밤에 쓰는 편지 밤에 쓰는 편지

1. Maximum likelihood (EM) Gaussian Mixture ModelGenerative model: we specify p(data|parameters)- The distribution that generated the data is a weighted sum of K Gaussians- Each of the K Gaussians has its own mean and variance:  µk, Σk- the likelihood for each data point is:To generate samples from this model (given the parameters) we could:1. Use some sampling method with the full probability d..

수식을 이해해도 다소 막연했던 알고리즘은 실제 구현을 해서 돌려보면, 와닿는 경우가 많다. 1. Gibbs Sampling automatic_samples = np.random.multivariate_normal([0,0],[[1,0.5],[0.5,1]], 10000) plt.scatter(automatic_samples[:,0], automatic_samples[:,1], s=0.9); samples = {'x':[1], 'y':[-1]} num_samples = 10000 for _ in range(num_samples): curr_y = samples['y'][-1] new_x = np.random.normal(curr_y/2, np.sqrt(3/4)) new_y = np.random.normal(..

※ 출처: https://medium.com/towards-data-science/brief-introduction-to-markov-chains-2c8cab9c98ab Introduction In 1998, Lawrence Page, Sergey Brin, Rajeev Motwani and Terry Winograd published “The PageRank Citation Ranking: Bringing Order to the Web”, an article in which they introduced the now famous PageRank algorithm at the origin of Google. A little bit more than two decades later, Google has b..

※ https://medium.com/towards-data-science/bayesian-inference-problem-mcmc-and-variational-inference-25a8aa9bce29 Introduction Bayesian inference is a major problem in statistics that is also encountered in many machine learning methods. For example, Gaussian mixture models, for classification, or Latent Dirichlet Allocation, for topic modelling, are both graphical models requiring to solve such ..

※ https://medium.com/@chloebee/the-em-algorithm-explained-52182dbb19d9 The expectation-Maximization algorithm (or EM, for short) is probably one of the most influential and widely used machine learning algorithms in the field. When I first came to learn about the EM algorithm, it is surprisingly difficult to find a tutorial that offers an intuitive explanation about what is attempts to achieve a..

※ 출처: 딥 러닝을 이용한 자연어 처리 입문 RNN을 이용한 인코더-디코더 하나의 RNN을 인코더, 또 다른 하나의 RNN을 디코더라는 모듈로 명명하고 두 개의 RNN을 연결해서 사용하는 인코더-디코더 구조는 주로 입력 문장과 출력 문장의 길이가 다를 경우에 사용하는데, 대표적 분야가 번역기나 텍스트 요약입니다. 영어 문장을 한국어 문장으로 번역한다고 하였을 때, 입력 문장인 영어 문장과 번역된 결과인 한국어 문장의 길이는 똑같을 필요가 없습니다. 텍스트 요약의 경우에는 출력 문장이 요약된 문장이므로 입력 문장보다는 당연히 길이가 짧을 것입니다. 시퀀스-투-시퀀스 (Sequence-to-Sequence, seq2seq) 시퀀스-투-시퀀스 (Sequence-to-Sequence, seq2seq)는 입력된..

※ 출처: https://nlp.seas.harvard.edu/annotated-transformer/#part-1-model-architecture The Transformer has been on a lot of people's minds over the last five years. This post presents an annotated version of the paper in the form of a line-by-line implementation. It reorders and deletes some sections from the original paper and adds comments throughout. This document itself is a working notebook, a..

출처: https://jalammar.github.io/visualizing-neural-machine-translation-mechanics-of-seq2seq-models-with-attention/ Sequence-to-sequence models are deep learning models that have achieved a lot of success in tasks like machine translation, text summarization, and image captioning. Google Translate started using such a model in production in late 2016. These models are explained in the two pioneeri..