Long short-term memory (LSTM)[1] is a type of recurrent neural network (RNN) aimed at dealing with the vanishing gradient problem[2] present in traditional RNNs. Its relative insensitivity to gap length is its advantage over other RNNs, hidden Markov models and other sequence learning methods. It aims to provide a short-term memory for RNN that can last thousands of timesteps, thus "long short-term memory".[1] The name is made in analogy with long-term memory and short-term memory and their relationship, studied by cognitive psychologists since early 20th century.
A common LSTM unit is composed of a cell, an input gate, an output gate[14] and a forget gate.[15] The cell remembers values over arbitrary time intervals and the three gates regulate the flow of information into and out of the cell. Forget gates decide what information to discard from the previous state by mapping the previous state and the current input to a value between 0 and 1. A (rounded) value of 1 means to keep the information, and a value of 0 means to discard it. Input gates decide which pieces of new information to store in the current cell state, using the same system as forget gates. Output gates control which pieces of information in the current cell state to output by assigning a value from 0 to 1 to the information, considering the previous and current states. Selectively outputting relevant information from the current state allows the LSTM network to maintain useful, long-term dependencies to make predictions, both in current and future time-steps.
In theory, classic RNNs can keep track of arbitrary long-term dependencies in the input sequences. The problem with classic RNNs is computational (or practical) in nature: when training a classic RNN using back-propagation, the long-term gradients which are back-propagated can "vanish", meaning they can tend to zero due to very small numbers creeping into the computations, causing the model to effectively stop learning. RNNs using LSTM units partially solve the vanishing gradient problem, because LSTM units allow gradients to also flow with little to no attenuation. However, LSTM networks can still suffer from the exploding gradient problem.[16]
The intuition behind the LSTM architecture is to create an additional module in a neural network that learns when to remember and when to forget pertinent information.[15] In other words, the network effectively learns which information might be needed later on in a sequence and when that information is no longer needed. For instance, in the context of natural language processing, the network can learn grammatical dependencies.[17] An LSTM might process the sentence "Dave, as a result of his controversial claims, is now a pariah" by remembering the (statistically likely) grammatical gender and number of the subject Dave, note that this information is pertinent for the pronoun his and note that this information is no longer important after the verb is.
In the equations below, the lowercase variables represent vectors. Matrices and contain, respectively, the weights of the input and recurrent connections, where the subscript can either be the input gate , output gate , the forget gate or the memory cell , depending on the activation being calculated. In this section, we are thus using a "vector notation". So, for example, is not just one unit of one LSTM cell, but contains LSTM cell's units.
The figure on the right is a graphical representation of an LSTM unit with peephole connections (i.e. a peephole LSTM).[18][19] Peephole connections allow the gates to access the constant error carousel (CEC), whose activation is the cell state.[18] is not used, is used instead in most places.
Each of the gates can be thought as a "standard" neuron in a feed-forward (or multi-layer) neural network: that is, they compute an activation (using an activation function) of a weighted sum. and represent the activations of respectively the input, output and forget gates, at time step .
The 3 exit arrows from the memory cell to the 3 gates and represent the peephole connections. These peephole connections actually denote the contributions of the activation of the memory cell at time step , i.e. the contribution of (and not , as the picture may suggest). In other words, the gates and calculate their activations at time step (i.e., respectively, and ) also considering the activation of the memory cell at time step , i.e. .
The single left-to-right arrow exiting the memory cell is not a peephole connection and denotes .
The little circles containing a symbol represent an element-wise multiplication between its inputs. The big circles containing an S-like curve represent the application of a differentiable function (like the sigmoid function) to a weighted sum.
An RNN using LSTM units can be trained in a supervised fashion on a set of training sequences, using an optimization algorithm like gradient descent combined with backpropagation through time to compute the gradients needed during the optimization process, in order to change each weight of the LSTM network in proportion to the derivative of the error (at the output layer of the LSTM network) with respect to corresponding weight.
A problem with using gradient descent for standard RNNs is that error gradients vanish exponentially quickly with the size of the time lag between important events. This is due to if the spectral radius of is smaller than 1.[2][21]
However, with LSTM units, when error values are back-propagated from the output layer, the error remains in the LSTM unit's cell. This "error carousel" continuously feeds error back to each of the LSTM unit's gates, until they learn to cut off the value.
Many applications use stacks of LSTM RNNs[22] and train them by connectionist temporal classification (CTC)[23] to find an RNN weight matrix that maximizes the probability of the label sequences in a training set, given the corresponding input sequences. CTC achieves both alignment and recognition.
Sometimes, it can be advantageous to train (parts of) an LSTM by neuroevolution[24] or by policy gradient methods, especially when there is no "teacher" (that is, training labels).
There have been several successful stories of training, in a non-supervised fashion, RNNs with LSTM units.
In 2018, Bill Gates called it a "huge milestone in advancing artificial intelligence" when bots developed by OpenAI were able to beat humans in the game of Dota 2.[11] OpenAI Five consists of five independent but coordinated neural networks. Each network is trained by a policy gradient method without supervising teacher and contains a single-layer, 1024-unit Long-Short-Term-Memory that sees the current game state and emits actions through several possible action heads.[11]
In 2018, OpenAI also trained a similar LSTM by policy gradients to control a human-like robot hand that manipulates physical objects with unprecedented dexterity.[10]
In 2019, DeepMind's program AlphaStar used a deep LSTM core to excel at the complex video game Starcraft II.[12] This was viewed as significant progress towards Artificial General Intelligence.[12]
1996: LSTM is published at NIPS'1996, a peer-reviewed conference.[14]
1997: The main LSTM paper is published in the journal Neural Computation.[1] By introducing Constant Error Carousel (CEC) units, LSTM deals with the vanishing gradient problem. The initial version of LSTM block included cells, input and output gates.[52]
1999:Felix Gers, Jürgen Schmidhuber, and Fred Cummins introduced the forget gate (also called "keep gate") into the LSTM architecture,[53]
enabling the LSTM to reset its own state.[52]
2000: Gers, Schmidhuber, and Cummins added peephole connections (connections from the cell to the gates) into the architecture.[18][19] Additionally, the output activation function was omitted.[52]
2001: Gers and Schmidhuber trained LSTM to learn languages unlearnable by traditional models such as Hidden Markov Models.[18][54]
Hochreiter et al. used LSTM for meta-learning (i.e. learning a learning algorithm).[55]
2004: First successful application of LSTM to speech Alex Graves et al.[56][54]
2005: Daan Wierstra, Faustino Gomez, and Schmidhuber trained LSTM by neuroevolution without a teacher.[24]
2006: Graves, Fernandez, Gomez, and Schmidhuber introduce a new error function for LSTM: Connectionist Temporal Classification (CTC) for simultaneous alignment and recognition of sequences.[23] CTC-trained LSTM led to breakthroughs in speech recognition.[26][57][58][59]
2007: Wierstra, Foerster, Peters, and Schmidhuber trained LSTM by policy gradients for reinforcement learning without a teacher.[60]
Hochreiter, Heuesel, and Obermayr applied LSTM to protein homology detection the field of biology.[36]
2009: An LSTM trained by CTC won the ICDAR connected handwriting recognition competition. Three such models were submitted by a team led by Alex Graves.[3] One was the most accurate model in the competition and another was the fastest.[61] This was the first time an RNN won international competitions.[54]
2013: Alex Graves, Abdel-rahman Mohamed, and Geoffrey Hinton used LSTM networks as a major component of a network that achieved a record 17.7% phoneme error rate on the classic TIMIT natural speech dataset.[27]
2014: Kyunghyun Cho et al. put forward a simplified variant of the forget gate LSTM[53] called Gated recurrent unit (GRU).[63]
2015: Google started using an LSTM trained by CTC for speech recognition on Google Voice.[57][58] According to the official blog post, the new model cut transcription errors by 49%.[64]
2015: Rupesh Kumar Srivastava, Klaus Greff, and Schmidhuber used LSTM principles[53] to create the Highway network, a feedforward neural network with hundreds of layers, much deeper than previous networks.[65][66][67] 7 months later, Kaiming He, Xiangyu Zhang; Shaoqing Ren, and Jian Sun won the ImageNet 2015 competition with an open-gated or gateless Highway network variant called Residual neural network.[68] This has become the most cited neural network of the 21st century.[67]
2016: Google started using an LSTM to suggest messages in the Allo conversation app.[69] In the same year, Google released the Google Neural Machine Translation system for Google Translate which used LSTMs to reduce translation errors by 60%.[6][70][71]
Microsoft reported reaching 94.9% recognition accuracy on the Switchboard corpus, incorporating a vocabulary of 165,000 words. The approach used "dialog session-based long-short-term memory".[59]
2018:OpenAI used LSTM trained by policy gradients to beat humans in the complex video game of Dota 2,[11] and to control a human-like robot hand that manipulates physical objects with unprecedented dexterity.[10][54]
2021: According to Google Scholar, in 2021, LSTM was cited over 16,000 times within a single year. This reflects applications of LSTM in many different fields including healthcare.[13]
2024: An evolution of LSTM called xLSTM is published by a team leaded by Sepp Hochreiter.[81][82]
^Li, Xiangang; Wu, Xihong (2014-10-15). "Constructing Long Short-Term Memory based Deep Recurrent Neural Networks for Large Vocabulary Speech Recognition". arXiv:1410.4281 [cs.CL].
^ abWu, Yonghui; Schuster, Mike; Chen, Zhifeng; Le, Quoc V.; Norouzi, Mohammad; Macherey, Wolfgang; Krikun, Maxim; Cao, Yuan; Gao, Qin (2016-09-26). "Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation". arXiv:1609.08144 [cs.CL].
^ abcMayer, H.; Gomez, F.; Wierstra, D.; Nagy, I.; Knoll, A.; Schmidhuber, J. (October 2006). "A System for Robotic Heart Surgery that Learns to Tie Knots Using Recurrent Neural Networks". 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 543–548. CiteSeerX10.1.1.218.3399. doi:10.1109/IROS.2006.282190. ISBN978-1-4244-0258-8. S2CID12284900.
^Fernández, Santiago; Graves, Alex; Schmidhuber, Jürgen (2007). "Sequence labelling in structured domains with hierarchical recurrent neural networks". Proc. 20th Int. Joint Conf. On Artificial Intelligence, Ijcai 2007: 774–779. CiteSeerX10.1.1.79.1887.
^ abGraves, Alex; Fernández, Santiago; Gomez, Faustino; Schmidhuber, Jürgen (2006). "Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks". In Proceedings of the International Conference on Machine Learning, ICML 2006: 369–376. CiteSeerX10.1.1.75.6306.
^Perez-Ortiz, J. A.; Gers, F. A.; Eck, D.; Schmidhuber, J. (2003). "Kalman filters improve LSTM network performance in problems unsolvable by traditional recurrent nets". Neural Networks. 16 (2): 241–250. CiteSeerX10.1.1.381.1992. doi:10.1016/s0893-6080(02)00219-8. PMID12628609.
^A. Graves, J. Schmidhuber. Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks. Advances in Neural Information Processing Systems 22, NIPS'22, pp 545–552, Vancouver, MIT Press, 2009.
^Baccouche, M.; Mamalet, F.; Wolf, C.; Garcia, C.; Baskurt, A. (2011). "Sequential Deep Learning for Human Action Recognition". In Salah, A. A.; Lepri, B. (eds.). 2nd International Workshop on Human Behavior Understanding (HBU). Lecture Notes in Computer Science. Vol. 7065. Amsterdam, Netherlands: Springer. pp. 29–39. doi:10.1007/978-3-642-25446-8_4. ISBN978-3-642-25445-1.
^Huang, Jie; Zhou, Wengang; Zhang, Qilin; Li, Houqiang; Li, Weiping (2018-01-30). "Video-based Sign Language Recognition without Temporal Segmentation". arXiv:1801.10111 [cs.CV].
^Thireou, T.; Reczko, M. (2007). "Bidirectional Long Short-Term Memory Networks for predicting the subcellular localization of eukaryotic proteins". IEEE/ACM Transactions on Computational Biology and Bioinformatics. 4 (3): 441–446. doi:10.1109/tcbb.2007.1015. PMID17666763. S2CID11787259.
^Malhotra, Pankaj; Vig, Lovekesh; Shroff, Gautam; Agarwal, Puneet (April 2015). "Long Short Term Memory Networks for Anomaly Detection in Time Series"(PDF). European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning — ESANN 2015. Archived from the original(PDF) on 2020-10-30. Retrieved 2018-02-21.
^Duan, Xuhuan; Wang, Le; Zhai, Changbo; Zheng, Nanning; Zhang, Qilin; Niu, Zhenxing; Hua, Gang (2018). "Joint Spatio-Temporal Action Localization in Untrimmed Videos with Per-Frame Segmentation". 2018 25th IEEE International Conference on Image Processing (ICIP). 25th IEEE International Conference on Image Processing (ICIP). pp. 918–922. doi:10.1109/icip.2018.8451692. ISBN978-1-4799-7061-2.
^Orsini, F.; Gastaldi, M.; Mantecchini, L.; Rossi, R. (2019). Neural networks trained with WiFi traces to predict airport passenger behavior. 6th International Conference on Models and Technologies for Intelligent Transportation Systems. Krakow: IEEE. arXiv:1910.14026. doi:10.1109/MTITS.2019.8883365. 8883365.
^Zhao, Z.; Chen, W.; Wu, X.; Chen, P.C.Y.; Liu, J. (2017). "LSTM network: A deep learning approach for Short-term traffic forecast". IET Intelligent Transport Systems. 11 (2): 68–75. doi:10.1049/iet-its.2016.0208. S2CID114567527.
^((Cite Abbey Martin, Andrew J. Hill, Konstantin M. Seiler & Mehala Balamurali (2023) Automatic excavator action recognition and localisation for untrimmed video using hybrid LSTM-Transformer networks, International Journal of Mining, Reclamation and Environment, DOI: 10.1080/17480930.2023.2290364))
^Mozer, Mike (1989). "A Focused Backpropagation Algorithm for Temporal Pattern Recognition". Complex Systems.
^ abcGers, Felix; Schmidhuber, Jürgen; Cummins, Fred (1999). "Learning to forget: Continual prediction with LSTM". 9th International Conference on Artificial Neural Networks: ICANN '99. Vol. 1999. pp. 850–855. doi:10.1049/cp:19991218. ISBN0-85296-721-7.
^ abcdefgSchmidhuber, Juergen (10 May 2021). "Deep Learning: Our Miraculous Year 1990-1991". arXiv:2005.05744 [cs.NE].
^Abbey Martin, Andrew J. Hill, Konstantin M. Seiler & Mehala Balamurali (2023) Automatic excavator action recognition and localisation for untrimmed video using hybrid LSTM-Transformer networks, International Journal of Mining, Reclamation and Environment, DOI: 10.1080/17480930.2023.2290364