knowledge and skill to design, develop, and deliver an online learning experience, the same applies to blended learning. Making informed decisions about how to blend is deep, deep, deep.

Yeah, this is scary.

Wen和Bordes介绍了一种基于网络的端到端的任务型对话系统，这个系统将对话系统学习看成是从历史回话到系统回复的映射关系的学习问题，并且应用了一个编码解码器来训练整个系统。

这个思路很有意思，和我之前想构建一个电销员的语料库来做用户回复响应很像。这个很有可能。

I'm so glad I read Dave's post after having just read Rob Horning's great post, "The Sea Was Not a Mask", also addressing algorithms and YouTube.

Great discussion of what questions to ask about artificial intelligence and learning from Dave Cormier.

基于模型参数的迁移学习对 proformace 影响不大，当然训练更快啦。有趣的是，迁移 trained 模型参数的均值/方差统计性也可以得到不错的迁移效果。

Students can help each other learn by collaborating their efforts. Each student can bring a certain strength to the group so that they can all work out problems together

The importance of understanding the full extent of the content we are teaching is to give our students correct information. Learning incorrect information and having "misconceptions about the content area" is detrimental to our students' learning.

Helping students to have a relatable interest with their learning can help them to succeed in their futures. Making our lessons more understandable and related to their interests is important when setting up their learning environment.

When we start off the year using different teaching methods and establishing healthy relationships with our students, we can help them to grow immensely in the small amount of time that we know them.

Providing our students with opportunities to learn outside the classroom or using technology as a tool when we are teaching are good ways to get them involved in their learning and can eventually help them to take control of their learning experience all together, with us being the facilitators of knowledge.

When we understand our student's interests and build relationships with them, we can help them to love learning. We can do this collectively with our classes and we can do it with each student individually.

MXNet设计和实现简介

Recent Trends in Deep Learning BasedNatural Language Processing

GloVe: Global Vectors for Word Representation

Natural Language Processing (Almost) from Scratch

Short Text Similarity with Word Embeddings

Attention Is All You Need

Distributed Representations of Sentences and Documents - Doc2Vec

Deep contextualized word representations

BERT: Pre-training of Deep Bidirectional Transformers forLanguage Understanding

dummy encoding

Learning through layered asociations.Learning as hyperlinking.

I think we feel this way now with the internet. But its really just a certain kind of knowing and a certain kind of knowledge. It also depends on how you choose to approach that knowledge.

In other words: culture. I'm pretty sure that Engelbart would agree with the statement that someone who could order a book from a library would likely not know the best way to find a nearby water source, as the right kind of aborigine would know. Collective intelligence is a monotonically increasing store of knowledge that is maintained through social learning -- not just social learning, but teaching. Many species engage in social learning, but humans are the only primates with visible sclera -- the whites of our eyeballs -- which enables even infants to track where their teacher/parent is looking. I think this function of culture is what Engelbart would call "C work"

A Activity: 'Business as Usual'. The organization's day to day core business activity, such as customer engagement and support, product development, R&D, marketing, sales, accounting, legal, manufacturing (if any), etc. Examples: Aerospace - all the activities involved in producing a plane; Congress - passing legislation; Medicine - researching a cure for disease; Education - teaching and mentoring students; Professional Societies - advancing a field or discipline; Initiatives or Nonprofits - advancing a cause.

B Activity: Improving how we do that. Improving how A work is done, asking 'How can we do this better?' Examples: adopting a new tool(s) or technique(s) for how we go about working together, pursuing leads, conducting research, designing, planning, understanding the customer, coordinating efforts, tracking issues, managing budgets, delivering internal services. Could be an individual introducing a new technique gleaned from reading, conferences, or networking with peers, or an internal initiative tasked with improving core capability within or across various A Activities.

C Activity: Improving how we improve. Improving how B work is done, asking 'How can we improve the way we improve?' Examples: improving effectiveness of B Activity teams in how they foster relations with their A Activity customers, collaborate to identify needs and opportunities, research, innovate, and implement available solutions, incorporate input, feedback, and lessons learned, run pilot projects, etc. Could be a B Activity individual learning about new techniques for innovation teams (reading, conferences, networking), or an initiative, innovation team or improvement community engaging with B Activity and other key stakeholders to implement new/improved capability for one or more B activities.

In other words, human culture, using language, artifacts, methodology, and training, bootstrapped collective intelligence; what Engelbart proposed, then was to apply C work to culture's bootstrapping capabilities.

To me, this is the most prescient of Engelbart's future visions, and the seed for future study of culture-technology co-evolution. I talked with Engelbart about this passage over the years and we agreed that although the power of the artifacts, from RAM to CPU speed to network bandwidth, had improved by the billionfold since 1962, the "softer" parts of the formula -- the language, methodology, and training -- have not advanced so much. Certainly language, training methods and pedagogy, and collaborative strategies have evolved with the growth and spread of digital media, but are still lagging. H/LAMT interests me even more today than it did thirty years ago because Engelbart unknowingly forecast the fundamental elements of what has come to be called cultural-biological co-evolution. I gave a TED talk in 2005, calling for an interdisciplinary study of human cooperation -- and obstacles to cooperation. It seems that in recent years an interdisciplinary understanding has begun to emerge. Joseph Henrich at Harvard, for one, in his recent book, The Secret of Our Success, noted:

Drawing insights from lost European Explorers, clever chimpanzees, hunter-gatherers, cultural neuroscience, ancient bones and the human genome, Henrich shows that it’s not our general intelligence, innate brain power, or specialized mental abilities that explain our success. Instead, it’s our collective brains, which arise from a combination of our ability to learn selectively from each and our sociality. Our collective brains, which often operate outside of any individual’s conscious awareness, gradually produce increasingly complex, nuanced and subtle technological, linguistic and social products over generations.

Tracking this back into the mist of our evolutionary past, and to the remote corners of the globe, Henrich shows how this non-genetic system of cultural inheritance has long driven human genetic evolution. By producing fire, cooking, water containers, tracking know-how, plant knowledge, words, hunting strategies and projectiles, culture-driven genetic evolution expanded our brains, shaped our anatomy and physiology, and influenced our psychology, making us into the world’s only living cultural species. Only by understanding cultural evolution, can we understand human genetic evolution.

Henrich, Boyd, and RIcherson wrote, about the social fundamentals that distinguish human culture's methods of evolving collective intelligence in The Origin and Evolution of Culture:

Surely, without punishment, language, technology, individual intelligence and inventiveness, ready establishment of reciprocal arrangements, prestige systems and solutions to games of coordination, our societies would take on a distinctly different cast. Thus, a major constraint on explanations of human sociality is its systemic structure

Executive capabilities/processes compared. Latter are tied to metacognition in some learning theories. Here tacit knowledge is included in capabilities and excluded from processes (i.e. as inferior or biasing?)- if I'm reading correctly.

Learning theory.

Absolutely thrilled to see Mezirow and transformative learning theory in the spotlight. Transformative learning theory happens to be my favorite educational theory.

What are some ways we who seem to "grok" participatory connected learning (or think we do) can make this concept more accessible to colleagues who lament the failure of "sit-and-get" faculty development/PD, but don't know what to do next? I was reminded of this a few days ago in a "mixed" meeting of faculty, staff, and administrators. We all meant well, but could have done better in planning some upcoming sessions that (we hope) will become a Community of Practice. I think a way to describe participatory culture in a room full of people who don't already know Henry Jenkins and Mimi Ito would help.

Knowing how to navigate the internet is very different than understanding the internet and internet uses. Understanding how to use the internet is essential to communication, sharing, and learning on the internet.

This explains my personal experience here: https://hyp.is/EIfMfivUEem7SFcAiWxUpA/epic.org/privacy/global_entry/default.html (Why someone who had Global Entry was flagged for a police incident before he applied for Global Entry).

Intelligence checks is very concerning here as it suggests pretty much what has already been leaked, that the US is running complex autonomous screening of all of this data all the time. This also opens up the possibility for discriminatory algorithms since most of these are probably rooted in machine learning techniques and the criminal justice system in the US today tends to be fairly biased towards certain groups of people to begin with.

This re-affirms the previous annotation that the set of training data for the intelligence checks the US runs on global entry data is biased towards certain groups of people.

good summary

Absolutely. I've also been thinking about this in terms of "learning out loud" openly online.

The ProfHacker article by Maha Bali and Steve Greenlaw explores this more concretely. Active learning for inclusion needs to be scaffolded in such a way that it does not reinforce the privilege of dominant cultures and personalities.

This paragraph mentions that employers are offering more training online, so having online course experience will benefit students once they enter the job market, What are some other potential benefits of students learning online?

The continuity between these new modular features and the rest of the DAW’s workflow probably has unexpected consequences. Before getting information about BWS3, one might have thought that the “Native Modular System” promised since the first version might still be an add-on. What the marketing copy around this “killer feature” makes clear, it’s the result of a very deliberate process from the start and it’ll make for a qualitatively different workflow.

Service-learning approaches - real-life application of skills gained in class to make society better.

In my opinion this is what education is all about child centered learning

MOOC for your learning pleasure.

Evokes LPP

从此paper了解到了叫 Active learning 的有趣概念，这似乎和自己设计的连续参数训练数据采样池很接近。。。。

这篇文章的主要工作是给出了一个在图像分类中关于训练样本重要性的研究，对于样本的重要度采用基于梯度的方法进行度量。文章的结论可能表明在深度学习中主动学习或许并不总是有效的。

I've been recently returning to descriptive processes and "looking" at student work (a la Carini) just to keep abreast of what youth are doing and creating today. It's always changing and I think we can support this honoring and respecting by spending time learning from the work itself. I also think it opens up the ideas that would support educators in creating the opportunities for students to write in multiple ways, for multiple purposes, etc.

Christina Puntel's piece on Looking with the Heart is one of my favorites that I return to/share time and time again: https://thecurrent.educatorinnovator.org/resource/looking-with-the-heart-celebrating-the-human-in-the-digital