专栏首页CreateAMindEpisodic memory 认知笔记

Episodic memory 认知笔记


Episodic memory is the memory of autobiographical events (times, places, associated emotions, and other contextual who, what, when, where, why knowledge) who what when 的抽象 有语义信息的整合记忆存储吗??还是序列的抽象记忆。

Episodic memory is association memory;

He was referring to the distinction between knowing and remembering. Knowing is more factual (semantic) whereas remembering is a feeling that is located in the past (episodic).[3]

three key properties of episodic memory recollection. These are a subjective sense of time (or mental time travel), connection to the self, and autonoetic consciousness. Autonoetic consciousness refers to a special kind of consciousness that accompanies the act of remembering which enables an individual to be aware of the self in a subjective time. others named the important aspects of recollection which includes visual imagery, narrative structure, retrieval of semantic information and the feelings of familiarity.[4]

Events that are recorded into episodic memory may trigger episodic learning, i.e. a change in behavior that occurs as a result of an event.[5][6] For example, a fear of dogs after being bitten by a dog is a result of episodic learning.

There are essentially nine properties of episodic memory that collectively distinguish it from other types of memory. Other types of memory may exhibit a few of these properties, but only episodic memory has all nine:[7]

  1. Contain summary records of sensory-perceptual-conceptual-affective processing. 多维度数据感知存储提取及方便联想提取。
  2. Retain patterns of activation/inhibition over long periods. 反复调用训练?
  3. Often represented in the form of (visual) images. 视觉压缩
  4. They always have a perspective (field or observer).
  5. Represent short time slices of experience. reward提取关键事件的一小段事件。
  6. They are represented on a temporal dimension roughly in order of occurrence.
  7. They are subject to rapid forgetting. reward相关程度对记忆的影响。
  8. They make autobiographical remembering specific.
  9. They are recollectively experienced when accessed. 可以反复拿来训练rl

3-4岁出现episode mem;


Relationship to semantic memory[edit]

Endel Tulving originally described episodic memory as a record of a person's experience that held temporally dated information and spatio-temporal relations.[11] A feature of episodic memory that Tulving later elaborates on is that it allows an agent to imagine traveling back in time.[12] A current situation may cue retrieval of a previous episode, so that context that colours the previous episode is experienced at the immediate moment. The agent is provided with a means of associating previous feelings with current situations. Semantic memory, on the other hand, is a structured record of facts, concepts, and skills that we have acquired. Semantic information is derived from accumulated episodic memory. Episodic memory can be thought of as a "map" that ties together items in semantic memory. For example, all encounters with how a "dog" looks and sounds will make up the semantic representation of that word. All episodic memories concerning a dog will then reference this single semantic representation of "dog" and, likewise, all new experiences with the dog will modify the single semantic representation of that dog.

Together, semantic and episodic memory make up our declarative memory.[13] They each represent different parts of context to form a complete picture. As such, something that affects episodic memory can also affect semantic memory. For example, anterograde amnesia, from damage of the medial temporal lobe, is an impairment of declarative memory that affects both episodic and semantic memory operations.[14] Originally, Tulving proposed that episodic and semantic memory were separate systems that competed with each other in retrieval. However, this theory was rejected when Howard and Kahana completed experiments on latent semantic analysis (LSA) that supported the opposite. Instead of an increase in semantic similarity when there was a decrease in the strength of temporal associations, the two worked together so semantic cues on retrieval were strongest when episodic cues were strong as well.[15]



The use of semantic memory is quite different from that of episodic memory. Semantic memory refers to general facts and meanings one shares with others whereas episodic memory refers to unique and concrete personal experiences

Recent research has focused on the idea that when people access a word's meaning, sensorimotor information that is used to perceive and act on the concrete object the word suggests is automatically activated. In the theory of grounded cognition, the meaning of a particular word is grounded in the sensorimotor systems.[12] For example, when one thinks of a pear, knowledge of grasping, chewing, sights, sounds, and tastes used to encode episodic experiences of a pear are recalled through sensorimotor simulation. A grounded simulation approach refers to context-specific re-activations that integrate the important features of episodic experience into a current depiction. Such research has challenged previously utilized amodal views. The brain encodes multiple inputs such as words and pictures to integrate and create a larger conceptual idea by using amodal views (also known as amodal perceptions).

Instead of being representations in modality-specific systems, semantic memory representations had previously been viewed as redescriptions of modality-specific states. Some accounts of category-specific semantic deficits that are amodal remain even though researchers are beginning to find support for theories in which knowledge is tied to modality-specific brain regions. This research defines a clear link between episodic experiences and semantic memory. The concept that semantic representations are grounded across modality-specific brain regions can be supported by the fact that episodic and semantic memory appear to function in different yet mutually dependent ways. The distinction between semantic and episodic memory has become a part of the broader scientific discourse. For example, it has been speculated that semantic memory captures the stable aspects of our personality while episodes of illness may have a more episodic nature.[13]



Associative models[edit]

The "association"—a relationship between two pieces of information—is a fundamental concept in psychology, and associations at various levels of mental representation are essential to models of memory and cognition in general. The set of associations among a collection of items in memory is equivalent to the links between nodes in a network, where each node corresponds to a unique item in memory. Indeed, neural networks and semantic networks may be characterized as associative models of cognition. However, associations are often more clearly represented as an N×N matrix, where N is the number of items in memory. Thus, each cell of the matrix corresponds to the strength of the association between the row item and the column item.

Learning of associations is generally believed to be a Hebbian process; that is, whenever two items in memory are simultaneously active, the association between them grows stronger, and the more likely either item is to activate the other. See below for specific operationalizations of associative models.


Amodal perception is the perception of the whole of a physical structure when only parts of it affect the sensory receptors. For example, a table will be perceived as a complete volumetric structure even if only part of it—the facing surface—projects to the retina; it is perceived as possessing internal volume and hidden rear surfaces despite the fact that only the near surfaces are exposed to view.

https://en.wikipedia.org/wiki/Implicit_memory 内隐记忆更多是动作控制等技能的学习; model-free很像是内隐记忆能力的学习;


本文分享自微信公众号 - CreateAMind(createamind)

原文出处及转载信息见文内详细说明,如有侵权,请联系 yunjia_community@tencent.com 删除。




0 条评论
登录 后参与评论


  • SPN 图片可视化 代码 应用 model 等介绍


  • 4篇前沿强化学习论文


  • Learning to Remember More with Less Memorization

    (Submitted on 5 Jan 2019 (v1), last revised 20 Mar 2019 (this version, v2))

  • 大脑记忆的建模


  • 【译】Python中的数据清洗 |Pythonic Data Cleaning With NumPy and Pandas(一)

    python中的数据清洗 | Pythonic Data Cleaning With NumPy and Pandas[1]

  • 语音表征的视觉引导自监督学习(Multimedia)


  • 语音表征的视觉引导自监督学习(Multimedia)


  • 涡轮叶片设计的THB样条近似与局部B样条近似(CS NA)

    我们考虑基于两阶段方法的工业模型的自适应重构,采用带有截短的分层B样条(THB样条)的自适应分散数据拟合方案。 该方案的第一阶段致力于计算局部最小二乘B样条曲线...

  • 我们需要联系追踪应用程序吗?(CS CS)


  • Swift 读标准库源码笔记 -- Integers(基本数据类型篇)

    版权声明:本文为博主原创文章,遵循 CC 4.0 BY-SA 版权协议,转载请附上原文出处链接和本声明。 ...