02 2020 档案
摘要：Waiting list: basic knowledge: http://adamlineberry.ai/vae-series/vae-code-experiments
阅读全文
摘要：Basic knowledge: 1. Class VS Instance; class Student(object): def __init__(self, name, score): self.__name = name self.__score = score def print_score
阅读全文
摘要：python中的装饰器分为两类：函数装饰器和类装饰器. here and here2
阅读全文
摘要：Problem: unsupervised anomaly detection Model: VAE-reEncoder VAE with two encoders and one decoder. They use bidirectional bow-tie LSTM for each part.
阅读全文
摘要：Problem: unsupervised anomaly detection for seasonal KPIs in web applications. Donut: an unsupervised anomaly detection algorithm based on VAE. Backgr
阅读全文
摘要：“pattern driven” (PD) is based on enumerating candidate patterns in a given solution space and picking out the ones with high fitness;
阅读全文
摘要：1. a heuristic function h(n): a heuristic value of n, that is the estimated cost of reaching goal from n. 2. How to design heuristics? properties of a
阅读全文
摘要：Problem: unsupervised clustering represent data in feature space; learn a non-linear mapping from data space X to feature space Z. Problem formulation
阅读全文
摘要：Problem: TSC, time series classification; Traditional TSC: find global similarities or local patterns/subsequence(shapelet). We extract statistical fe
阅读全文
摘要：From: 虽然今天的主题是早睡早起，但是细想就会发现这是关于时间管理、关于一个习惯怎么养成。 1. 早起的重要性 所以如果您今天明白了早起，那减肥、跑步……任何坚持您都能明白。早起不仅仅是一个习惯，而是一种生活，它跟一天的状态有关。它意味着幸福感、时间、精力、效率、精神状态。网上有早起不一定有科学
阅读全文
摘要：Problem: new loss Label: new loss; Abstract: A differentiable learning loss; Introduction: supervised learning: learn a mapping that links an input to
阅读全文
摘要：from here 论文Timeseries data mining（2012）中提出：时间序列数据挖掘包括7个基本任务和3个基础问题： 7 tasks: query by content clustering classification segmentation?? prediction ano
阅读全文
摘要：From: Yoshua Bengio Problem: time series forecasting. Supplementary knowledge: 1. what is meta-learning: https://www.zhihu.com/question/264595128 2. w
阅读全文
摘要：Problem: clustering A clustering network transforms the data into another space and then selects one of the clusters. Next, the autoencoder associated
阅读全文
摘要：1. NLP/IR/DM/ML Conference Deadlines(Updating) Two Principles of Deadlines:1. All deadlines converge on the same day—Deadline Day.2. Every day is Dead
阅读全文
摘要：Problem: how to cluster non-stationary multivariate time series. What are stationary time series How to consider both global features and local featur
阅读全文
摘要：1. uncertainty aleatoric uncertainty 偶然不确定性 epistemic uncertainty 认知不确定性 2. probability VS likelihood Pr(data|distribution); L(distribution|data); The
阅读全文
摘要：From Michael I. Jordan As with many phrases that cross over from technical academic fields into general circulation, there is significant misunderstan
阅读全文
摘要：Problem: Design problem parameters consist of the search space of your model. Scientists design experiments to gain insights into physical and social
阅读全文
摘要：probability VS likelihood: https://zhuanlan.zhihu.com/p/25768606 http://sdsy888.me/%E9%9A%8F%E7%AC%94-Writing/2018/%E4%BC%BC%E7%84%B6%EF%BC%88likeliho
阅读全文
摘要：Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. Before: a discr
阅读全文
摘要：From: KU Leuven; ESAT-STADIUS比利时鲁汶大学 ?? How to model real-world multidimensional time series? especially, when these are sporadically observed data. ?
阅读全文
摘要：What is a search problem: A solution to a search problem is a sequence of actions (a path) from s0 to a goal state. It is optimal if it has minimum su
阅读全文
摘要：Why: real-world data are typically noisy, enormous in volume, and may originate from a hodgepodge of heterogeneous sources. mean; median; mode(most co
阅读全文
摘要：docker:组装 使用Postgres数据库的web应用 登录到psql的方法： psql --host=localhost --dbname=database-name --username=postgres sudo -u postgres psql database-name docker
阅读全文
摘要：From Kaggle: Appliances Energy Prediction Energy consumption of the Netherlands International Energy Statistics Hourly Energy Consumption
阅读全文
摘要：Problem: time series forecasting Challenge: forecasting for non-stationary signals and multiple future steps prediction ?? how to deal with non-statio
阅读全文
摘要：金灿荣：中国人民大学 国际关系学院 副院长，著名国际关系问题专家 武汉疫情： 公共卫生的预警机制有很大的缺陷。 中国应对危机的能力很强，但是预防危机的能力不够好 郑永年 媒体更加公开一些 政府更加透明一些 有些地方干部的素质还是很差的。 中国人形成的恶习：吃野生动物 SARS时候，达菲是比较管用的，
阅读全文
摘要：由武汉疫情事件引发的关注中国政治制度，新闻等，开始感叹各种事物都是殊途同归，于是选择阅读毛'泽东选集。 关于纠正党内的错误思想 1. 打胜仗就骄傲，打败仗就消极。 2. 上级机关要明了下级机关的情况和群众生活的情况，成为正确指导的客观基础. 扩展到科研上，个人作为上级，领域知识作为群众，要通过大量粗
阅读全文
摘要：Key points: don't procrastinate; review the notes and renew the literature search; determine who your audience is - most likely reviewers, so get thei
阅读全文
摘要：DON'T DO THAT: read the article from the beginning to end; it's a waste of time! READ A PAPER IN TWO-PHASE PROCESS: Phase 1: surveying the article! to
阅读全文
摘要：创新的源头来自于思考，尤其是深度思考： 1. 读博过程必然会经历痛苦，思考，深度思考这么一个过程，其中思考是最重要的，尤其是深度思考。 思考之后才是创新。 2. 借用其他的知识弥补这个领域的知识，不简单的是应用，而是能应用的好并且发现新的东西。 Insight VS research Put ano
阅读全文
摘要：data stream;streampipe/nifi data flow:1. algorithm;;;; top-k pattern, motif and so on 2. implication. 3. performance studyquery on stream; hue.sciclou
阅读全文
摘要：Chapter 1 data mining is knowledge discovery from data; The knowledge discovery process is an iterative sequence of 7 steps: data cleaning: to remove
阅读全文
摘要：Problem: high-dimensional time series forecasting ?? what is "high-dimensional" time series forecasting? one dimension for each individual time-series
阅读全文
摘要：Problem: time series clustering TSC - unsupervised learning/ category information is not available. time-series clustering for anomaly detection/ patt
阅读全文
摘要：relational data. Neural collaborative filtering and recurrent recommender systems have been successful in modeling user-item relational data. However,
阅读全文
摘要：Problem: time series classification shallow RNNs: the first layer splits the input sequence and runs several independent RNNs. The second layer consum
阅读全文
摘要：Problem: time series classification shapelet-based method: two issues 1. for multi-class imbalanced classification tasks, these methods will ignore th
阅读全文
摘要：1. develop a strong publications record early, so do what you can to make that happen. 2. 粗读：abstract, introduction, conclusion. 90%的文章都不值得读 因为当你不具备评价
阅读全文
摘要：Problem: time series prediction The nonlinear autoregressive exogenous model: The Nonlinear autoregressive exogenous (NARX) model, which predicts the
阅读全文
摘要：From google institution; 1. Before this, DNN cannot be used to map sequences to sequences. In this paper, we propose a sequence learning that makes mi
阅读全文
摘要：From: University of Maryland encode time series as different types of images. reformulate features of time series as visual clues. three representatio
阅读全文
摘要：MQTT[1]消息队列遥测传输(Message Queuing Telemetry Transport) subscribe to a topic and publish messages on that topic. The core of the client library is the cl
阅读全文
PP: Deep r -th Root of Rank Supervised Joint Binary Embedding for Multivariate Time Series Retrieval
摘要：from: Dacheng Tao 悉尼大学 PROBLEM: time series retrieval: given the current multivariate time series segment, how to obtain its relevant time series segm
阅读全文
摘要：Problem: the important frequency information is lack of effective modelling. ?? what is frequency information in time series? and why other models don
阅读全文
摘要：FROM Amazon research Germany PROBLEM probabilistic forecasting: estimate the probability distribution of a time series in future. INTRODUCTION a globa
阅读全文
摘要：2017 NIPS, time series workshop traditional methods: ARIMA. Seq2Seq quantile forecast; RELATED WORK DeepAR, probabilistic forecasting with encoder-dec
阅读全文
摘要：data > knowledge Are all patterns interesting? No. only a small fraction of the patterns potentially generated would actually be of interest to a give
阅读全文
摘要：Problem: ?? mining relationships in time series data; A new class of relationships in time series data. traditional methods: discover pair-wise relati
阅读全文
摘要：第四篇 阅读的最终目标 第二十章 阅读的第四个层次：主题阅读 在做主题阅读时的要求： 1. 知道：知道应该读哪些书。 在面对如此庞大的相关资料时，我们要如何决定我们要研究的主题是什么呢。 如何判断属于同一个主题。矛盾，就是有些书不属于这个主题。 2. 阅读的层次是渐进累积的。较高层次的阅读也包括了较
阅读全文
摘要：第一章 阅读的活力与艺术 主动阅读 VS 被动阅读（新闻媒体） 作者与读者：投手与捕手的关系。 阅读的目标： 为获得资讯而读，以及为求得理解而读。 为获得咨讯而读，不会增加我们的理解力。比如阅读报纸，杂志等。 所谓吸收资讯，就只是某件事发生了。 所读东西比reader高上一截，能增进阅读者的理解力。
阅读全文
